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 transmembrane-domain 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 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 sates feature as sealed PG, a partially open ADP-bound canonical NBS, and a tightly ATP-occluded noncanonical NBS¹¹.

Single-turnover experiments established that ATP binding, rather than hydrolysis, drives the IF-to-OF transition, while phosphate release precedes the OF-to-IF switch^12,13. Complementary ensemble approaches, including pulsed electron–electron double resonance (PELDOR/DEER) spectroscopy, have further characterized ATP-dependent conformational changes^14,15. 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^16,17. In particular, single-molecule Förster resonance energy transfer (smFRET) enables real-time monitoring of protein conformational changes with nanometer precision^18–21. 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^22–24.

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 catalytic mutant^11,12, (ii) Mg²⁺ depletion^14,25, 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 FRET variants targeting the nucleotide-binding domains (NBDs) and the periplasmic gate (PG). The NBDs undergo ATP-dependent dimerization followed by 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 the transport cycle.

Labeling positions were selected based on prior PELDOR/DEER studies¹⁵. The NBD reporter variant (TmrA^C416B^L458C, referred to as TmrAB^NBD) monitors conformational changes at the noncanonical nucleotide-binding site (NBS), while the PG reporter (TmrA^{C416A, T61C}B^R56C, hereafter TmrAB^PG) tracks 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. Selecting the noncanonical rather than the canonical NBS prevents distinguishing outward-facing occluded (OF^occluded) from asymmetric unlocked return states (UR^asym and UR^asym*)¹¹, but reduces the number of resolvable FRET states and thereby simplifies data interpretation.

Both variants were labeled with photostable fluorophores, LD555 (donor) and LD655 (acceptor), containing a 1,3,5,7-cyclooctatetraene moiety to suppress photobleaching and blinking^28,29. Accessible-volume (AV) simulations³⁰ performed on nine cryo-EM structures¹¹ confirmed that donor-acceptor distances (R_DA) and simulated FRET efficiencies (E_sim) fall within the FRET-sensitive range (Fig. 1). For TmrAB^NBD, E_sim ranged from 0.62 ± 0.02 (57.9 ± 0.7 Å, NBDs separated) to 0.84 ± 0.01 (45.2 ± 0.4 Å, NBDs dimerized). For TmrAB^PG, E_sim shifted from 0.96 ± 0.02 (30.6 ± 0.4 Å, PG closed) to 0.69 ± 0.03 (53.8 ± 2.0 Å, PG open). These transitions correspond to ΔE_sim values of 0.22 and 0.27 and ΔR_DA of 12.7 Å and 23.2 Å, for TmrAB^NBD and TmrAB^PG, respectively, predicting robust and experimentally resolvable FRET changes.

Figure 1.

Figure 1 Figure supplement 1.

Figure 1 Figure supplement 2.

Additionally, we employed a slow-turnover TmrAB variant that reports on PG opening (TmrA^{C416A, E523Q, T61C}B^R56C, hereafter TmrAB^PG_EQ). Substituting the catalytic glutamate with glutamine removed the carboxylate required to activate water for nucleophilic attack on ATP, thus drastically reducing the rate of ATP hydrolysis. This mutation slows down the catalytic turnover (∼1000-fold) to a half-life of approximately 25 min at 45 °C^11,12,14,25, enabling stabilization of ATP-bound outward-facing conformations.

TmrAB variants are suitable for FRET studies

TmrAB variants were expressed in E. coli and purified using immobilized metal-affinity chromatography. SDS-PAGE and size-exclusion chromatography (SEC) confirmed high sample purity and monodispersity (Fig. 1–Fig. S1a,b). ATPase assays of TmrAB^wt verified that enzymatic activity was fully retained after purification, yielding a Michaelis-Menten constant (K_m) of 0.97 ± 0.28 mM and catalytic ATP turnover rate (k_cat) of 2.57 ± 0.38 s⁻¹ at 40 °C (Fig. 1–Fig. S1c).

Cysteine-maleimide labeling of detergent-solubilized TmrAB variants achieved site-specific labeling efficiencies exceeding 90% (Fig. 1–Fig. S1d–f). Fluorescence lifetime (τ) analysis of conjugated fluorophores confirmed that their photophysical properties were preserved and that they retained sufficient rotational freedom for reliable FRET measurements. τ histograms of both conjugated and free fluorophores were fitted with 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), while 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 1.11 ± 0.02 ns (LD555) and 1.29 ±0.01 ns (LD655). Because the fluorescence lifetimes of both, the conjugated dyes and the free dyes, remain on the ∼1 ns timescale, we conclude that the fluorophores remain photophysically active and are not affected by protein-induced quenching^17,31. Moreover, the measured lifetimes on the nanosecond timescale are only marginally affected and, most importantly, identical between the TmrAB variants, indicating dynamics orientational averaging of the transition dipoles and confirming that the labeled constructs are suitable for quantitative FRET studies^32–34.

Ensemble ATP titration (0–10 mM ATP) revealed the expected concentration-dependent donor quenching and acceptor sensitization (Fig. 1–Fig. S2b–d). ATP-induced fractional fluorescence changes provided as a quantitative readout of conformational transitions, allowing estimation of equilibrium dissociation constants for ATP binding (K_{d, ATP}) to labeled TmrAB variants (Fig. 1–Fig. S2e–g). The measured apparent K_{d, ATP} values––51 ± 38 μM for TmrAB^NBD, 68 ± 25 μM for TmrAB^PG, and 95 ± 26 μM for the slow-turnover variant TmrAB^PG_EQ––are in good agreement with previously reported values (∼100 µM for TmrA^E523QB)¹², indicating that fluorophore labeling does not perturb ATP binding. Notably, TmrAB^PG_EQ exhibited a larger shift in ATP-induced fluorescence change than TmrAB^PG, consistent with stabilization of the ATP-bound conformation and reduced catalytic turnover.

ATP-induced conformational switching resolved by single-molecule FRET

TmrAB variants were site-specifically immobilized on PEGylated coverslips using a conformation-independent, TmrB-specific nanobody (Nb9F10^S63C)¹¹ conjugated to maleimide-PEG₁₁-biotin (Fig. 2a). Previous studies confirmed that this nanobody does not perturb TmrAB transport or ATPase activity^11,25. 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 difference in detection-efficiency^17,35,36 (Fig. 2–Fig. S1 and 2). FRET efficiency (E) and stoichiometry (S) were calculated from the corrected fluorescence-trajectories (see Methods, Eq. 1 and Eq. 2).

Figure 2.

Figure 2 Figure supplement 1.

Figure 2 Figure supplement 2.

FRET efficiency (E) histograms revealed two Gaussian populations corresponding to the apo and ATP-bound states (Fig. 2d,e). 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 calculated 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, indicating that the ATP-bound population at this site represents a mixture of rapidly interconverting conformations rather than a single well-defined state.

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). smFRET measurements revealed dose-dependent population shifts: TmrAB^NBD transitioned from a low-FRET apo state (E = 0.58) to 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 K_{d, ATP} values of 13 ± 1 μM for TmrAB^NBD and 2 ± 1 μM for TmrAB^PG (Fig. 3b,d), indicating saturation at ATP concentrations well below physiological levels (3 mM). For the TmrAB^PG variant, population quantification and subsequent determination of K_{d, ATP} are unreliable due to insufficient separation of the two FRET populations below 1 mM ATP, consistent with the smaller ΔE observed for this labeling configuration.

Figure 3.

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

At physiological ATP concentrations (3 mM), ∼80% of TmrAB^PG molecules populated the ATP-bound state (E = 0.86), closely matching the ATP-bound population (∼77% of 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; Fig. 1c). This discrepancy indicates 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 are clearly indistinguishable from OF^occluded within the current FRET geometry¹¹.

To test 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 catalytic mutant (TmrAB^PG_EQ), (ii) Mg²⁺ depletion using EDTA, and (iii) reverse inhibition by high concentrations of the substrate peptide RRYQKSTEL (R9L) (Fig. 4). Slow-turnover variants have previously enabled structural separation of OF^open and OF^occluded states^11,12,14,25. Mg²⁺ depletion blocks ATP hydrolysis while preserving ATP binding, allowing rapid and reversible trapping²⁵. We further hypothesized that trans-inhibition by peptide binding sterically restricts PG closure and is therefore expected to stabilize OF^open in a dose-dependent manner^26,27.

Figure 4.

In conditions lacking ATP, either in the absence of nucleotides or in the presence of ADP (3 mM ADP), 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 NBD dimerization nor PG opening, consistent with previous biochemical and structural observations^11,12. Upon ATP addition (3 mM ATP), however, the conformational landscape diverged sharply from that of wild-type TmrAB^PG. Instead of the two-state distribution observed for wild-type (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 E = 0.97 (∼14%), E = 0.86 (∼55%), and E = 0.63 (∼31%) (Fig. 4a, bottom).

As in wild-type 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 of 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 due to its drastically reduced ATP hydrolysis rate¹². Notably, the low-FRET ATP-bound population (E = 0.63) was entirely absent in wild-type. 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), thereby postulating E = 0.63 as 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 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 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 measurements¹⁴ (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, indicating that most conformational transitions occur at or below our temporal resolution.

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 conformation state occupancies and dwell times throughout the catalytic cycle of a heterodimeric ABC transporter under active turnover conditions, providing valuable insights into 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 been confined to monomeric or homodimeric systems^22–24, leaving the dynamic behavior of asymmetric, heterodimeric transporter comparatively unexplored. Here, we apply smFRET to the heterodimeric type IV ABC transporter TmrAB, extending single-molecule analysis to an asymmetric transporter system and uncovering 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, these structural rearrangements are not strictly correlated: NBD dimerization can give rise to either an outward-facing open (OF^open) or occluded (OF^occluded) conformation^11,12. This decoupling underscores the need to monitor both cytosolic and periplasmic regions to resolve the transport mechanism.

Labeling sites previously validated for PELDOR/DEER spectroscopy^10,14,15 were adapted for smFRET and rigorously benchmarked using accessible-volume (AV) simulations³⁰, fluorescence lifetime analysis, and ensemble FRET titrations. This additional validation is essential because fluorophores impose stricter steric and rotational constraints than nitroxide spin labels³². Collectively, these controls demonstrate that fluorophore attachment preserves native-like conformational behavior and ATP binding, establishing TmrAB as a robust system for quantitative single-molecule analysis. Consistent with this conclusion, fluorescence lifetime analysis showed no evidence for substantial protein-fluorophore quenching or restricted dye motion, as indicated by donor lifetime shortening and prolonged acceptor lifetimes characteristic of efficient energy transfer¹⁷.

Single-molecule measurements resolved two dominant FRET populations corresponding to apo and ATP-bound states for both reporter variants TmrAB^NBD and TmrAB^PG. ATP titrations spanning concentrations well below the reported apparent K_{d, ATP} (∼100 µM for TmrA^E523QB)¹² up to physiologically relevant levels (3 mM ATP) revealed gradual, concentration-dependent shifts between these populations, demonstrating the high sensitivity of the FRET constructs to ATP binding. Notably, apparent K_{d, ATP} values derived from smFRET (2–13 µM) were substantially lower than those obtained from ensemble FRET measurements (50–100 µM). This difference likely reflects the inherent selectively of single-molecule analyses for properly folded and catalytically competent transporters. In smFRET experiments, aggregated or inactive species can be identified and excluded during trace selection based on fluorescence intensity, stoichiometry, and photobleaching behavior, thereby enriching the analyzed population for functional molecules.

Crucially, the ATP-bound population may reflect a rapidly interconverting ensemble of conformations, rather than a single static structure, potentially including OF^open, OF^occluded, and post-hydrolysis unlocked return states (UR^asym and UR^asym*). These transitions occur faster than the ∼200 ms temporal resolution of our measurements, resulting in averaged FRET efficiencies under turnover conditions. Using three independent trapping strategies—slow-turnover catalysis, Mg²⁺ depletion, and substrate trans-inhibition—we stabilized and directly resolved a previously hidden OF^open conformation. The associated distance changes are generally consistent with cryo-EM structures¹¹, PELDOR/DEER data^10,14,15, and simulation-based predictions, suggesting that detergent-solubilized TmrAB samples a largely native-like conformational landscape.

Although detergent-solubilized and lipid nanodisc-reconstituted TmrAB exhibit similar global conformational states, the conformational space accessible to attached fluorophores may be differentially influenced by membrane-associated environments^19,39,40. In particular, fluorophores attached near the periplasmic gate may experience steric restrictions due to partial overlap with the membrane region, as suggested by AV simulations. While such effects are negligible under the detergent conditions used here, they should be carefully evaluated in future studies employing membrane-embedded systems. Single-molecule measurements further revealed that addition of the peptide substrate induces concentration-dependent shifts in the conformational equilibrium. Increasing substrate concentrations progressively stabilized the OF^open state, consistent with trans-inhibition behavior observed in human TAP1/2²⁶ and bovine ABCC1²⁷, and reflecting the finite substrate-loading capacity of the transporter²⁶.

Quantitative deconvolution of FRET populations enabled direct determination of conformational state occupancies under physiological ATP concentrations. During active turnover, TmrAB populates the IF state (∼20%), the OF^open state (∼25%), and OF^occluded/post-hydrolysis states (UR^asym and UR^asym*) (∼55%) (Fig. 5). The current reporter geometries do not allow direct discrimination between OF^occluded and post-hydrolysis states because fluorophores were placed at the noncanonical nucleotide-binding site. However, contributions from post-hydrolysis states are expected to be minimal for slow-turnover TmrAB variant (TmrAB^PG_EQ), owing to its drastically reduced ATP hydrolysis rate. To our knowledge, this represents the first single-molecule quantification of conformational equilibria for a heterodimeric ABC transporter under catalytic conditions. While these distributions broadly align with cryo-EM particle classifications^11,25, smFRET resolves fewer IF states than the cryo-EM distinction between IF^wide and IF^narrow, and additionally captures ATP-bound intermediates that are challenging to resolve structurally due to rapid interconversion.

Figure 5.

All smFRET measurements were performed at 40 °C to maintain consistency with prior biochemical studies and ensure fluorophore stability. At the physiological temperature of T. thermophilus (68 °C), absolute rates of ATP turnover and conformational transitions are expected to increase, although relative state occupancies may remain conserved if the underlying free-energy landscape is preserved. Despite a substantial fraction of static single-molecule trajectories, ATP-dependent population shifts and catalytic rates indicate that TmrAB operates near the temporal resolution limit of our measurements. Integrating smFRET-derived state occupancies with biochemical turnover rates yields an ATP-bound dwell time of approximately 300 ms, in good agreement with previous biochemical estimates^9,12.

Emerging microsecond-resolution smFRET approaches offer the potential to directly visualize short-lived intermediates within the transport cycle⁴¹. Future studies could further benefit from three- or four-color FRET strategies^42,43, which would allow simultaneous monitoring of multiple structural elements. In particular, dual labeling of the NBDs and PG could provide direct detection of the OF^occluded state, while probes placed at both canonical and noncanonical nucleotide-binding sites could capture post-hydrolysis conformational dynamics.

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, our study resolved an important aspect of the transport mechanism how chemical energy is transduced into directional transport in heterodimeric ABC systems. Integration of native lipid environments, higher temporal resolution, and substrate engagement will further illuminate the coordination of ATP hydrolysis and substrate translocation during transport.

Methods

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 (12%) were prepared using 12% (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^TM 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 (0.6 μM) 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.

Functionalization of glass slides for single-molecule FRET analysis

Glass coverslips used for TmrAB immobilization in smFRET experiments were functionalized by PEGylation as described previously⁴⁵. 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, sides 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 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 conformation transitions do not occur on timescales longer that 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^9,11,14, 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.

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 their 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 on the manuscript and proofreading.

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