Conformational equilibrium of an ABC transporter analyzed by luminescence resonance energy transfer

Abstract

Humans have three known ATP-binding cassette (ABC) transporters in the inner mitochondrial membrane (ABCB7, ABCB8, and ABCB10). ABCB10, the most studied of them thus far, is essential for normal red blood cell development and protection against oxidative stress, and it was recently found to export biliverdin, a heme degradation product with antioxidant properties. The molecular mechanism underlying the function of ABC transporters remains controversial. Their nucleotide binding domains (NBDs) must dimerize to hydrolyze ATP, but capturing the transporters in such conformation for structural studies has been experimentally difficult, especially for ABCB10 and related eukaryotic transporters. Purified transporters are commonly studied in detergent micelles, or after their reconstitution in nanodiscs, usually at nonphysiological temperature and using nonhydrolyzable ATP analogs or mutations that prevent ATP hydrolysis. Here, we have used luminescence resonance energy transfer to evaluate the effect of experimental conditions on the NBD dimerization of ABCB10. Our results indicate that all conditions used for determination of currently available ABCB10 structures have failed to induce NBD dimerization. ABCB10 in detergent responded only to MgATP at 37°C, whereas reconstituted protein shifted toward dimeric NBDs more easily, including in response to MgAMP-PNP and even present NBD dimerization with MgATP at room temperature. The nanodisc’s size affects the nucleotide-free conformational equilibrium of ABCB10 and the response to ATP in the absence of magnesium, but for all analyzed sizes (scaffold proteins MSP1D1, MSP1E3D1, and MSP2N2), a conformation with dimeric NBDs is clearly preferred during active ATP hydrolysis (MgATP, 37°C). These results highlight the sensitivity of this human ABC transporter to experimental conditions and the need for a more cautious interpretation of structural models obtained under far from physiological conditions. A dimeric NBD conformation that has been elusive in previous studies seems to be dominant during MgATP hydrolysis at physiological temperature.

Significance

Our understanding of conformational changes underlying the molecular functioning of proteins has heavily relied on x-ray or cryo-EM atomic models that usually represent static snapshots of proteins trapped in conformations promoted by necessary experimental conditions, which are often nonphysiologically relevant. For proteins that have shown large conformational flexibility, such as ATP-binding cassette transporters, the interpretation of certain conformations is controversial. Here, we used luminescence resonance energy transfer to demonstrate a dramatic effect of experimental conditions on the conformational equilibrium of the ABCB10 transporter. This spectroscopic approach is notably useful to examine protein’s dynamic conformational changes in real time, at physiological temperature, allowing us to clearly detect for the first time a preferential switch toward dimeric NBDs during ABCB10’s active ATP hydrolysis.

Introduction

ATP-binding cassette (ABC) transporters are membrane proteins found in all kingdoms of life (1,2). Their basic architecture consists of two highly conserved nucleotide binding domains (NBDs) that bind and hydrolyze ATP, and two transmembrane domains (TMDs) that constitute the pathway for the passage of substrate. Depending on the organism, these domains can be expressed as a single polypeptide (full transporter), as half-transporters that must dimerize to form a functional transporter, or as four subunits that must form a tetramer (3). Humans mainly express full transporters (like the multidrug resistance P-glycoprotein) and some half-transporters that are mostly found in the membranes of organelles (such the homodimeric ABCB7, ABCB8, and ABCB10, in the mitochondrial inner membrane) (3,4,5). The commonly accepted molecular mechanism for ABC exporters suggests a transition from an inward-facing (IF) conformation with separated NBDs that allow space for the entrance of substrate to a binding cavity in the TMDs, to an outward-facing (OF) conformation with dimeric NBDs capable of ATP hydrolysis and a TMD reorganization that allows the release of substrate to the other side of the membrane (6,7,8). Other intermediate conformations, including occluded states with closed NBDs and TMDs, also exist. The exact molecular triggers of such transitions and the magnitude of the conformational changes remain controversial. For example, ATP binding to the NBDs is traditionally believed to trigger NBD dimerization, although some studies where ATP analogs have failed to induce dimerization have been interpreted as substrate binding being needed for the transition (9,10,11). The trigger from IF to OF is also unclear, with some evidence suggesting that binding of ATP is enough to induce NBD dimerization and OF conformation, whereas others suggest that the hydrolysis of ATP is the actual trigger (12,13). Another widely disputed issue is how far do the NBDs separate during the ATP hydrolysis cycle, with some models proposing a separation of tens of Angstroms (monomer-dimer models) and others proposing instead that the NBDs do not separate (constant-contact models) (14,15,16,17). Initial x-ray structural models showed widely separated NBDs for nucleotide-free transporters in detergent micelles, but the now widespread use of lipid nanodiscs for reconstitution of the purified transporters into a lipid bilayer, has coincided with a more common observation of IF structures with narrower NBD separation, fueling a growing concern that the small size of the nanodiscs might restrict the protein’s conformational changes. A recent spectroscopic study has shown evidence that the bacterial transporter MsbA can adopt a conformation with widely separated NBDs in the plasma membrane of cells depleted of ATP, suggesting that this conformation can indeed occur in the cells (14). Despite years of effort trying to understand how these proteins work at the molecular level, scientific studies often yield contradictory evidence (18,19). This confusion in part may arise from combining data from several different proteins, studied with various techniques that involved the use of very dissimilar experimental conditions, and often these conditions are not physiologically relevant.

Here, we have used ABCB10 as a model protein to study how the NBD dimerization is affected by ATP binding and hydrolysis and for determining the effect of usual experimental conditions such as nonhydrolyzable ATP analogs, low versus physiological temperature, and the use of detergent micelles versus nanodiscs of different sizes. Four x-ray crystallography structures of ABCB10 were published in 2013, and all of them show IF conformation even with the ATP analogs AMP-PNP or AMP-PCP bound to the NBDs (9). Cryoelectron microscopy (cryo-EM) structures of ABCB10 (nucleotide-free and biliverdin [BV] bound) in detergent micelles were recently published and they look like the known IF structures (20). We are using a complementary spectroscopy approach (luminescence resonance energy transfer [LRET]) that allows us to study the transporter’s structural changes in real time, while the protein undergoes ATP hydrolysis at physiological temperature, or in response to other relevant experimental conditions. To monitor the changes in NBD separation we chemically attach reporter probes (donor and acceptor) to a single cysteine introduced at a desired position using a previously created functional cysteine-less ABCB10 (21). Since ABCB10 is a homodimer, the newly introduced single cysteine is present in each monomer, allowing the distance measurements between the donor and acceptor probes attached to the same molecule. According to available x-ray models of ABCB10 in IF and a prediction made in AlphaFold for the OF conformation, position Q706 could be a good location to attach the probes to monitor these putative conformational changes at the NBD level during the ATP hydrolysis cycle. Conformations with separated NBDs could have a Cα-Cα distance of ∼25–35 Å (depending on the PDB structure considered), whereas the translational and rotational movements that lead to NBD dimerization would increase this distance to ∼45 Å (Fig. 1). Thus, for the specific position monitored in this study, NBD dimerization is observed as an increase in the distance. LRET data have an intrinsic uncertainty due to the unknown orientation of the reporter probes attached to the thiol group in the protein. Fig. S1 shows estimated distances between thiol groups in the Q706C mutant, which are ∼1–3 Å away from the estimated Cα-Cα distances, and the chemical structure of the LRET probes we use, which are relatively small. The experimental LRET distance measurements can be used in the future as constraints to generate atomic models of the protein with attached probes by molecular dynamics simulations. Nevertheless, since the efficiency of energy transfer depends on the sixth power of the distance, LRET is an exceptionally sensitive technique to monitor distance changes associated with conformational changes in a protein (see materials and methods). Thus, this study can efficiently report about the dynamic changes experienced by an ABC transporter during the ATP hydrolysis cycle, beyond focusing on absolute distance values. For the ABCB10-Q706C protein, changes of the measured experimental distance from a smaller value toward a larger value is a good indication of a transition from separated to dimeric NBDs.

Figure 1.

Structural representation of the ABCB10 homodimer. Monomers in green and cyan. TMDs, transmembrane domains; NBDs, nucleotide binding domains. Position of Q706 residue in magenta, with distance between the α carbons of these two residues indicated for each conformation. Models in inward-facing conformation are based on PDB crystal structures 4AYW (AMP-PNP bound) and 3ZDQ (nucleotide-free). The dimeric NBD conformation was modeled in the AlphaFold server with bound ATP (in red), and Mg (orange spheres).

Lanthanide-based probes are valuable for studying biological systems, including their usefulness for imaging, immunoassays, and many other applications (22). In the past, LRET probes strategically placed in regions of interest have been used to quantify conformational changes of various proteins. For example, LRET between a terbium chelate and fluorescein attached to cysteines was used to measure distances between the identical subunits in a voltage-dependent tetrameric potassium channel (23). A year earlier, LRET between a nucleotide terbium chelate and a rhodamine attached to a cysteine had been used to measure the lever arm swing in muscle myosin in conditions that imitated different states during the ATP hydrolysis cycle and in actin-bound states (24). LRET has also been used to measure distance changes induced by ligand binding to glutamate receptors (25,26,27), as well as for distance changes associated with pH-induced opening of a mammalian inward-rectifier potassium channel in liposomes (28). Specifically regarding the ABC transporters field, we have previously used LRET in the archaea MJ0796 NBDs to answer some fundamental questions about the requirements for ATP-induced dimerization, and to evaluate conformational changes in the lipid A exporter MsbA and the murine multidrug resistance P-glycoprotein, including comparisons between detergent micelles and nanodiscs, effect of the temperature, and response to substrate binding (29,30,31,32,33,34,35). We are now expanding these studies to a human transporter and describing the methodology in more detail, to illustrate the power and simplicity of LRET for studying conformational changes of membrane proteins under physiologically relevant conditions.

Our data show that the population of ABCB10 molecules displays a dynamic equilibrium that is highly sensitive to experimental conditions. Various experimental conditions commonly used during x-ray crystallography and cryo-EM studies failed to induce NBD dimerization, but when ABCB10 is at physiological temperature, with its natural substrate (MgATP), and in a phospholipid membrane, the equilibrium shifts toward a longer distance conformation compatible with dimeric NBDs. The size of the nanodiscs also influences the conformational equilibrium of nucleotide-free ABCB10 and its response to ATP. Our experimental findings illustrate the need to study these transporters under more physiologically relevant conditions and suggest a more cautious interpretation of available atomic models.

Material and methods

Protein production, reconstitution, and ATPase activity

Human ABCB10 without the mitochondrial targeting sequence (ABCB10^152−735; named wild-type [WT] in this work) and its single-cysteine mutant (C215S-C224L-C582G-Q706C; named Q706C) were expressed in bacteria (pET19 plasmid) and purified by affinity chromatography according to our previously published procedures, with some modifications (9,21,36). In brief, C-terminal His-tagged proteins were expressed overnight in E. coli (Rosetta 2, DE3) in autoinduction medium at 20°C. Cells were grown at 37°C until OD₆₀₀ was near 1.6, after which the temperature was dropped to 20°C, and the cells were harvested 24 h later. We have found that these expression conditions improve the yield and stability of the purified proteins. The membranes were solubilized in purification buffer (300 mM NaCl, 20 mM Tris-HCl [pH 8], 10% glycerol, 0.5 mM tris(2-carboxyethyl) phosphine) with 1% dodecyl-maltoside (DDM) (Inalco, CA, USA), 0.1% cholesteryl hemisuccinate (CHS) (Avanti Research, AL, USA), and 1 mM phenylmethylsulfonyl fluoride for 1 h at room temperature. After batch incubation with Ni-NTA resin (QIAGEN, MD, USA), the resin was transferred to a column and washed with buffer containing 20 mM imidazole and 0.06%/0.006% DDM/CHS, and eluted with 300 mM imidazole in the same buffer (pH 7.5). Elution fractions were pooled for His-tag removal during overnight dialysis at 4°C, followed by a second Ni-NTA purification step to remove the rTEV protease. The concentration of the final purified protein was determined by bicinchoninic acid (Pierce Protein Biology, MA, USA; molar concentration calculated with the monomeric molecular weight of 65,905 Da). Protein aliquots were stored at −80°C. Reconstitution in nanodiscs was done following standard protocols, using E. coli polar lipids extract in chloroform (Avanti Research, AL, USA; dried with a stream of nitrogen gas, left under vacuum overnight, and then resuspended in buffer with 3% DDM) and membrane scaffold protein (MSP) of different sizes (MSP1D1, MSP1E3D1, and MSP2N2; all plasmids from Addgene, MA, USA). The molar ratio of ABCB10 to MSP was 1:5, and the molar ratio of lipids to MSP was empirically determined (1:50 for MSP1D1, 1:100 for MSP1E3D1, and 1:240 for MSP2N2). Membrane scaffold proteins were produced following previously published protocols (37). ATPase activity was determined using a colorimetric assay measuring released Pi (38). We used 0.5–1 μg of transporter (usually contained in less than 5 μL of sample) in a 25 μL total reaction volume, in 96-well plates. The composition of the buffer was 100 mM KCl, 20 mM Tris-HCl (pH 7.5), with a final concentration of 2 mM ATP and 4 mM MgCl₂. The buffer was supplemented with 0.06%/0.006% DDM/CHS for samples in detergent. To study the activation by substrate, we added BV (Sigma-Aldrich, MO, USA; stock made in DMSO) to a saturating final concentration of 5 μM. Duplicate samples were incubated for 50 min at 37°C, and the reaction stopped by addition of SDS, followed by color development. For measurements at 20°C, we used 2 μg of transporter per well and longer incubation times. After color development, absorbance (850 nm) was measured in a microplate reader (SpectroStar Nano, BMG LabTech, Ortenberg, Germany). For ATPase measurements we decided to use incubation times compatible with the full duration of typical luminescence resonance energy transfer experiments (described below). Calculations done assuming an excessive ATP hydrolysis rate of 500 nmol Pi/mg protein.min during basal hydrolysis at 37°C suggest that, during a 50-min incubation, 1 μg of ABCB10 could consume ∼50% of the available ATP, lowering [ATP] to ∼1 mM by the end of the assay, which is still near saturating if we consider the ∼0.3 mM K_m we have measured for ABCB10 under comparable conditions (39). Notice that, for measurements with BV, when the ATPase activity is higher, we adjust the amount of ABCB10 to 0.5 instead of 1 μg per well. Our control experiments indicate that the protein is stable during the duration of the ATPase assays (values for activity reported here are like those obtained during shorter incubations). Most reagents were purchased from Fisher (MA, USA), Thermo Scientific (MA, USA), and Sigma-Aldrich (MO, USA).

LRET experiments

One hundred microliters (6–10 μM) of the single-cysteine mutant (ABCB10-Q706C) in detergent micelles was labeled with a molar excess (50 μM) of the luminescent donor (LanthaScreen maleimide Tb-chelate, ThermoFisher Scientific) and the fluorescent acceptor (Bodipy-FL maleimide, ThermoFisher Scientific) for 30 min at room temperature, and the free, unreacted labels removed by size-exclusion chromatography (Bio-Rad ENrich 650 column equilibrated with nanodisc buffer: 100 mM KCl, 20 mM Tris-HCl [pH 7.5], supplemented with 0.06%/0.006% DDM/CHS for samples in detergent). For LRET experiments with reconstituted ABCB10, the protein was labeled after reconstitution using a similar procedure and using nanodisc buffer without detergent for the size-exclusion chromatography step. Labeling did not affect the ATPase activity of the Q706C protein, and the order of labeling (before or after reconstitution) had no effect on the experimental results. The elution fractions containing the transporter (see Fig. S2) were combined and directly used for the experiments. We also studied samples labeled with only Tb-chelate donor, as required for appropriate data analysis (see below).

Emission spectra (see Fig. S2) were recorded in a spectrophotometer (QM-4CW, Photon Technology International, NJ, USA) with 337 nm excitation. LRET experiments were done as described before (40). In brief, a 150 μL of sample in a quartz cuvette was placed in a temperature-controlled Easylife L phosphorescence lifetime spectrometer (Horiba, CA, USA), with bandpass filters (ThorLabs, NJ, USA) for Tb-chelate excitation (340/10 nm) and for collection of either the Tb-chelate (490/10 nm) or the Bodipy (510/10 nm) emission. The emissions were recorded 200 μs after a short (4 μs) excitation pulse (gated mode) to minimize artifacts caused by direct excitation of the acceptor, allowing preferential acquisition of the fluorescence of Bodipy that is being excited by long-lived emission (ms) of the luminescent donor (sensitized acceptor emission). This sensitized acceptor emission depends on the distance that separates the probes: a short distance between the probes results in a high fluorescence acceptor emission signal that decays fast and, conversely, a longer distance between the probes results in a lower intensity that lives longer (41). For the position Q706C labeled in this study, dimerization of the NBDs is expected to increase the distance between the probes due to a rotational movement in these domains as the protein moves from IF to OF conformation (Fig. 1). The LRET signals were acquired (FluoreScan software) using two different modes: time-based mode and phosphorescence decays. When using the time-based mode, we followed the kinetics of conformational changes produced in response to manual additions of ATP, or other molecules, to the cuvette (excitation set at 100 Hz, and the acceptor emitted intensity is integrated to record ∼1 data point/s for several minutes). Once samples have reached steady state for the desired experimental condition, we acquired LRET phosphorescence decays to determine the time that the sensitized acceptor emission takes to relax after a single excitation pulse (excitation also set to 100 Hz, but 1 data point is successively acquired until information for up to 10 ms is recorded, resulting in 1 point every 0.8 μs). These LRET decays were recorded by triplicate for each one of the various experimental conditions used to mimic different steps throughout the ATP hydrolysis cycle.

In a typical experiment, the sample is initially measured in the absence of nucleotide (Apo). Then, 0.6 μL of a 0.5 M sodium ATP (adenosine 5′-triphosphate disodium salt) (Fisher, MA, USA) stock was added to the cuvette for a 2 mM final concentration of nucleotide. Alternatively, we used the nonhydrolyzable ATP analogs AMP-PNP (β-γ-imidoadenosine 5′-triphosphate lithium salt hydrate) (Roche, Penzberg, Germany) and AMP-PCP (β-γ-methyleneadenosine 5′-triphosphate disodium salt) (Sigma-Aldrich, MO, USA). Third, we added MgCl₂ (4 mM final concentration) to promote the active ATP hydrolysis state. Finally, we added sodium orthovanadate (VO₄; 0.25 mM final concentration) to inhibit ATP hydrolysis and trap the NBDs in the dimeric conformation. The pH of all nucleotide stocks was adjusted to pH 7.5, were aliquoted, and kept at −80°C until use. Vanadate stock was made as described previously (42). In some experiments, biliverdin (Sigma-Aldrich, MO, USA) was added to a final concentration of 5 μM to evaluate the changes induced by substrate. The protein samples were allowed to equilibrate for 5 min at the experimental temperature (20 or 37°C), and after each addition, before LRET decays were acquired (unless otherwise stated).

LRET data analysis

By fitting the LRET decays with exponential functions, we can obtain information about the lifetime (τ) and amplitude (A) of each exponential function present in the data (41). The lifetime of the acceptor-sensitized emission (τ_DA) can be used to calculate distances (r) according to the equation: r = R₀ (E⁻¹ − 1)^1/6, where E = 1 − τ_DA/τ_D. The donor-only lifetime (τ_D) was obtained from fitting the decays of protein samples labeled only with the Tb-chelate, acquired under each of the evaluated experimental conditions (as a reference, for the Apo state in detergent, the τ_D was 2111 ± 4 μs at 20°C and 1818 ± 13 μs at 37°C). The Förster distance (R₀) for the terbium-Bodipy-FL pair is 41 Å, so this pair is useful to measure distance changes in the ∼25–55 Å linear range (31). A good fitting by a single exponential function would indicate a single conformational state (distance) in our sample. However, as we have observed before for other ABC transporters (29,31), the decays are more complex and require multiple exponential functions for an appropriate fitting, indicating the existence of multiple coexisting conformations. The exponential series method (Felix 32 software; Photon Technology International, NJ, USA) allows a multiple exponential fitting of the phosphorescence decays, providing a distribution of the lifetimes (τ) and amplitude (A) components for each exponential function (see Figs. S3 and S4 for additional details) (43). The goodness of fit was evaluated by residual errors plot (it must be flat) and a low chi-squared value. For the analysis of our data, we have allowed the exponential series method to use 200 possible lifetimes logarithmically distributed from 1 to 3000 μs (which is within the range of lifetimes we can measure using Tb-chelate as a donor), to report the amplitude values found for each of those lifetimes. We then convert the lifetimes to distances, as described above, to obtain the distribution of distances for each decay. We have analyzed these distance distributions in two ways: 1) multipeak Gaussian fit (Origin 2020, Origin Lab software) to obtain the center of each peak in the distribution (which tell us the average distance represented by each peak), and the area of each peak (which we used to calculate the relative proportion of each peak within the distribution for each one of the experimental conditions). The peak areas of multiple independent experiments for each experimental condition were averaged to quantify the relative contribution of each peak. 2) Each of the distance distributions was normalized to the amplitude of their highest peak and then all the normalized distance distributions of the multiple data acquired for each experimental condition were averaged, resulting in the averaged distance distribution that summarizes the distances that are consistently detected for each condition. For clarity, standard error bars have been omitted for all these averaged distance distribution plots, but an example displaying error bars is presented in Fig. S5. For comparison, selected data were also analyzed using the alternative maximum entropy method (Felix 32 software; Photon Technology International, NJ, USA) and the results were very similar to those obtained with the exponential series method (Fig. S6), which is in agreement with the previously reported similar ability of both methods for recovering underlying distributions from complex fluorescence decay data (43).

Statistical analysis and data presentation

Experiments were done with at least three fully independent purifications for each WT and Q706C protein. Data points represent the average and standard deviation of at least three independent experiments, unless otherwise indicated. Comparisons between conditions were done with a paired t-test in Origin Lab. Plots were graphed in Origin Lab and PDB structures visualized with PyMOL. The model for ABCB10 in OF conformation was done with AlphaFold 3 including ATP and Mg as ligands (44).

Results

ATPase activity of the single cysteine Q706C mutant in detergent micelles and nanodiscs

We have evaluated the ATP hydrolysis activity of our purified ABCB10-WT and single cysteine ABCB10-Q706C transporter in detergent micelles and after reconstitution in nanodiscs of different diameters. Since we first published our methodology to produce the human ABCB10 transporter in bacteria (36), we have made improvements that have led to an increased stability of the purified protein in detergent micelles. We are now expressing the transporter in autoinduction medium at 20°C, instead of Luria Broth at 30°C, and the purified transporter shows robust basal, and substrate activated ATPase activity in DDM-CHS detergent micelles at physiological temperature (Fig. 2 A, left plot). In addition, we have found that positioning of the poly-histidine purification tag in the C-terminus (instead of the N-terminus) is essential for the stability of purified mutant proteins such as the Q706C single cysteine mutant used in this study, which also presents ATPase activity in detergent micelles (Fig. 2 A, right plot). The mutant conserves the response to substrate observed for the WT transporter and labeling of the single cysteine in this position does not affect the ATPase activity, indicating that this position is a good reporter to track ABCB10’s conformational changes. In addition, we have reconstituted the WT and Q706C proteins in nanodiscs of different diameters (∼10 nm if using MSP1D1, ∼12 nm for MSP1E3D1, and ∼16 nm for MSP2N2) (37). As a reference, the basal and substrate activated ATPase activity of the WT transporter in each nanodisc system is shown in Fig. 2 B. The data show that the basal ATPase rate and the degree of activation by BV depends on the size of the nanodiscs, with MSP1D1 nanodiscs presenting larger basal activity and lower activation by BV than MSP2N2 nanodiscs. The reconstituted Q706C transporter shows a similar tendency (Fig. 2 C), suggesting that this single cysteine mutant provides a good representation of the behavior of the WT transporter. Our data also show that labeling does not affect the ATPase activity of the reconstituted Q706C protein in any of the nanodiscs systems analyzed. Our measurements at 20°C (Fig. 2 D) indicate that the ATPase activity of the transporter in MSP1E3D1 nanodiscs is about six times lower than that at physiological temperature, whereas for the protein in detergent the temperature effect is more dramatic, with a decrease of about an order of magnitude.

Figure 2.

ATP hydrolysis activity of ABCB10-WT and the single cysteine mutant ABCB10-Q706C in the absence (−) and presence (+) of substrate (BV). (A) ATPase activity in detergent micelles for the WT (left plot) and Q706C unlabeled and labeled (L) mutant (right plot). (B) ATPase activity of WT protein reconstituted in nanodiscs of different diameter. (C) ATPase activity of the unlabeled and labeled (L) Q706C reconstituted in nanodiscs of different diameters. All measurements were performed at 37°C. (D) ATPase activity of Q706C at 20°C, either in detergent micelles or in MSP1E3D1 nanodiscs. Values are mean and standard deviation of at least two independent protein purifications. p values for paired t-test (one tail) are shown. p values between nonlabeled and labeled samples were all >0.05.

Conformational changes in detergent micelles

We first studied the conformational changes of the transporter in DDM-CHS detergent micelles (Fig. 3). A sample of the labeled Q706C protein was placed in the cuvette of a lifetime fluorometer at 37°C. Initially the sample was in Apo condition (no nucleotide). The time base recordings (Fig. 3 A) show the changes in LRET intensity after consecutive manual addition of nucleotide (Apo to ATP, to imitate prehydrolysis state), then magnesium (ATP to MgATP, to promote ATP hydrolysis), and finally vanadate (MgATP to MgADPO₄, to trap the posthydrolysis state). Under these conditions, ATP did not seem to induce a stable change (top plot) until Mg was added, when a small increase in intensity was observed (middle plot). Addition of vanadate did cause a pronounced increase in intensity (lower plot). To quantify the distance changes behind the observed LRET intensity changes, we also acquired LRET decays (Fig. 3 B) in the initial Apo condition , and after steady state had been achieved following the addition of ATP , Mg , and vanadate . These representative decays show identical curves in Apo (black) and ATP (red), with a very sharp and fast decaying component that diminished in the presence of Mg (blue) and almost disappeared with vanadate (orange). The very fast component likely results from ABCB10 molecules adopting a conformation where the donor and acceptor probes are nearby, as predicted from available structures with separated NBDs (Fig. 1). After incubation with vanadate, this fast component was replaced by a slower signal, consistent with the longer distance between the donor and acceptor probes predicted to occur in ABCB10 molecules with dimeric NBDs. Exponential fitting analysis of these decays (Fig. 3 C) confirmed that, in Apo and ATP, most of the LRET signal came from molecules adopting a conformation where the probes attached to Q706C in the NBDs were separated by ∼23 Å (where the main peak is centered in these distance distributions). Since the Förster distance for the Tb-Bodipy pair is 41 Å, this ∼23 Å distance is near the lower limit for the linear measuring range with this pair. Such a short distance for the Q706C position is consistent with crystal structures of ABCB10 in IF conformation (see PDB: 4AYW in Fig. 1) and, in this article, we are referring to this ∼23 Å peak as the “Open” conformation, with separated NBDs. In the presence of MgATP, there was a shift in the distance distribution, where the amplitude of the ∼23 Å peak decreased, whereas a peak centered at ∼47 Å became the most prominent. This distance is compatible with the predicted 45 Å distance between Q706 α carbons in the ABCB10 with dimeric NBDs, which we are referring to as the “Closed” conformation in this article. Incubation with vanadate shifted the distance distribution almost completely toward the closed conformation, as expected for the stabilization of the NBDs dimer.

Figure 3.

LRET experiments for ABCB10-Q706C in detergent micelles. (A) Time base recordings follow the kinetics of LRET intensity in response to the conformational changes induced by manual addition (indicated by vertical arrows) of ATP (top), MgCl₂ (middle), or vanadate (VO₄, bottom). (B) LRET decays in semilog scale show the decrease in the sensitized emission of the acceptor after a single excitation pulse. These decays were acquired at the steady-state time frames for each condition (Apo, ATP, MgATP, and MgADPVO₄), as represented by circled numbers 1 through 4 in (A). (C) Exponential series method analysis of the decays presented in (B). The lifetime values were converted to distances as described in materials and methods. The empty circles show the amplitude for each one of the 200 logarithmically spaced lifetimes (converted to distances) of the multiexponential fit, whereas the continuous lines show the Gaussian fits of the multiple peaks in the distributions. The inserted structural models represent the possible conformations compatible with Open (top; PDB: 4AYW model) and Closed (bottom; AlphaFold model) conformations corresponding to the indicated peaks in the distributions. Experiments representative of multiple experiments performed at 37°C.

To perform a quantitative analysis of the observed shifts in distance distributions, we did multipeak Gaussian fittings to determine the relative area for each peak. Fig. 4 A shows the relative proportion of the two main peaks (Open conformation, ∼23 Å; Closed conformation, ∼47 Å) averaged from multiple independent experiments at physiological temperature. The Closed conformation increases at the expense of the Open conformation as the protein transitions from Apo (∼90% Open) to MgATP (∼20% Open; ∼60% Closed), and then vanadate (∼90% Closed). The average distance distribution from multiple independent experiments (Fig. 4 B) shows mostly a single conformation in Apo (Open), ATP (Open), and vanadate (Closed), whereas in MgATP there is a mix of Open, Closed, and other intermediate conformations. Together, these results suggest that a proportion of ABCB10 molecules in detergent micelles have dimerized NBDs during active hydrolysis conditions (MgATP, at physiological temperature), but ATP without Mg was not able to promote NBD dimerization. Vanadate (VO₄) stabilizes the NBD dimer by replacing the phosphate (PO₄) after ATP hydrolysis has taken place, therefore trapping the catalytically active ABCB10 molecules in the Closed conformation.

Figure 4.

Effect of temperature and nonhydrolyzable ATP analogs on the conformational changes of ABCB10-Q706C in detergent micelles. (A) Relative proportional area of the two main peaks observed in the distance distributions at 37°C (circles represent the area of the peak centered at ∼23 Å, identified as Open conformation, and triangles represent the area of the peak centered at ∼47 Å, identified as Closed conformation). Values are the mean and standard deviation of multiple experiments performed with at least 3 independent protein purifications at the indicated conditions (Apo, ATP, MgATP, and MgADPVO₄). (B) Average distance distributions at 37°C, determined from multiple independent experiments under the indicated conditions. The amplitudes were normalized to the maximum amplitude of each distribution before performing the average. Error bars are omitted for clarity. (C) Relative proportional area of the Open and Closed peaks, as described in (A), but for experiments performed at 20°C. Notice that the values for the vanadate condition were acquired after a 1-h incubation instead of the 5 min used for the rest of the conditions. (D) Representative time base recording showing the slow LRET intensity increment after manual addition of vanadate (vertical arrow) to the protein in the presence of MgATP at 20°C. (E) Average distance distributions as described in (B), but in response to the nonhydrolyzable ATP analogs AMP-PNP (left) and AMP-PCP (right). ATP was added at the end of each experiment as a positive control (VO₄ + ATP).

We have carried out similar experiments at room temperature and found that at this nonphysiological temperature most of the ABCB10 molecules remained in Open conformation even in the presence of MgATP (Fig. 4 C). This is in marked contrast to the significant shift toward the Closed conformation induced by MgATP at physiological temperature (compare Fig. 4, A and C). Incubation with vanadate at 20°C promotes a slow conformational change (Fig. 4 D), which is compatible with a slower ATP hydrolysis rate at lower temperature. Notice that, even after almost 20 min, the change in intensity had not yet reached steady state. After a prolonged (1 h) incubation with vanadate, ∼80% of Closed conformation had been reached (Fig. 4 C). Control experiments suggest that longer incubation times with MgATP (Fig. S7) are needed to allow a partial shift toward the peak corresponding to the Closed conformation, an observation that is compatible with the low ATPase activity detected for the protein in detergent at 20°C (Fig. 2 D). These results suggest that at room temperature and in detergent, the NBD dimerization in response to MgATP is slower and less efficient than that detected at physiological temperature, but the Closed conformation peak can become significantly populated after this conformation is allowed to be stabilized by vanadate. We have also investigated the ability of two nonhydrolyzable ATP analogs to promote NBD dimerization. X-ray crystal structures of ABCB10 with bound AMP-PNP or AMP-PCP have shown only IF conformations, and our results in detergent micelles have confirmed that these analogs failed to induce conformational changes at physiological temperature (Fig. 4 E) or room temperature (Fig. S8). The average distance distributions of multiple independent experiments looked very similar to the Apo condition until ATP was added at the end of the experiments, as a control to prove that the ABCB10 protein indeed could undergo conformational changes. Therefore, our data indicate that AMP-PNP and AMP-PCP cannot imitate the effect of the physiological substrate (MgATP) when ABCB10 is in detergent micelles.

Conformational changes in MSP1E3D1 nanodiscs

Lipid nanodiscs have become a very useful tool for the study of membrane proteins in a phospholipid bilayer. Therefore, we decided to reconstitute our detergent purified ABCB10 in nanodiscs to study the conformational changes of the transporter in a more “native-like” environment. Fig. 5 shows the kinetics of conformational changes induced by successive manual addition of ATP, Mg, and vanadate, and the corresponding LRET decays for representative experiments performed at physiological and room temperature in nanodiscs made with the membrane scaffold protein MSP1E3D1. At physiological temperature, the LRET intensity increased successively after each addition (Fig. 5 A), and the increment elicited by incubation with vanadate was clearly faster than the slow increment observed for the protein in detergent (compare with Fig. 3 A). The LRET decays acquired at physiological temperature (Fig. 5 B, left plot) show that the ATP decay (red) is different from the Apo decay (black), which indicates that ATP without Mg can induce conformational changes when ABCB10 is in these nanodiscs. Addition of Mg further changes the decay (blue) by drastically reducing the fast initial component to a level close to the one observed for the decay of the protein incubated with vanadate (orange). When experiments were carried out at room temperature (Fig. 5 B, right plot) the initial addition of ATP had no obvious effect, which can be seen by the identical Apo (black) and ATP (red) decays. Addition of Mg induced a change in the decay (blue), but not to the same level as the MgATP decay acquired at 37°C. Interestingly, the decay obtained after incubation with vanadate (orange) looks very similar at both temperatures, suggesting that the conformation stabilized by vanadate is comparable, independently of the temperature. However, the time base recording (Fig. 5 C) clearly shows that the change induced by vanadate is slower at 20°C, which is again consistent with a lower ATP hydrolysis rate at room temperature.

Figure 5.

Representative LRET experiments for ABCB10-Q706C reconstituted in MSP1E3D1 nanodiscs. (A) Time base recordings show the kinetics of the changes induced by successive manual addition (vertical arrows) of ATP (top), MgCl₂ (center), and vanadate (bottom), at 37°C. (B) LRET decays of experiments shown in (A) at 37°C (top) and of the same protein sample studied at 20°C (bottom). (C) Time base recording of the response of the sample at 20°C after manual addition (vertical arrow) of vanadate to the protein in the presence of MgATP.

Quantitative analysis of the LRET decays of multiple independent experiments (Fig. 6) show that, similarly to the behavior observed in detergent, the protein in nanodiscs preferentially adopts either the Open (∼23 Å) or Closed conformation (47 Å), depending on the experimental condition. The relative proportion of each of these two main conformations measured at physiologically relevant condition is shown Fig. 6 A. An initial interesting observation is that in Apo the ∼23 Å Open conformation is only ∼60% (whereas this value was ∼90% for the protein in detergent), with less than 10% of Closed conformation. Therefore, only ∼70% of the signal comes from Open and Closed conformations, whereas the ∼30% remaining must be adopting other conformation(s). The average distance distribution plot in Fig. 6 B shows a clear contribution of a new distinctive peak centered at ∼32 Å in the Apo state (black trace), which is the conformation adopted by that remaining ∼30% in these conditions. This peak centered at ∼32 Å was essentially absent for the Apo protein in detergent (Fig. 4 B). This ∼32 Å distance is closer to the 35 Å distance measured in the IF ABCB10 crystal structure in nucleotide-free state (PDB: 3ZDQ, see Fig. 1), suggesting that the nucleotide-free ABCB10 transporter reconstituted in nanodiscs displays an equilibrium between different IF conformations. Fig. 6, A and B clearly shows the increment in Closed conformation induced by ATP (∼40%) when the transporter is in MP1E3D1 nanodiscs, whereas ATP had essentially no effect in detergent. In the presence of MgATP, the closed conformation became predominant (∼80%), with vanadate further promoting the transition from the Open conformations to the stabilized NBD dimer.

Figure 6.

Effect of temperature and nonhydrolyzable ATP analogs on the conformational changes of ABCB10-Q706C reconstituted in MSP1E3D1 nanodiscs. (A) Relative proportional area of the Open and Closed conformations at 37°C. (B) Average distance distributions at 37°C. (C) Relative proportional area of the Open and Closed conformations at 20°C. (D) Average distance distributions at 20°C. (E) Average distance distributions in response to the nonhydrolyzable ATP analogs AMP-PNP (left) and AMP-PCP (right). All data represent averages from multiple experiments performed with at least 2 independent protein purifications.

The temperature had an important effect for ABCB10 reconstituted in nanodiscs (Fig. 6, C and D). At room temperature, ATP failed to induce conformational changes unless Mg was present (confirming the identical Apo versus ATP decays at 20°C, as shown in Fig. 5). MgATP induces transition to the Closed conformation, but only up to ∼30% (instead of the ∼80% observed at physiological temperature). Since the formation of the NBD dimer is required for ATP hydrolysis, the equilibrium preference toward Open conformations at room temperature could relate to a lower ATPase activity of the transporter at 20°C compared with physiological temperature, at which the Closed conformation predominates. An additional interesting finding is presented in Fig. 6 E, which summarizes the effect of nonhydrolyzable ATP analogs on the reconstituted transporter at physiological temperature. The average distance distributions of multiple independent experiments show that AMP-PNP alone did not induce any major effect, but in the presence of Mg it was very efficient at promoting NBD dimerization, causing a clear shift toward the Closed conformation (Fig. 6 E, left plot). This effect in nanodiscs is in sharp contrast to the lack of effect of AMP-PNP for the protein in detergent, even in the presence of Mg (Fig. 4 E). On the contrary, AMP-PCP was not effective at inducing the Closed conformation of ABCB10 in nanodiscs under any tested condition (Fig. 6 E, right plot). The addition of ATP at the end of the experiments proved that the protein was indeed functional, and that ATP but not AMP-PCP could induce conformational change. We also evaluated the response to these ATP analogs at room temperature and found no effect even with MgAMP-PNP (Fig. S8).

In summary, our data demonstrate that the conditions that have been used for x-ray crystallography and cryo-EM studies (detergent, nonphysiological temperature, and nonhydrolyzable ATP analogs) failed to induce NBD dimerization in ABCB10, explaining the difficulty to capture a putative OF conformation under such experimental conditions. The conformational changes of this transporter are highly sensitive to the temperature and the presence of a lipid bilayer, with the Closed conformation being favored in nanodiscs during active hydrolysis conditions (MgATP, at physiological temperature).

Does the size of the nanodisc affect the behavior of ABCB10?

Given the stated sensitivity of ABCB10 to the experimental conditions and the increasing controversy about a possible restriction of protein’s conformational changes due to the small size of the nanodiscs, we decided to compare the response of ABCB10 reconstituted in nanodiscs of different sizes. We initiated these studies (all experiments above) using the membrane scaffold protein MSP1E3D1, which generates nanodiscs of ∼12 nm diameter. Next, we used MSP1D1 for smaller size nanodiscs (∼10 nm diameter) and MSP2N2 for larger size nanodiscs (∼16 nm diameter). Fig. 7 shows the average distance distribution of multiple independent experiments performed at physiological temperature in Apo (Fig. 7 A), ATP (Fig. 7 B), and MgATP (Fig. 7 C) conditions for ABCB10 reconstituted in the different nanodiscs (MSP1D1 in orange, MSP13D1 in blue, and MSP2N2 in red). For comparison, the average distribution for the protein in detergent micelles is also included (black). The first striking difference is visible in the Apo condition, where the peak centered at ∼32 Å, which is minimal in detergent, became more relevant as the diameter of the nanodisc decreases. Therefore, the choice of nanodisc does influence the conformational equilibrium of the nucleotide-free transporter. To our surprise, the response of the protein to ATP is highly dependent on the choice of nanodisc, with the largest ATP-induced dimerization observed for the smallest nanodiscs (Fig. 7 B). The distributions clearly show that the Closed conformation peak (∼47 Å) increases at the expense of the Open conformation peak (∼23 Å), with the largest distance distribution shift observed for MSP1D1 nanodiscs. The relative proportional area for the Closed conformation peak (determined from Gaussian fittings of multiple independent experiments) for each condition are presented next to their distance distribution in Fig. 7 B. The area ranges from no detection in detergent, to 26% for MSP2N2 (red), 37% for MSP1E3D1 (blue), and 62% for MSP1D1 (orange). Notice as well that, in MSP2N2 nanodiscs, the larger distance peak is not centered at the ∼47 Å position, but it is shifted to a ∼42 Å distance, suggesting that, in these larger nanodiscs, ATP without Mg does not fully induce the conformational transition, and the “Closed” proteins might be adopting an intermediate conformation not yet like that of the catalytically active NBD dimer. These results strongly suggest that, whereas the transporter is in the nucleotide-free or ATP states, the smaller-size nanodiscs facilitate the transition toward conformations that are not favorable in detergent or in larger-size nanodiscs. However, during more physiologically relevant conditions (MgATP), our data did not show any significant difference in the distance distributions for the different nanodiscs (Fig. 7 C). The relative area for the Closed conformation peak extends from ∼60% in detergent (black), to 87% for MSP2N2 (red), 84% for MSP1E3D1 (blue), and 83% for MSP1D1 (orange). Notice that, in the larger nanodiscs, MgATP fully shifted the Closed conformation to the ∼47 Å peak, like the one observed in the other conditions. Further studies are needed to determine if the size of the nanodiscs influences the conformational changes in other regions of the transporter during active ATP hydrolysis.

Figure 7.

Effect of nanodisc size on the conformational equilibrium of ABCB10-Q706C at 37°C. (A) Average distance distributions in the nucleotide-free protein. (B) Average distance distributions in the presence of ATP. (C) Average distance distributions in the presence of MgATP. Averages obtained from multiple independent experiments in detergent micelles (black), or protein reconstituted in nanodiscs of different diameters (small MSP1D1, orange; medium MSP1E3D1, blue; large MSP2N2, red). The average areas of the Closed conformation peak calculated from Gaussian fitting of the individual distance distribution of each independent experiment are also indicated (as mean ± standard deviation) in (B) and (C), following the corresponding color code.

Response of ABCB10 to substrate binding

Together with our collaborators, we recently identified BV as the physiological substrate transported by ABCB10 (45). Therefore, here we have investigated the response of the transporter to BV at physiological temperature (Fig. 8). Addition of BV to the nucleotide-free transporter reconstituted in MSP1E3D1 nanodiscs triggered a massive change, with the appearance of a new and distinctive predominant peak centered at ∼38 Å, as shown in the average distance distribution of multiple independent experiments (Fig. 8 A, top). This new ∼38 Å conformation is comparable with the distance estimated from the recently published cryo-EM structure of BV-bound ABCB10 in detergent (PDB: 7Y48), which adopts an IF conformation more compact than the Apo protein. The bottom plot in Fig. 8 A shows the overlay of the distributions in the presence of BV (green) with the distributions in Apo (black), and BV plus ATP (red), Mg (blue), and vanadate (orange) conditions. Addition of BV clearly induces a dramatic change (compare black and green traces) suggesting a large change in conformational equilibrium induced by substrate binding. The subsequent addition of ATP (red), Mg (blue), and vanadate (orange) also induced conformational changes, with a large shift toward the Closed conformation in the MgATP and vanadate conditions. An interesting observation is the widening of the Closed conformation peak in the presence of BV (notice that the Closed conformation peak is wider and extends up to distances of 60 Å). However, the relevance of this observation remains to be elucidated. Importantly, we have detected the effect of BV on nanodiscs of different sizes and in detergent (Fig. 8 B). In all these experimental conditions, the addition of BV promoted the appearance of the peak centered at ∼38 Å (boxed area in Fig. 8 B), although this peak is less prominent in detergent. Future experiments measuring conformational changes in other regions of the transporter are required for a better understanding of the conformational equilibrium of the transporter during the basal ATPase cycle and how the equilibrium could be modified in the presence of substrate.

Figure 8.

Response of ABCB10-Q706C to biliverdin (BV) at 37°C. (A) Average distance distributions of the transporter reconstituted in MSP1E3D1 nanodiscs. The top plot shows the average distribution (circles) and Gaussian fit (yellow line) in the presence of BV. The vertical black arrow points to the new predominant peak observed in the presence of substrate. A cartoon representation of ABCB10 (PDB: 7Y48) with predicted distance for Q706C residues is also shown. The plot below overlays the distance distributions in Apo (nucleotide and BV-free), and in the presence of BV for the rest of the experimental conditions (BV, and BV plus ATP, MgATP, and MgADPVO₄). (B) Average distance distributions in the presence of BV show the new distinctive peak (area enclosed by dashed rectangle) for protein in detergent (black) or reconstituted in nanodiscs of different diameter (orange, blue, and red for MSP1D1, MSP1E3D1, and MSP2N2, respectively). All data are averages from multiple experiments from at least 2 independent protein purifications. Error bars omitted for clarity.

Discussion

Since the 2000s, many structures of ABC transporters have been published, mostly thanks to x-ray crystallography and, more recently, to the amazing advances of cryo-EM. Unfortunately, controversy has surrounded the interpretation and physiological significance of many of these models. These techniques commonly represent conformations of selected molecules stabilized under certain conditions that are often far from what these proteins experience in their native physiological environment. Spectroscopic techniques, such as DEER, have the advantage of providing information about the distribution of conformations within the population of molecules, instead of single selected conformations, but commonly these studies have not shown conformational changes associated with NBD dimerization unless the protein is inhibited by vanadate or mutation of the catalytic glutamate (46,47,48,49). In the past, we have used an alternative spectroscopic approach (LRET) to study the conformational changes of the bacterial ABC transporter MsbA and of murine P-glycoprotein (ABCB1), and we detected changes in the dynamic conformational equilibrium of these transporters under various experimental conditions, including measurements while the transporters are actively undergoing ATP hydrolysis. In this work, we have used this powerful approach to evaluate the conformational changes of ABCB10, the best-known human mitochondrial ABC transporter thus far. ABCB10, similarly to the other two inner mitochondrial membrane transporters ABCB7 and ABCB8, has been studied by x-ray crystallography and cryo-EM, resulting only in structures that adopt IF conformation with separated NBDs, even when nonhydrolyzable ATP analogs are bound to their NBDs (9,10,11). Surprisingly, the lack of NBD dimerization in the available structures of ABCB7 (MgAMP-PNP), ABCB8 (MgAMP-PNP), and ABCB10 (AMP-PNP or MgAMP-PCP) has been interpreted as a requirement for simultaneous binding of both nucleotide and substrate, although such a model is clearly not compatible with the transporters presenting basal ATP hydrolysis activity in the absence of substrate (9,10,11). We hypothesize instead that the recurrent inability to capture these human transporters in the dimeric NBD conformation is due to the experimental conditions used in such studies. Here, we use LRET to determine the conformational changes of ABCB10 undergoing active ATP hydrolysis and to evaluate the effect of nonhydrolyzable ATP analogs, temperature, and membrane mimetic systems commonly used for available structural studies.

In the nucleotide-free state, ABCB10 seems to preferentially adopt conformations with separated NBDs, and our data clearly demonstrate that the temperature has a profound effect on the ability of the transporter to shift toward dimeric NBDs. This effect is especially dramatic for the protein in detergent micelles, where no conformational changes were detected at room temperature in response to nucleotide (either ATP or its nonhydrolyzable analogs AMP-PNP or AMP-PCP). This lack of detectable response at room temperature explains the recurrent inability to trap the dimeric NBD conformation in x-ray crystals and cryo-EM samples studied thus far (9,20). At room temperature, MgATP-induced transition to the dimeric NBD conformation is modest and occurs slowly, which correlates with the low ATPase activity in these conditions. A more robust switch toward the dimeric NBD conformation was achieved after inhibition with vanadate to stabilize this state. Contrarily, at physiological temperature, the presence of MgATP promotes the establishment of a new equilibrium that includes a significant contribution of molecules with dimeric NBDs and at a faster timescale. Reconstitution of ABCB10 in nanodiscs also has a dramatic effect on the behavior of the transporter, with MgATP inducing a partial shift toward dimeric NBDs even at room temperature. Another striking difference is that MgAMP-PNP can displace the equilibrium toward dimeric NBDs in the reconstituted transporter at physiological temperature, whereas no change was observed for the transporter in detergent micelles in otherwise equivalent conditions. AMP-PCP was not effective at inducing NBDs dimerization under any of the tested conditions.

Bacterial and yeast ABCB7 homologs with bound AMP-PNP have been observed in occluded conformations with dimeric NBDs (50,51) whereas the bacterial transporter MsbA and Sav1866 have been crystallized in OF conformation when AMP-PNP is bound (52,53). Therefore, ABC transporters seem to respond differently to ATP analogs and other experimental conditions. Our results on ABCB10 are an example of how the behavior of a human ABC transporter can be drastically affected by the temperature, the presence of a lipid bilayer, and the use of nonphysiological nucleotides. Essentially, our data show that the conditions used for available ABCB10 crystal structures (detergent micelles, low temperature, and either MgAMP-PCP or AMP-PNP without Mg) are not good at promoting NBD dimerization. The ABC transporters field is full of controversies that undermine our mechanistic understanding of these proteins, and our data suggest that the role played by experimental conditions is critical. A systematic study comparing the sensitivity of various ABC transporters (bacterial, yeast, mammalian, for example) to the temperature, detergents, lipid composition, and nucleotides would likely help the interpretation of available data and the experimental design of future experiments using more “native-like” conditions. For example, a cryo-EM study of a Thermus thermophilus heteromeric transporter reconstituted in MSP1D1 nanodiscs and preincubated with MgATP at elevated temperature before freeze plunging has shown a mix of conformations, including structures with dimeric NBDs (54). Thus, high-resolution structural techniques can be combined with more physiologically relevant conditions to improve our understanding of the molecular mechanisms of environmentally sensitive proteins such as ABC transporters.

Since the introduction of nanodisc technology, many laboratories have embraced this approach due to the significant advantages they bring for the study of purified membrane proteins in a lipid bilayer. However, their use has not been exempted from controversies. A couple of recent publications highlight the need to cautiously evaluate the effect that nanodiscs may have on the reconstituted membrane proteins. The first example is the case of MsbA, whose detection in a wide open IF conformation in the native membrane of E. coli cells depleted of ATP has been considered as strong evidence that such widely separated NBDs can be physiologically relevant and actually needed for the entrance of large substrates to the substrate binding cavity in the transporter, and that the more compact conformations that have been reported in nanodiscs are likely due to constraints caused by the small diameter of these discs (14). However, this effect can be species dependent, since MsbA of A. baumannii in MSP1D1 nanodiscs has been detected in a wide open IF conformation (55). The second relevant example is the recent publication demonstrating the effect of the nanodiscs on the structure of pentameric-ligand gated ion channel, either due to alterations of the bilayer properties depending on the size of the discs, or to direct interactions of the channel with the membrane scaffold protein (56). Possible artifacts caused by the nanodiscs need to be seriously considered, especially since most of the available studies thus far have been done using the small MSP1D1 or the slightly larger MSP1E3D1. Therefore, in this study we have also compared the effect of nanodiscs of different sizes on the conformational equilibrium of ABCB10. Our data show that the size of the discs affects the nucleotide-free transporter, with the smaller nanodiscs promoting the appearance of an additional distinctive peak in the distance distribution that can suggest the presence of a more compact conformation. This peak is essentially absent in detergent micelles and becomes smaller as we move to the larger MSP2N2. Even more remarkable is the difference in the response to ATP without Mg, which almost fully shifts the conformational equilibrium toward the dimeric NBDs in the smallest nanodiscs, whereas the protein in the largest MSP2N2 has only minor response. These results are compatible with the idea of the small nanodiscs favoring more compact conformations of the transporter that could alter the response to nucleotide binding. Interestingly, we did not observe obvious differences in the various nanodiscs when ABCB10-Q706C was in the presence of MgATP. Undergoing measurements in other regions of the transporter can help us better understand if the size of the nanodiscs remains influencing the conformational changes of the transporter undergoing active ATP hydrolysis in basal conditions and in the presence of substrate.

The exact reasons behind the effect of a phospholipid’s membrane and/or the size of nanodiscs on the protein’s conformational equilibrium are still unknown. Protein-lipid interactions can regulate the activity of membrane proteins. A native mass spectrometry indicated that ABCB10 binds cardiolipin with higher affinity than that of the bulk phospholipids commonly found in the mitochondrial inner membrane, and that the ATPase activity of the purified transporter is inhibited by cardiolipin in a dose-dependent fashion (57). The E. coli polar lipid extract we have used for reconstitution contains ∼10% cardiolipin (as reported by the vendor, Avanti Research, AL, USA), but alternative control reconstitutions using soy polar lipids (which lack cardiolipin) have shown no effect of the lipid composition on ABCB10’s ATPase activity (39). Because we have expressed ABCB10 in E. coli, it is likely that the purified transporter has bound cardiolipin, which is advantageous in terms of providing a physiologically relevant environment for the transporter. In addition to specific protein-lipid interactions, the physicochemical properties of the phospholipid bilayer can affect the activity of membrane proteins. The thickness and order parameters of the bilayer can be radially heterogeneous in nanodiscs (58). The smaller MSP1D1 nanodiscs seem to exhibit larger stiffness in the center compared with the edges, whereas the spatial changes for MSP1E3D1 and MSP2N2 nanodiscs can be milder (59). Such larger stiffness in the smaller nanodiscs could affect the lateral forces on the reconstituted transporter and partially explain the differences in the conformational dynamic equilibrium we have detected for nanodiscs of different sizes. Such regulations of protein’s conformation and activity might occur in the physiological environment, as macromolecular crowding in the inner mitochondrial membrane can affect the properties of the bilayer and alter the conformation of proteins (60).

Lastly, comparison of our experimental data on ABCB10 (a mammalian homodimer) with our previous studies on the closely related mouse P-glycoprotein (Abcb1; a mammalian single polypeptide) and MsbA (a bacterial homodimer) exposes dissimilar preferences for each of these proteins (29,31). We have used LRET to monitor the NBDs separation for each of these transporters reconstituted in MSP1E3D1 nanodiscs at physiological temperature, and our data suggest a mix of molecules with distances compatible with either dimeric or separated NBDs. The existence of such complex mix of coexisting conformations has been reported before for various ABC transporters using diverse experimental approaches (46,54,61,62,63). Interestingly, whereas our current data in ABCB10 during active ATP hydrolysis shows a marked preference for the dimeric NBD conformation, the equivalent data for MsbA and P-glycoprotein showed a higher contribution of molecules with separated NBDs. Further studies will help determine if these differences could reflect diverse molecular mechanisms for these transporters involving small versus large separation between the NBDs. Another important variable yet to explore is how the composition of the lipid bilayer may affect the conformational changes or the conformational equilibrium of different ABC transporters.

Conclusions

We have investigated the conformational changes of a human ABC transporter using LRET as a very useful spectroscopic approach that allows the analysis of the protein’s conformational equilibrium in real time and at physiological temperature, while the transporter undergoes conformational changes associated with ATP hydrolysis. This study, initially directed at the NBD level, has demonstrated the profound effect that experimental conditions impose on the dynamic equilibrium and the crucial necessity to investigate these transporters under physiologically relevant conditions. Undergoing studies in our laboratory using LRET to evaluate conformational changes in other regions of the transporter will allow us to provide a general mechanistic view during the basal and substrate-activated ATP hydrolysis cycle at physiological temperature.

Data and code availability

All data are included in this article. If any additional details about methodology, data collection, data analysis, or raw data files are desired, request them to Maria E. Zoghbi (mzoghbi@ucmerced.edu).

Acknowledgments

Dr. Luis G. Cuello (Department of Cell Physiology and Molecular Biophysics, Texas Tech University Health Science Center, Lubbock, TX, USA) for suggesting the use of autoinduction media, and for useful scientific discussions and comments about the manuscript. Funding from the National Institutes of Health (NIH), Institute of General Medical Sciences (1R01GM145938 to M.E.Z.) is acknowledged. A.L.H. was partially supported by IBioSTeP: NIH GRISE at UC Merced (T32GM141862). The content is solely the responsibility of the authors and does not necessarily represent the official views of the NIH.

Author contributions

M.E.Z. designed the research, performed the research, analyzed the data, and wrote the manuscript. A.L.H. and A.N performed the research, analyzed the data, and edited the manuscript.

Declaration of interests

The authors declare no competing interests.

Editor: Merritt Maduke.