Interpretation of spectroscopic data using molecular simulations for the secondary active transporter BetP

Abstract

Mechanistic understanding of dynamic membrane proteins such as transporters, receptors, and channels requires accurate depictions of conformational ensembles, and the manner in which they interchange as a function of environmental factors including substrates, lipids, and inhibitors. Spectroscopic techniques such as electron spin resonance (ESR) pulsed electron-electron double resonance (PELDOR), also known as double electron-electron resonance (DEER), provide a complement to atomistic structures obtained from x-ray crystallography or cryo-EM, since spectroscopic data reflect an ensemble and can be measured in more native solvents, unperturbed by a crystal lattice. However, attempts to interpret DEER data are frequently stymied by discrepancies with the structural data, which may arise due to differences in conditions, the dynamics of the protein, or the flexibility of the attached paramagnetic spin labels. Recently, molecular simulation techniques such as EBMetaD have been developed that create a conformational ensemble matching an experimental distance distribution while applying the minimal possible bias. Moreover, it has been proposed that the work required during an EBMetaD simulation to match an experimentally determined distribution could be used as a metric with which to assign conformational states to a given measurement. Here, we demonstrate the application of this concept for a sodium-coupled transport protein, BetP. Because the probe, protein, and lipid bilayer are all represented in atomic detail, the different contributions to the work, such as the extent of protein backbone movements, can be separated. This work therefore illustrates how ranking simulations based on EBMetaD can help to bridge the gap between structural and biophysical data and thereby enhance our understanding of membrane protein conformational mechanisms.

Figure 1.

Structure of BetP and spin-label locations. (A and B) The structure of BetP shown as cartoon helices, highlighting TM helices 1′, 5′, 6′, and 8′ lining the substrate pathway (blue) and helices 3′, 4′, 8′, and 9′ in the hash domain (pale blue). Helices −2, −1, and 7b (white) contribute to the trimer interface. Each BetP protein chain was labeled at two cysteines on the periplasmic surface, introduced at positions G450 and S516. (A) Trimer structure of BetP (PDB accession no. 4C7R). Cα atoms of labeled positions are shown as orange spheres. (B) Comparison of MMM-based predicted spin-label positions in structures of the two extreme conformations of monomeric BetP, outward-open (PDB accession no. 4LLH, chain A) or inward-open (PDB accession no. 4C7R, chain A), with nitroxide N atom positions colored red and gold, respectively. (C–F) Simulation system for spin-labeled monomeric BetP. Snapshots indicate the location of the spin labels with betaine bound (C and D) or without betaine (E and F), in either outward-open (C and E) or inward-open (D and F) conformations. The protein is viewed from within the plane of the membrane. The MTS labels and bound betaine are shown as sticks, and sodium ions are shown as blue spheres. Representative snapshots were selected from the center of a cluster of conformations in which the probe distance is at the peak of the distribution (∼42 Å or ∼27 Å for the betaine-bound or betaine-free systems, respectively). Water molecules within 10 Å of the ligand are shown as an orange surface, highlighting the pathways on the extracellular or intracellular sides. Helices 3′, 4′, 8′ (including G450C), and 9′, which comprise the so-called hash domain, are colored blue. Helices 1′, 5′, 6′, and 10′ (including S516C) are colored yellow. Lipids and surrounding solvent are omitted for clarity. (G and H) Sodium ion binding sites (Na1′ and Na2, labeled 1′ and 2, respectively) in the outward- (G) and inward-open (H) conformations of BetP in the presence of substrate, viewed from the extracellular side. The density occupied by each ion during the EBMetaD simulations is shown as orange mesh. Nearby water molecules are shown in stick representation.

Figure 2.

Measurements for BetP in E. coli polar lipid proteoliposomes with spin labels introduced at positions 450 and 516. (A) Uptake of betaine in nanomolar per milligram protein was measured at 0.6 osmol/kg as a function of time for WT and MTSSL-labeled BetP cysteine variant C252T/G450C/S516C reconstituted into E. coli polar lipid liposomes. Uptake was initiated by adding saturating concentrations of 50 µM [¹⁴C]-betaine. Each value is the mean ± SEM of at least six independent measurements. (B) PELDOR normalized echo amplitude (Norm. Echo Amp.) time traces and assumed background traces (thinner smooth lines) for MTSSL-labeled BetP measured in the presence of 500 mM NaCl (black) or 300 mM NaCl and 5 mM betaine (green). (C and D) Probability of a distance P(r) versus distance (r) between spins derived from Tikonov regularization of the PELDOR time traces in B, before (C) and after (D) subtraction of a Gaussian distribution with a peak at ∼60 Å (see Fig. S2).

Figure 3.

Convergence of simulated to experimental distance distributions. The probability of a distance P(r) is plotted versus distance (r). The PELDOR-based distances (black lines), measured in the presence of 500 mM NaCl (A) or 300 mM NaCl plus 5 mM betaine (B), are compared with distances obtained in 1-µs-long EBMetaD MD simulations, performed for BetP monomers in the presence of two sodium ions (A) or with two sodium ions plus a betaine substrate (B). The simulations were started with BetP structures of either outward-facing (PDB accession no. 4LLH chain A, red circles) or inward-facing (PDB accession no. 4C7R, blue circles) conformations.

Figure 4.

The EBMetaD bias does not change the overall conformation of the protein. (A–D) Structural similarity (in RMSD) of each simulated ensemble with respect to the two extreme conformations of BetP. EBMetaD simulation trajectories initiated with either outward-facing (A and B) or inward-facing (C and D) conformation in the presence of two sodium (A and C) or two sodium and betaine (B and D) are compared with either the initial structure (blue) or with the structure of the opposite state (orange).

Figure 5.

Work required by each simulation system to reproduce the experimental distance distribution. The work was computed by averaging the bias potential applied over the last 0.8 µs of each simulation of G450R5/S516R5. Simulations were performed with two bound sodium ions (2Na⁺), in either the absence (left) or presence (right) of betaine. The BetP conformation used to start the simulation was either inward- (PDB accession no. 4C7R, blue) or outward-facing (PDB accession no. 4LLH chain A, red). Error bars represent the SD of work values obtained by averaging over the two halves of the last 0.8 µs for the entire width of the distance distribution (A), or averaging the bias potential applied for distances <37 Å (B; see Materials and methods and Fig. S2), as expected for exclusively interprotomer interactions.

Figure 6.

Work required by each D470-protonated simulation system to reproduce the experimental distance distribution. The work was computed by averaging the bias potential applied over the last 0.8 µs of each simulation of G450R5/S516R5 with Asp470 protonated. Simulations were performed with two bound sodium ions (2Na⁺), in either the absence (left) or presence (right) of betaine. The BetP conformation used to start the simulation was either inward- (PDB accession no. 4C7R, blue) or outward-facing (PDB accession no. 4LLH chain A, red). Error bars represent the SD of work values obtained by averaging over the two halves of the last 0.8 µs for the entire width of the distance distribution (A), or averaging the bias potential applied for distances <37 Å (B; see Materials and methods), as expected for exclusively inter-protomer interactions.

Figure 7.

Breakdown of contributions to the work for each molecular simulation system. Work computed with the Kullback-Leibler divergence between the biased and debiased trajectories (see Materials and methods) for the distance between the Cα atoms of residues 450 and 516 (A); the relative orientation of the spin labels (B); the side-chain dihedral angles χ₁ to χ₅ (C) or only χ₂ (D); and interactions between the probes at positions 450 and 516 and nearby residues in loop EL5 (residues 436–438 or 444–446), TM8′ (455–457), or TM-1′ (85–93; E). Simulations were performed with two bound sodium ions (Na⁺), in either the absence (left) or presence (right) of betaine. The initial protein conformation was either outward- (PDB accession no. 4LLH chain A, red) or inward-facing (PDB accession no. 4C7R, blue).

Figure 8.

Analysis of backbone distances and spin-label orientations. Distributions of Cα–Cα atom distances (A–D) and spin-label side chain orientations calculated for the biased trajectories (dashed lines) and after de-biasing those same trajectories (solid lines; E–H). Simulations were performed with two sodium bound (left panels) or with two sodium ions plus one betaine molecule bound (right panels). Data were obtained for simulations of BetP in outward- (PDB accession no. 4LLH chain A, red), or inward-open conformations (PDB accession no. 4C7R, blue). (A–D) Reference x-ray structure data (vertical gray lines) are taken from Table S2. (E–H) The angle was computed between the axes connecting the backbone Cα and nitroxide N atoms in each spin label; negative numbers indicate that the probe rings point toward one another, whereas positive numbers indicate that the probes are oriented in opposite directions.