Introduction

ATP-binding cassette (ABC) transporters constitute the largest family of primary active membrane transport systems, conserved across all domains of life^1-3. Despite considerable structural diversity, all ABC transporters share a modular architecture comprising two conserved nucleotide-binding domains (NBDs)–the defining hallmark of the family–and two transmembrane domains (TMDs) that form the substrate translocation pathway^3,4. Based on their TMD architecture, ABC transporters are classified into seven types that encompass importers, exporters, extractors, and mechanotransmitters⁵. Substrate translocation is driven by large conformational changes that are chemo-mechanically coupled to ATP binding, hydrolysis, and phosphate/ADP release^2,3. ABC transporters play central roles in cellular homeostasis, nutrient uptake, waste removal, and toxin defense. Their dysfunction and misregulation are linked to numerous diseases and drug resistance⁶.

The heterodimeric type IV ABC transporter TmrAB from Thermus thermophilus has emerged as a powerful model system due to its exceptional thermal stability and functional homology to the transporter associated with antigen processing (TAP1/2), a key component of adaptive immunity^7-9. Notably, TmrAB shares overlapping peptide specificity with TAP and can restore antigen presentation in TAP-deficient human cells¹⁰. Its inherent asymmetry, with one catalytically active (canonical) and one inactive (noncanonical) nucleotide-binding site (NBS), provides a unique opportunity to investigate functional specialization and asymmetry in ABC transport mechanisms.

Extensive structural studies, particularly using cryogenic electron microscopy (cryo-EM), have delineated the conformational landscape of TmrAB and yielded a detailed model of its translocation cycle^11,12. In this model, TmrAB fluctuates between inward-facing wide and narrow conformations (IF^wide and IF^narrow), characterized by a sealed periplasmic gate (PG) and well-separated NBDs, thereby permitting substrate access to the central binding cavity. ATP binding to both NBDs induces NBD dimerization and drives the transition into the outward-facing (OF) states, including an OF open (OF^open) conformation with an open PG that enables substrate release into the periplasm, as well as an OF occluded (OF^occluded) state characterized by a sealed PG and dimerized NBDs. Subsequent ATP hydrolysis and phosphate release lead to asymmetric unlocked return states (UR^asym and UR^asym*), before the transporter returns to the IF conformation. These UR states feature a sealed PG, a partially open ADP-bound canonical NBS, and a tightly ATP-occluded noncanonical NBS¹¹.

Single-turnover experiments have shown that ATP binding, rather than hydrolysis, drives the IF-to-OF transition, while phosphate release precedes the OF-to-IF switch^12-14. Complementary ensemble approaches, including pulsed electron–electron double resonance (PELDOR/DEER) spectroscopy, have further characterized ATP-dependent conformational changes^15,16. However, ensemble averaging inherently masks molecular heterogeneity, obscures inactive or misfolded subpopulations, and limits access to kinetic information.

Single-molecule techniques overcome these limitations by resolving conformational dynamics at the level of individual molecules^17,18. In particular, single-molecule Förster resonance energy transfer (smFRET) enables real-time monitoring of protein conformational changes with nanometer precision^19-22. Applied to ABC transporters, smFRET provides a unique opportunity to dissect transport cycles, resolve transient intermediates, and extract kinetic and mechanistic insights that remain inaccessible to ensemble-based measurement approaches^23-25.

Here, we apply total internal reflection fluorescence (TIRF) microscopy combined with alternating laser excitation (ALEX)-based smFRET to detergent-solubilized heterodimeric ABC transporter TmrAB, providing the first single-molecule characterization of this system. By strategically positioning fluorophore pairs, we directly monitor ATP-dependent NBD dimerization and periplasmic gate (PG) opening, quantify conformational state occupancies across ATP concentrations ranging from nucleotide-free to physiological levels (3 mM), and uncover conformational dynamics previously masked by ensemble averaging. Using three orthogonal trapping strategies–(i) a slow-turnover mutant^11,12, (ii) Mg²⁺ depletion^14,15, and (iii) substrate trans-inhibition^26,27–we resolved a previously hidden outward-facing open (OF^open) state that rapidly exchanges with the outward-facing occluded (OF^occluded) state. Distance measurements derived from smFRET closely matched predictions from accessible-volume (AV) simulations, cryo-EM structures, and PELDOR/DEER spectroscopy, confirming that detergent-solubilized TmrAB retains a native-like conformational landscape. Together, these results provide the first single-molecule quantification of conformational state occupancies for a heterodimeric type IV ABC transporter and establish TmrAB as a versatile model for single-molecule studies of ABC transport systems.

Results

Design of FRET-labeled TmrAB variants to probe conformational dynamics

To monitor conformational changes in distinct regions of TmrAB, we engineered double-cysteine variants for smFRET targeting the nucleotide-binding domains (NBDs) and the periplasmic gate (PG). The NBDs undergo ATP-dependent dimerization and post-hydrolysis dissociation, whereas the PG opening and closing controls substrate release into the periplasm^2,3,11. Probing both regions provides complementary readouts of cytosolic and periplasmic coupling during transport.

Labeling positions were selected based on prior PELDOR/DEER studies¹⁶. The NBD reporter variant (TmrA^C416B^L458C, referred to as TmrAB^NBD) monitors the noncanonical nucleotide-binding site (NBS), while the PG reporter (TmrA^{C416A, T61C}B^R56C, hereafter TmrAB^PG) reports PG opening. In TmrAB^NBD, the native single cysteine (C416) was retained for labeling, whereas in TmrAB^PG it was substituted by alanine to prevent off-target labeling. Using the noncanonical NBS prevents direct distinction between OF^occluded and asymmetric unlocked return states (UR^asym and UR^asym*)¹¹, reducing the number of resolvable FRET states and simplifying data interpretation.

Both variants were labeled with photostable LD555/LD655 fluorophores containing a 1,3,5,7-cyclooctatetraene moiety to suppress photobleaching and blinking^28,29. Accessible-volume (AV) simulations³⁰ across nine cryo-EM structures¹¹ confirmed that donor–acceptor distances (R_DA) and simulated FRET efficiencies (E_sim) lie within the measurable range (Fig. 1 and Table 1). For TmrAB^NBD, E_sim shifts from 0.62 ± 0.02 (57.9 ± 0.7 Å, NBDs separated) to 0.84 ± 0.01 (45.2 ± 0.4 Å, NBDs dimerized), corresponding to predicted ΔE_sim of 0.22 ± 0.02 and ΔR_DA of 12.7 ± 0.8 Å. For TmrAB^PG, E_sim changes from 0.96 ± 0.02 (30.6 ± 0.4 Å, closed PG) to 0.69 ± 0.03 (53.8 ± 2.0 Å, open PG), corresponding to predicted ΔE_sim of 0.27 ± 0.04 and ΔR_DA of 23.2 ± 2.0 Å.

Table 1:

Figure 1.

A slow-turnover TmrAB variant (TmrA^{C416A, E523Q, T61C}B^R56C, hereafter TmrAB^PG_EQ) was also generated to kinetically stabilize ATP-bound conformations. This mutation reduces ATP hydrolysis by ∼1000-fold, extending the catalytic half-life of ∼25 min at 45 °C^11,12,14,15.

TmrAB variants are suitable for FRET studies

TmrAB variants were expressed in E. coli and purified by immobilized metal-affinity chromatography. SDS-PAGE and size-exclusion chromatography (SEC) confirmed purity and monodispersity of the samples (Fig. 1–Fig. S1a,b). ATPase assays showed that TmrAB^wt retained enzymatic activity after purification, yielding a Michaelis-Menten constant (K_m) of 0.97 ± 0.28 mM and turnover rate (k_cat) of 2.57 ± 0.38 s^-1 at 40 °C (Fig. 1–Fig. S1c), consistent with previous reports^9,10,12.

Site-specific labeling yielded >90% labeling efficiency per cysteine, with ∼40–50% per-site labeling efficiency for donor-only and acceptor-only populations (Fig. 1–Fig. S1d–f). Stochastic labeling of the TmrAB variants results in both homo (donor-donor, acceptor-acceptor) and hetero (donor-acceptor) labeled species. Only complexes with appropriate donor–acceptor stoichiometry were included in subsequent analyses.

Fluorescence lifetime (τ) analysis confirmed that the conjugated fluorophores retained their photophysical properties and sufficient rotational freedom for reliable FRET measurements. τ histograms of both conjugated and free fluorophores were fitted using a biexponential decay model, from which amplitude-weighted average lifetimes were calculated. For TmrAB^NBD, average τ values were 0.93 ± 0.02 ns (LD555) and 1.52 ± 0.01 ns (LD655), whereas TmrAB^PG exhibited average τ values of 0.95 ± 0.02 ns (LD555) and 1.65 ± 0.01 ns (LD655) (Fig. 1–Fig. S2a). By comparison, free dyes in buffer displayed lifetimes of 1.12 ± 0.02 ns (LD555) and 1.29 ± 0.01 ns (LD655). The similarity between free and conjugated dye lifetimes, together with comparable values across variants, indicates minimal protein-induced quenching and supports sufficient orientational averaging for quantitative smFRET analysis^18,31-34.

Ensemble ATP titration (0–10 mM ATP) showed ATP-dependent donor quenching and acceptor sensitization (Fig. 1–Fig. S2b–d). ATP-induced fluorescence changes provided a quantitative readout of conformational transitions and enabled estimation of apparent equilibrium dissociation constants for ATP binding (K_{d, ATP}) in labeled TmrAB variants (Fig. 1–Fig. S2e–g). The measured apparent K_{d, ATP} values were 51 ± 38 μM for TmrAB^NBD, 68 ± 25 μM for TmrAB^PG, and 95 ± 26 μM for the slow-turnover variant TmrAB^PG_EQ, consistent with prior biochemical measurements (∼100 µM for TmrA^E523QB)¹². Labeling did not measurably perturb ATP binding.

Previous studies on spin-labeled TmrAB^NBD demonstrated transport activity comparable to wild-type TmrAB¹⁶, while AV simulations confirmed that fluorophores at these positions do not interfere with ATP- or substrate-binding sites. For labeled TmrAB^PG variant, no previous transport activity had been reported; we therefore performed transport assays (Fig. 1–Fig. S3) which confirmed that LD555/LD655 labeling at the PG does not impair function, as labeled and wild-type TmrAB showed indistinguishable substrate transport in single-liposome assays (Fig. 1–Fig. S3c).

ATP-induced conformational switching resolved by smFRET

TmrAB was immobilized on PEGylated surfaces using a conformation-independent, TmrB-specific nanobody (Nb9F10^S63C)¹¹. Previous studies confirmed that Nb9F10^S63C does not perturb TmrAB transport^11,14, consistent with single-liposome transport assays performed in this study (Fig. 1–Fig. S3d). For surface immobilization, Nb9F10^S63C was conjugated to maleimide-PEG₁₁-biotin, serving as an anchor to attach TmrAB to surface-immobilized streptavidin (Fig. 2a). Donor and acceptor photons were recorded by a total-internal reflection fluorescence (TIRF) microscope using alternating laser excitation (ALEX; NanoImager) at 40 °C (Fig. 2b). Single-molecule localization, fluorescence-trajectory extraction, and background correction were performed using NanoImager software, followed by DeepFRET-based machine-learning trace classification and corrections for donor leakage, direct acceptor excitation, and differences in detection-efficiency^18,35,36 (Fig. 2–Fig. S1 and Fig. 2–Fig. S2). FRET efficiency (E) and stoichiometry (S) were calculated from the corrected fluorescence trajectories (see Methods, Eq. 1 and Eq. 2), and 2D E versus S diagrams were constructed to verify that only complexes with appropriate donor–acceptor stoichiometry (S ≈ 0.5) were included in the smFRET analysis (Fig. 2c).

Figure 2.

FRET efficiency (E) histograms revealed two dominant Gaussian populations corresponding to apo and ATP-bound states (Fig. 2d,e). Low-amplitude features were excluded from quantitative analysis because they were not reproducibly observed across replicates, lacked structural support from cryo-EM and DEER/PELDOR data, introduced poorly constrained fitting parameters, and increased the risk of overfitting. In the absence of ATP, only the apo population was observed, whereas addition of 3 mM ATP induced the appearance of a second ATP-bound population. For TmrAB^NBD, the apo and ATP-bound populations exhibited mean E values of 0.58 and 0.88 (ΔE = 0.30), respectively, with ∼77% of molecules occupying the ATP-bound state. For TmrAB^PG, mean E values were 0.97 (apo) and 0.86 (ATP-bound) (ΔE = 0.11), with ∼80% of molecules in the ATP-bound state.

Distance estimates using a Förster radius of R₀ = 63.5 Å (ref.³⁷) yielded apparent distances of 60.2 Å (apo) and 45.6 Å (ATP-bound) for TmrAB^NBD, and 35.6 Å (apo) and 46.9 Å (ATP-bound) for TmrAB^PG. For TmrAB^NBD, the experimentally derived ΔR of 14.6 Å closely agreed with the AV simulations. In contrast, the smaller ΔR of 11.4 Å observed for TmrAB^PG deviated from simulated values, suggesting that the ATP-bound population at this site represents a mixture of rapidly interconverting states rather than a single well-defined state.

ATP dependence of conformational equilibria

To assess the ATP sensitivity, we quantified conformational responses across a wide range of ATP concentrations, spanning well below the reported K_{d, ATP} (∼100 µM for TmrA^E523QB)¹² up to physiologically relevant levels (3 mM ATP). ATP concentrations up to 3 mM were selected to approximate near-physiological conditions commonly used in in-vitro studies, where millimolar ATP levels ensure saturation of nucleotide-dependent conformational transitions and facilitate comparison with previous biochemical and structural analyses^10-12,14. smFRET measurements revealed dose-dependent population shifts: TmrAB^NBD shifted from a low-FRET apo state (E = 0.58) to a high-FRET ATP-bound state (E = 0.88), whereas TmrAB^PG shifted from a high-FRET apo state (E = 0.97) to a lower-FRET ATP-bound state (E = 0.86) (Fig. 3a,c). Langmuir isotherm fits yielded apparent K_{d, ATP} values of 13 ± 1 μM for TmrAB^NBD and 2 ± 1 μM for TmrAB^PG (Fig. 3b,d), significantly lower than K_{d ATP} determined in ensemble FRET measurements (Fig. 1–Fig. S2e–g). This difference likely arises from the difficulty of deconvoluting overlapping FRET populations at sub-K_{d, ATP} concentrations, particularly for TmrAB^PG, where state assignment is less well separated. Despite this quantitative offset, both approaches consistently indicate ATP saturation well below physiological concentrations and therefore support the same mechanistic conclusion that ATP binding drives conformational switching in TmrAB.

Figure 3.

Trapping of TmrAB^PG reveals a previously hidden outward-facing open state

At 3 mM, ∼80% of TmrAB^PG complexes populated the ATP-bound state (E = 0.86), closely matching the ATP-bound population (∼77% NBD-dimerized) observed for TmrAB^NBD (Fig. 3). This agreement indicates that both labeling strategies consistently report ATP-dependent conformational changes. However, the ATP-induced shift observed for the periplasmic-gate reporter TmrAB^PG (ΔE = 0.11, ΔR = 11.4 Å; Fig. 3c) was substantially smaller than predicted by AV simulations (ΔE = 0.27, ΔR = 23.2 Å; Table 1). This discrepancy suggests that the ATP-bound population at E = 0.86 represents an unresolved ensemble, potentially comprising OF^open and OF^occluded conformations as well as the post-hydrolysis asymmetric unlocked return states (UR^asym and UR^asym*), which cannot be distinguished from OF^occluded within the current FRET geometry¹¹.

To determine whether these states are kinetically unresolved by smFRET, we applied three complementary strategies to arrest the OF^open conformation of TmrAB^PG: (i) a slow-turnover mutant (TmrAB^PG_EQ), (ii) Mg²⁺ depletion using EDTA, and (iii) trans-inhibition by high concentrations of the substrate peptide R9L (RRYQKSTEL) (Fig. 4). Slow-turnover variants have previously enabled structural separation of OF^open and OF^occluded states^11,12,14,15. Mg²⁺ depletion blocks ATP hydrolysis while preserving ATP binding, thereby enabling rapid and reversible trapping¹⁴. We further hypothesized that trans-inhibition by peptide binding sterically restricts PG closure and would therefore stabilize OF^open in a dose-dependent manner^26,27.

Figure 4.

In the absence of ATP, either without nucleotide or in the presence of ADP (3 mM), the slow-turnover variant TmrAB^PG_EQ populated a single high-FRET state (E = 0.97) (Fig. 4a, top and middle). These results indicate that ADP binding alone is insufficient to promote either NBD dimerization or PG opening, which is consistent with previous biochemical and structural observations^11,12. Upon ATP addition (3 mM), however, the conformational landscape diverged sharply from that of TmrAB^PG. Instead of the two-state distribution observed for TmrAB^PG (apo: E = 0.97, ∼20%; ATP-bound: E = 0.86, ∼80%) (Fig. 2e, bottom), TmrAB^PG_EQ exhibited three well-resolved populations with mean E values of 0.97 (∼14%), E = 0.86 (∼55%), and E = 0.63 (∼31%) (Fig. 4a, bottom).

As in TmrAB^PG, the high-FRET population (E = 0.97) corresponds to the IF state, whereas the intermediate-FRET population (E = 0.86) likely represents a dynamic equilibrium between OF^open and OF^occluded conformations, as suggested by cryo-EM analyses¹¹. Post-hydrolysis return states (UR^asym and UR^asym*) are expected to be minimally populated in TmrAB^PG_EQ because of its drastically reduced ATP hydrolysis rate¹². Notably, the low-FRET ATP-bound population (E = 0.63) was entirely absent in wild-type TmrAB^PG. The transition from E = 0.97 to E = 0.63 corresponds to ΔE = 0.34 and ΔR = 22.5 Å, in close agreement with the IF→OF^open distance predicted by AV simulations (ΔR = 23.2 Å; Fig. 1 and Table 1), thereby supporting assignment of the E = 0.63 population to the OF^open conformation.

Mg²⁺ depletion independently reproduced this three-state landscape. In the absence of Mg²⁺, wild-type TmrAB^PG transitioned from a single apo population (without ATP; E = 0.97) (Fig. 4b, top) to three ATP-bound populations (3 mM ATP; E = 0.97, ∼13%; E = 0.86, ∼56%; E = 0.63, ∼31%) (Fig. 4b, middle), closely resembling those observed for the slow-turnover variant (Fig. 4a, bottom). Reintroducing Mg²⁺ abolished the E = 0.63 population and restored the wild-type TmrAB^PG two-state distribution (Fig. 4b, bottom). This reversibility confirmed that the ATP-bound OF^open state (E = 0.63) is selectively revealed only when ATP hydrolysis is prevented.

Finally, we tested whether periplasmic substrate binding shifts the conformational equilibrium of wild-type TmrAB^PG toward OF^open. In the presence of ATP (3 mM), increasing concentrations of peptide substrate (0.3–2 mM R9L) progressively enriched the E = 0.63 population from ∼20% to ∼38% (Fig. 4c). This dose-dependent stabilization mirrors the trans-inhibition behavior reported for human TAP1/2 and reflects the upper substrate-loading capacity of the transporter²⁶.

Distance changes derived from smFRET closely match AV simulations, cryo-EM structures (PDB 6RAH, 6RAN)¹¹, and DEER/PELDOR measurements¹⁵ (Fig. 4d), together validating assignment of the E = 0.63 population as the OF^open conformation.

Kinetics and thermodynamics of the transport cycle

ALEX-smFRET data were acquired with an effective temporal resolution of 200 ms (100 ms per excitation channel). Shorter integration times compromised the signal-to-noise ratio and precluded reliable FRET determination. To quantify conformational dynamics, we applied Hidden Markov Modeling (HMM) using MASH-FRET³⁸, classifying traces as either static (single FRET state) or dynamic (multiple states). Approximately 95% of traces in each condition were classified as static. The predominance of static traces indicates that most conformational transitions occur faster than the 200 ms observation window rather than reflecting conformational arrest.

Although individual transitions could not be directly resolved, population-based analysis (Fig. 3), combined with biochemical turnover measurements (Fig. 1–Fig. S1c), allowed estimation of ATP-bound dwell times. At saturating ATP conditions well above the apparent K_{d, ATP} (3 mM, 40 °C), wild-type TmrAB exhibited a catalytic turnover rate of k_cat = 2.57 ± 0.38 s^-1 (Fig. 1–Fig. S1c), corresponding to a full transport cycle time (τ_cycle) of 395 ± 55 ms. ATP-bound dwell times (τ_d) were derived from population ratios obtained from Gaussian fits of the FRET efficiency histograms (Fig. 3; see Methods, Eq. 3 and Eq. 4). These analyses yielded ATP-bound dwell times of 304 ± 43 ms for TmrAB^NBD (∼77% ATP-bound) and 316 ± 44 ms for TmrAB^PG (∼80% ATP-bound), with the remaining ATP-free intervals (∼20–23%) accounting for 91 ms and 79 ms of the cycle, respectively.

Together, these measurements establish a quantitative, single-molecule description of conformational state occupancies and dwell times throughout the catalytic cycle of a heterodimeric ABC transporter under active turnover conditions, providing a quantitative single-molecule framework for describing the dynamic landscape of its translocation cycle.

Discussion

smFRET has become an indispensable tool for dissecting conformational dynamics of membrane proteins, including receptors, ion channels, and transporters, by directly linking structural transitions to functional states. With the ABC transporter family, however, smFRET studies have largely focused on monomeric or homodimeric systems^23-25, leaving asymmetric heterodimeric transporters comparatively underexplored. Here, we extend single-molecule analysis to the heterodimeric type IV ABC transporter TmrAB, revealing dynamic features of the transport cycle that are inaccessible to ensemble-averaged approaches.

By positioning FRET reporters at the nucleotide-binding domains (NBDs) and periplasmic gate (PG), we directly monitored ATP-dependent coupling between chemical energy input and global conformational rearrangements. Importantly, NBD dimerization is not uniquely coupled to a single outward-facing state but can lead to either an outward-facing open (OF^open) or occluded (OF^occluded) conformation^11,12. This decoupling highlights the need to monitor both cytosolic and periplasmic regions to resolve the transport mechanism.

A limitation of the present study is the use of two FRET reporter pairs. While these positions were selected to probe key functional regions and validated by structural and spectroscopic benchmarks, additional labeling positions could in principle resolve further intermediates. However, increasing the number of states would also complicate quantitative analysis due to partially overlapping FRET populations, particularly for TmrAB^PG. The current design therefore balances mechanistic resolution with robust state assignment.

Labeling sites previously validated for PELDOR/DEER spectroscopy^10,15,16 were adapted for smFRET and rigorously benchmarked using accessible-volume (AV) simulations³⁰, fluorescence lifetime analysis, and ensemble FRET titrations. This validation is essential because fluorophores impose stricter steric and rotational constraints than spin labels³². Collectively, these controls demonstrate that fluorophore attachment preserves native-like conformational behavior and ATP responsiveness. Fluorescence lifetime measurements further indicated efficient energy transfer without evidence of substantial protein-fluorophore quenching or restricted mobility¹⁸.

Single-molecule measurements resolved two dominant FRET populations corresponding to apo and ATP-bound states for both TmrAB^NBD and TmrAB^PG. ATP titrations from sub-K_{d, ATP} to physiological concentrations (3 mM ATP) produced gradual, concentration-dependent redistribution between these states, demonstrating high sensitivity of both reporters. Apparent K_{d, ATP} values derived from smFRET (2–13 µM) were lower than those obtained from ensemble FRET (50–100 µM). Importantly, both approaches agree that TmrAB reaches conformational saturation well below physiological ATP concentrations and retains Michaelis–Menten behavior.

Using three independent strategies—slow-turnover catalysis, Mg²⁺ depletion, and substrate trans-inhibition—we resolved this previously hidden OF^open conformation. The observed distance changes agree with cryo-EM structures¹¹, PELDOR/DEER data^10,15,16, and AV-based predictions, indicating that detergent-solubilized TmrAB samples a largely native-like conformational landscape.

Although detergent and lipid nanodisc environments yield similar global states, fluorophore behavior may still be influenced by membrane proximity in lipid systems^20,39,40. AV simulations suggest that dyes near the periplasmic gate may experience steric constraints due to partial membrane overlap. While negligible under detergent conditions, this effect should be considered in future studies of membrane-reconstituted systems.

Substrate addition shifted the conformational equilibrium in a concentration-dependent manner, progressively stabilizing the OF^open state. This behavior is consistent with trans-inhibition observed in other ABC transporters, including TAP1/2²⁶ and ABCC1²⁷, and reflects finite substrate loading capacity²⁶.

Quantitative deconvolution of FRET populations enabled estimation of conformational state occupancies under physiological ATP concentrations. During active turnover, TmrAB populates the IF state (∼20%), OF^open (∼25%), and OF^occluded/post-hydrolysis states (UR^asym and UR^asym*) (∼55%; Fig. 5). The current labeling scheme does not allow direct separation of OF^occluded from post-hydrolysis states due to fluorophore placement at the noncanonical NBS. However, post-hydrolysis contributions are expected to be reduced in the slow-turnover mutant TmrAB^PG_EQ. To our knowledge, this represents the first single-molecule quantification of conformational equilibria in a heterodimeric ABC transporter under catalytic conditions. While IF substructure resolution is lower than in cryo-EM classifications^11,14, smFRET uniquely captures ATP-bound intermediates that interconvert too rapidly for structural resolution.

Figure 5.

All experiments were performed at 40 °C for consistency with prior biochemical work and fluorophore stability. At the physiological temperature of T. thermophilus (68 °C), absolute rates are expected to increase, although relative state distributions may be preserved if the free-energy landscape remains unchanged. Despite a substantial fraction of static trajectories, ATP-dependent shifts indicate that TmrAB operates near the temporal resolution of the measurements. Combining state occupancies with enzymatic turnover yields an ATP-bound dwell time of ∼300 ms, consistent with previous biochemical estimates^9,12.

Emerging microsecond-resolution smFRET approaches could directly resolve short-lived intermediates⁴¹, and multi-color FRET strategies^42,43 may enable simultaneous monitoring of NBDs, PG, and nucleotide sites. These advances would allow direct detection of the OF^occluded states and post-hydrolysis dynamics.

Overall, the consistency between reporter positions, ensemble measurements, and structural data support the conclusion that the observed conformational states reflect native TmrAB behavior. While certain intermediates remain kinetically unresolved under turnover conditions, they are strongly supported by independent trapping approaches, reinforcing the robustness of the mechanistic model. Additional conditions, including nucleotide-state asymmetry and substrate-only binding, represent promising directions for future work.

In summary, this work establishes smFRET as a powerful approach for mapping the dynamic landscape of asymmetric ABC transporters. By quantitatively linking ATP binding, conformational equilibria, and kinetics at the single-molecule level, we define how chemical energy is converted into directional transport in heterodimeric ABC systems.

Methods

Key resources table

Expression, purification, and labeling of TmrAB

His₁₀-tagged TmrAB variants were expressed in E. coli BL21(DE3) (Thermo Fisher Scientific) as described previously¹². Cells were grown in high-salt LB media (Carl Roth) supplemented with 100 μg ml^-1 ampicillin (PAA Laboratories) at 37 °C. At an OD₆₀₀ of 0.5, expression was induced with 1 mM isopropyl β-D-thiogalactopyranoside (IPTG; Carl Roth), and cultures were incubated for 3 h at 37 °C. Cells were harvested by centrifugation (4,500 × g, 4 °C, 15 min) and stored at –80 °C.

For purification, cell pellets were resuspended in lysis buffer (20 mM HEPES-NaOH pH 7.5, 300 mM NaCl, 50 µg ml^-1 lysozyme, 0.2 mM phenylmethylsulfonyl fluoride (PMSF)) and lysed by sonication. Cell debris was removed by centrifugation (18,000 × g, 4 °C, 35 min), and membranes were collected by ultracentrifugation (100,000 × g, 4 °C, 30 min). Membranes were solubilized for 2 h at 4 °C in purification buffer (20 mM HEPES-NaOH pH 7.5, 300 mM NaCl) containing 20 mM n-dodecyl β-D-maltoside (β-DDM; Carl Roth). After ultracentrifugation (100,000 × g, 30 min, 4 °C), the supernatant was incubated with Ni-NTA agarose (Bio-Rad) for 1 h at 4 °C. The resin was washed with 20 column volumes of wash buffer (20 mM HEPES-NaOH pH 7.5, 300 mM NaCl, 1 mM β-DDM) containing 50 mM imidazole, and TmrAB was eluted with elution buffer (20 mM HEPES-NaOH pH 7.5, 300 mM NaCl, 1 mM β-DDM, 300 mM imidazole).

For fluorophore labeling, TmrAB variants were conjugated via maleimide chemistry using LD555 and LD655 (Lumidyne Technologies). Labeling was carried out at a 1:10:10 molar ratio of protein to each dye in elution buffer for 3 h at 4 °C. Excess dye was quenched with 2 mM β-mercaptoethanol (Sigma-Aldrich), and the labeled protein was buffer-exchanged into size-exclusion chromatography (SEC) buffer (20 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 1 mM β-DDM) using a Zeba Spin Desalting Column (Thermo Fisher Scientific). Unreacted fluorophores were removed by SEC on a Superdex 200 Increase 10/300 GL column (Cytiva). Labeling efficiency was determined by analytical SEC (Superdex 200 Increase 3.2/300; Cytiva) by monitoring absorbance at 280, 555, and 655 nm. To preserve sample integrity for smFRET measurements, TmrAB was purified and labeled within a single day, stored on ice, and imaged over the following two days.

Time-correlated single-photon counting (TCSPC)

Fluorescence lifetime measurements were performed using a FluoTime 100 spectrometer (PicoQuant) equipped for time-correlated single-photon counting (TCSPC). Experiments were carried out on labeled TmrAB in SEC buffer (20 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 1 mM β-DDM). LD555 and LD655 were excited at 510 nm and 610 nm, respectively. Emission was collected using a 620/60 nm bandpass filter for LD555 and a BG4 700 nm long-pass filter for LD655. Photon arrival times were accumulated until the TCSPC histogram reached a peak count of 50,000 photons. Fluorescence decay curves were analyzed by fitting mono- or bi-exponential decay models using FluoFit software (PicoQuant) and amplitude-weighted average of fluorescence lifetime was calculated.

Nanobody production and purification

The nanobody Nb9F10^S63C was expressed and purified as described previously¹¹. Briefly, Nb9F10^S63C was produced in E. coli BL21(DE3) cells grown in Terrific Broth (TB; Carl Roth) supplemented with 100 μg ml^-1 ampicillin at 37 °C. At an OD₆₀₀ of 0.6, expression was induced with 1 mM IPTG, followed by overnight incubation at 28 °C. Cells were harvested by centrifugation (4,500 × g, 4 °C, 15 min) and stored at –80 °C. For purification, cell pellets were resuspended in nanobody lysis buffer (25 mM HEPES-NaOH pH 7.4, 300 mM NaCl, 15 mM imidazole, 0.5 mM PMSF) and disrupted by sonication. Cell debris was removed by centrifugation (18,000 × g, 4 °C, 35 min), and the clarified lysate was applied to Ni-NTA agarose equilibrated in potassium phosphate (KP_i) buffer (25 mM KP_i pH 6.5, 100 mM KCl, and 0.5 mM tris(2-carboxyethyl) phosphine (TCEP)). Bound nanobody was washed with 10 column volumes (CV) of KP_i buffer and eluted with 8 CV of elution buffer (25 mM KP_i pH 6.0, 20 mM KCl, 300 mM imidazole, 0.5 mM TCEP). Eluted fractions were pooled and further purified by cation exchange chromatography (CEX) on a HiTrap SP column (Cytiva) using a linear gradient from low-salt buffer (25 mM KP_i pH 6.0, 20 mM KCl, 0.5 mM TCEP) to high-salt buffer (25 mM KP_i pH 6.0, 500 mM KCl, 0.5 mM TCEP). The purified nanobody was concentrated and buffer-exchanged into nanobody SEC buffer (20 mM HEPES-NaOH pH 7.5, 150 mM NaCl) using Zeba spin desalting columns, followed by SEC (Superdex 200 Increase 10/300 GL; Cytiva). For site-specific conjugation, Nb9F10^S63C was incubated with a biotin-PEG₁₁-maleimide linker (Sigma-Aldrich) at a 1.2:1 molar ratio of protein to linker in the presence of 0.5 mM TCEP for 2 h at 4 °C. Excess linker was removed by desalting on Zeba Spin Desalting Columns, followed by a final SEC step (Superdex 200 Increase 10/300 GL).

SDS-PAGE

The purity of TmrAB samples was assessed by SDS-PAGE. Resolving gels (10%) were prepared using 10% (w/v) acrylamide, 0.5 M Tris-HCl (pH 8.8), 0.13% (w/v) SDS, 0.05% (w/v) ammonium persulphate (APS), and 0.25% (v/v) N,N,N’,N’-tetramethylethylenediamine (TEMED). Stacking gels contained 4.3% (w/v) acrylamide, 0.5 M Tris/HCl (pH 6.8), 0.09% (w/v) SDS, 0.09% (w/v) APS, and 0.33% (v/v) TEMED. Gels were used immediately or stored at 4 °C for up to 4 weeks. Protein samples were mixed with 4× SDS loading buffer containing dithiothreitol (DTT; Sigma-Aldrich) and heated at 90 °C for 5 min before loading. Electrophoresis was performed at a constant voltage of 120 V using 1× SDS running buffer (25 mM Tris-HCl pH 8.8, 192 mM glycine, 0.1% SDS). Proteins were visualized by staining with InstantBlue^™ Protein Stain (Abcam) for 1 h at room temperature with gentle agitation and imaged using a Fusion FX system (Vilber).

ATPase activity assay

The ATPase activity of β-DDM-solubilized TmrAB^wt was quantified using a Malachite Green-based colorimetric assay as described previously⁴⁴. Detergent-solubilized TmrAB (60 nM) was incubated in ATPase buffer (20 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 2 mM MgCl₂, 1 mM β-DDM) containing 3 mM ATP (Sigma-Aldrich) at 40 °C for 7 min. Autohydrolysis controls were prepared by incubation of ATP in ATPase buffer without protein. Reactions were quenched by adding 20 mM H₂SO₄, followed by incubation with 3 mM Malachite Green (Thermo Fisher Scientific), 0.2% (v/v) Tween20 (Carl Roth), and 1.5% (w/v) ammonium molybdate (Carl Roth) for 10 min at room temperature. The absorbance at 620 nm was recorded on a CLARIOstar v.5.20 R5 plate reader (BMG LABTECH).

Ensemble FRET measurements

The ATP binding and FRET characteristics of selected TmrAB variants were assessed by ensemble FRET. Labeled TmrAB (100 nM) was incubated with increasing concentrations of ATP at 42 °C for 5 min. Donor-excited emission was recorded from 550–700 nm with an excitation wavelength of 520 nm using a CLARIOstar v.5.20 R5 plate reader (BMG LABTECH). Acceptor emission intensities at 675 nm were plotted against ATP concentration and fitted with a hyperbolic function to determine the apparent dissociation constant (K_{d, ATP}) for each variant.

Reconstitution of TmrAB

TmrAB variants were reconstituted into proteoliposomes composed of E. coli polar lipids and dioleoylphosphatidylcholine (DOPC) at a molar ratio of 7:3. A protein-to-lipid ratio of 1:20 (w/w) was used, corresponding to approximately 50 TmrAB complexes per liposome. Extruded large unilamellar vesicles (LUVs) were destabilized by incubation with 0.3% (v/v) Triton X-100 for 30 min. TmrAB was then added to the detergent-destabilized liposomes and incubated for 30 min at 8 °C under gentle rotation. Detergent was removed gradually using sequential addition of polystyrene Bio-Beads SM-2 (Bio-Rad): two incubations at 40 mg ml^-1 (1 h and overnight), followed by two incubations at 80 mg ml^-1 (1 h each). Proteoliposomes were subsequently harvested by ultracentrifugation (100,000 × g, 30 min, 4 °C) and resuspended in the appropriate buffer.

Single liposome transport assay

Single liposome transport assay was performed as described previously^12,13. Wild-type TmrAB or LD555/LD655-labeled TmrAB^PG (0.6 µM each), reconstituted into liposomes (∼50 TmrAB per liposome), was mixed with fluorescently labeled peptide RRYC^FKSTEL (30 µM C4F; ^F, fluorescein), 3 mM ATP or ADP, and 5 mM MgCl₂ in liposome buffer (25 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 5% (v/v) glycerol). Where indicated, nanobody Nb9F10^S63C (0.6 µM) was added. Samples were incubated for 5 min at 40 °C (unless specified otherwise), followed by addition of ice-cold ethylenediaminetetraacetic acid (EDTA; 10 mM). Proteoliposomes were then washed twice by centrifugation (270,000 x g, 30 min) to remove non-transported peptide. Mean fluorescence intensities (MFIs) of translocated substrates were quantified by flow cytometry using a FACSMelody (Waters Biosciences) and analyzed in FlowJo 10.6.1 (Waters Biosciences). The gating strategy used to determine fluorescence intensities of single liposomes from raw flow cytometry data consisted of three steps: First, forward scatter area versus side scatter area (FSC-A vs SSC-A) was used to separate populations based on size and granularity. Second, forward scatter height versus forward scatter area (FSC-H vs FSC-A) was used to exclude aggregates and retain singlet events. Finally, fluorescence intensities of the gated singlet population were quantified at the wavelength of interest (527 nm), and the resulting values were used to calculate mean fluorescence intensity (MFI) of the proteoliposome population.

Solid-phase peptide synthesis

The peptides used in this study were synthesized and labeled as described previously^14,45. The 9-mer peptides RRYQKSTEL (R9L) and RRYC^FKSTEL (C4F; ^F, fluorescein) were synthesized on preloaded Fmoc-L-Leu resin using a Liberty microwave-assisted peptide synthesizer (CEM) according to a standard protocol (54 W, 3 min, 75 °C). Each coupling step was performed twice using 0.2 M fluorenylmethoxycarbonyl (Fmoc)-protected amino acids, 0.5 M O-benzotriazole-N,N,N’,N’-tetramethyluronium hexafluoro-phosphate (HBTU), and 1-hydroxybenzotriazole hydrate (HOBt H₂O). Fmoc deprotection was carried out with 20% (v/v) piperidine and 0.1 M HOBt H₂O in dimethylformamide (DMF). Peptides were cleaved from the resin using a cleavage cocktail containing 92.5% (v/v) trifluoroacetic acid (TFA), 2.5% (v/v) H₂O, 4.5% (v/v) thioanisole, and 0.5% (v/v) 1,2-ethanedithiol (EDT) for 1.5 h at room temperature. Cleaved peptides were precipitated in ice-cold diethyl ether (Et₂O), pelleted, dissolved in tert-butanol (tBuOH)/water (4:1, v/v), and lyophilized. Peptides were purified by reverse-phase (RP) C18 HPLC (PerfectSil C18 column; MZ-Analysentechnik) using buffer A (Milli-Q water containing 0.05% (v/v) TFA) and buffer B (acetonitrile containing 0.05% (v/v) TFA). Peptide identity and purity were confirmed by LC-MS. For fluorescence labeling, the peptide RRYC^FKSTEL was dissolved in 33% (v/v) DMF in PBS (10 mM Na₂HPO₄, 1.8 mM KH₂PO₄, 137 mM NaCl, 2.7 mM KCl, pH 7.4) and incubated for 1 h at room temperature with a 1.3-fold molar excess of 5-iodoacetamido fluorescein (Merck).

Functionalization of glass slides for single-molecule FRET analysis

Glass coverslips used for TmrAB immobilization in smFRET experiments were functionalized by PEGylation as described previously⁴⁶. Coverslips (Carl Roth) were cleaned by sequential sonication in Milli-Q water and analytical-grade acetone (>99.9%; VWR International), followed by oxygen plasma treatment (0.3 mbar, 80% power, 15 min) using a Zepto plasma cleaner (Diener) and a 10 min incubation in methanol (Avantor, Gliwice, PL). Coverslips were then silanized by incubation for 30 min in a solution of 100 ml methanol, 5 ml acetic acid, and 3 ml 3-aminopropyltrimethoxysilane (APTES; Tokyo Chemical Industry). After silanization, slides were rinsed four times with methanol and dried under a nitrogen stream. Surface functionalization was achieved using a mixture of biotinylated-PEG (4 mol%) and nonbiotinylated-PEG (96 mol%; Rapp Polymer). The PEG solution was sandwiched between two coverslips and incubated overnight in a humidity chamber. Coverslips were then rinsed thoroughly with Milli-Q water and dried under nitrogen. To enhance passivation, a second PEGylation step was performed using 25 mM CH₃-PEG-NHS (333 Da; Thermo Fisher Scientific) under the same conditions. Finally, slides were rinsed with Milli-Q water, dried under nitrogen, and stored at –20 °C under argon until use.

Single-molecule FRET imaging

Single-molecule FRET (smFRET) experiments were performed using a flow chamber system (Ibidi). Chambers were assembled by placing a biotin-PEG-functionalized glass slide onto a μ-Slide I Luer Family flow channel (Ibidi), with the functionalized surface facing inward. All buffers and Milli-Q water were filtered through 0.2 μm filters (Sigma-Aldrich). Chambers were washed with 1 ml Milli-Q water and incubated with 0.2 mg ml^-1 streptavidin (Sigma-Aldrich) at 4 °C for 30 min to allow binding to the biotin-PEG surface. Unbound streptavidin was removed by washing with 1 ml Milli-Q water. The surface was then treated with 0.3 mg ml^-1 biotinylated-PEG₁₁-Nb9F10^S63C at 4 °C for 45 min, followed by flushing with 2 ml SEC buffer (20 mM HEPES-NaOH pH 7.5, 150 mM NaCl, 1 mM β-DDM). Detergent-solubilized, fluorophore-labeled TmrAB (100 nM) was added and incubated at 4 °C for 1 h. Unbound protein was removed by five washes with 1 ml TmrAB-SEC buffer. Chambers were equilibrated with 1 ml of imaging buffer containing 25 mM HEPES-NaOH (pH 7.5), 150 mM NaCl, 3 mM MgCl₂, 50 mM glucose, 5 mM Trolox, 7.5 U ml^-1 pyranose oxidase, and 1 kU ml^-1 catalase, supplemented with the desired ATP concentration. For EDTA trapping experiments, MgCl₂ was omitted and replaced with 3 mM ethylenediaminetetraacetic acid (EDTA; Sigma-Aldrich). smFRET data were acquired at 40 °C using alternating laser excitation (ALEX) on a total internal reflection fluorescence (TIRF) microscope (NanoImager S, ONI, Oxford, UK). Typically, 600 frames were recorded per region of interest (ROI) with 100 ms exposure time. Laser powers were 0.8 mW cm^-2 (532 nm) and 0.9 mW cm^-2 (640 nm). Data were recorded in 1-min intervals, except for TmrAB^NBD variant in apo and 3 mM ATP conditions, where 3-min intervals were used to confirm that conformational transitions do not occur on timescales longer than one minute due to the reduced temperature.

Single-molecule FRET data analysis

Single-molecule FRET (smFRET) measurements were performed using alternating laser excitation (ALEX), allowing assignment of detected photons based on both excitation and emission wavelengths. Photon counts for each molecule were extracted using NanoImager software (ONI NanoImager, Development Build) and classified into three detection channels: donor excitation with donor emission , donor excitation with acceptor emission , and acceptor excitation with acceptor emission . Traces were analyzed with DeepFRET³⁵ and manually curated. To minimize bias, a second researcher independently curated traces from both ATP-free and 3 mM ATP samples, yielding 98% overlap between curations. FRET efficiency (E) and stoichiometry (S) were calculated as:

Population analysis was performed by constructing one-dimensional histograms of FRET efficiency (E) and stoichiometry (S) using OriginPro 2024 (OriginLab). Histograms were fitted with two Gaussian distributions corresponding to the ATP-free state (defined from apo samples) and the ATP-bound state (defined by a two-component fit at saturating ATP). Hidden Markov Modeling (HMM) of individual traces was performed using MASH-FRET³⁸ to distinguish dynamic from static molecules within each sample.

ATP-bound dwell times and distribution of conformational states

The ATP-bound dwell time (τ_d) was estimated as:

where f_ATP-bound is the fraction of ATP-bound molecules derived from Gaussian fits of the FRET efficiency histograms (Fig. 3). This approach assumes (i) ATP hydrolysis occurs exclusively at the canonical NBS, with negligible contribution from the noncanonical site^9,11,15, and (ii) the majority of molecules are catalytically competent and continuously cycling.

The distribution of conformational states within ATP-bound FRET population (E = 0.86) of TmrAB^PG under turnover conditions (3 mM ATP) was determined using a two-state model:

where f_open is the fraction of OF^open state (E_open = 0.63) readily resolved under trapping conditions (Fig. 4), and f_other represents the combined fraction of PG-closed states (OF^occluded/UR^asym/UR^asym*, E_other = 0.97).

Data availability

Source data are provided with this paper: DOI: https://doi.org/10.25716/gude.0jbq-k1q6.

Figure supplements

Figure 1–Figure supplement 1.

Figure 1–Figure supplement 2.

Figure 1–Figure supplement 3.

Figure 2–Figure supplement 1.

Figure 2–Figure supplement 2.

Acknowledgements

This work was supported by the European Research Council (ERC Advanced Grant 101141396 to R.T.), the German Research Foundation via the Collaborative Research Center CRC 1507/P18 to R.T. and the Research Training Group (GRK 1986/B4.7 to R.T.). We thank Jan F.M. Stuke and Jonas Göhmann for support in automating trace extraction from ONI NanoImager software, Dr. David Glück for guidance on lifetime measurements, and Tobias Nocker for preparing nanobodies used in the study. We are also grateful to the Wachtveitl lab (Goethe University Frankfurt) for access to their FluoTime 100 spectrometer (PicoQuant). Finally, we thank Dr. Rupert Abele, Dr. David Glück, Dr. Simon Trowitzsch, Inga Nold, and Andrea Pott for helpful comments and proofreading of the manuscript.

Additional information

Author contributions

M.P. prepared all TmrAB samples and carried out the experiments for this study. M.P. performed data analysis. Curated traces were independently checked by C.N. to avoid human bias. M.P. and C.N. prepared functionalized glass slides for single-molecule FRET. M.P. and R.T. wrote the manuscript. R.T. conceived and supervised the work.

Funding

EC | European Research Council (ERC)

https://doi.org/10.3030/101141396

• Robert Tampé

Deutsche Forschungsgemeinschaft (DFG) (CRC 1507/P18)

• Robert Tampé

Deutsche Forschungsgemeinschaft (DFG) (GRK 1986/B4.7)

• Robert Tampé