Split tasks of asymmetric nucleotide‐binding sites in the heterodimeric ABC exporter EfrCD

Many heterodimeric ATP‐binding cassette (ABC) exporters evolved asymmetric ATP‐binding sites containing a degenerate site incapable of ATP hydrolysis due to noncanonical substitutions in conserved sequence motifs. Recent studies revealed that nucleotide binding to the degenerate site stabilizes contacts between the nucleotide‐binding domains (NBDs) of the inward‐facing transporter and regulates ATP hydrolysis at the consensus site via allosteric coupling mediated by the D‐loops. However, it is unclear whether nucleotide binding to the degenerate site is strictly required for substrate transport. In this study, we examined the functional consequences of a systematic set of mutations introduced at the degenerate and consensus site of the multidrug efflux pump EfrCD of Enterococcus faecalis. Mutating motifs which differ among the two ATP‐binding sites (Walker B, switch loop, and ABC signature) or which are involved in interdomain communication (D‐loop and Q‐loop) led to asymmetric results in the functional assays and were better tolerated at the degenerate site. This highlights the importance of the degenerate site to allosterically regulate the events at the consensus site. Mutating invariant motifs involved in ATP binding and NBD closure (A‐loop and Walker A) resulted in equally reduced transport activities, regardless at which ATP‐binding site they were introduced. In contrast to previously investigated heterodimeric ABC exporters, mutation of the degenerate site Walker A lysine completely inactivated ATPase activity and substrate transport, indicating that ATP binding to the degenerate site is essential for EfrCD. This study provides novel insights into the split tasks of asymmetric ATP‐binding sites of heterodimeric ABC exporters.


Introduction
ATP-binding cassette (ABC) exporters are ubiquitous transmembrane proteins found in all living cells [1]. They minimally consist of two nucleotide-binding domains (NBDs), responsible for ATP binding and hydrolysis and two transmembrane domains (TMDs), which form the substrate permeation pathway across the membrane. In a typical bacterial ABC exporter, a TMD is fused to a NBD constituting a half-transporter which homo-or heterodimerizes to form the active transporter. Many eukaryotic ABC exporters encode the four domains in a single polypeptide chain and can therefore be seen as fused heterodimers. In order to transport substrates, the TMDs alternate between an inward-and an outward-facing conformation. These movements are fueled by ATP hydrolysis at the NBDs, which are connected to the TMDs via coupling helices. The NBDs dimerize by sandwiching two nucleotides at their interface via Walker A, Walker B, A-loop, and switch loop from one NBD, and D-loop and ABC signature from the opposite NBD and vice versa, thereby creating two composite nucleotide-binding sites (NBSs; Fig. 1). Closed NBDs are associated with TMDs adopting the outward-facing or outward-occluded state. ATP hydrolysis at the closed NBDs destabilizes the NBD dimer, leading to their dissociation and a concomitant transition to inward-facing TMDs.
The conserved motifs fulfill various functions in the transport process ( Fig. 1) [2]. The A-loop carries a conserved aromatic residue (typically a tyrosine), which interacts with the adenine ring of bound ATP via p-stacking. The Walker A motif wraps around the phosphates of bound ATP and contains a highly conserved lysine residue, which interacts with the band c-phosphates of ATP. The interactions mediated by the Walker A lysine contribute to binding affinity and orient ATP such that they are optimally placed for NBD closure. In a closed NBD dimer, the ABC signature motif of the opposite NBD is juxtaposed against the Walker A motif thereby occluding the phosphates of ATP from both sides. ATP sandwiching at the NBD dimer interface is a prerequisite for ATP hydrolysis, which is mediated by a catalytic dyad consisting of the Walker B glutamate and the switch loop histidine. The D-loops play an important role in the allosteric coupling of the ATP-binding sites. The Q-loop senses the presence of the nucleotide and is noncovalently connected to the coupling helices of the opposite polypeptide chain thereby mediating cross-communication between TMDs and NBDs. Many ABC exporters, such as PatAB [3], LmrCD [4], TM287/288 [5], TmrAB [6], CFTR [7], and TAP1/2 [8], feature one NBS called the degenerate site, in which the Walker B glutamate, the switch loop histidine, and the ABC signature motif deviate from the consensus sequence. The second NBS agrees with the consensus sequence of ABC-type NBDs and is therefore denoted as the consensus site. The degenerate site binds nucleotides tightly, but is unable to hydrolyze ATP [9,10]. This notion was supported by the structure of inward-facing TM287/288, which featured adenosine 5 0 -(b,c-imido)triphosphate (AMP-PNP) bound exclusively to the degenerate ATP-binding site [5]. Further structural and biochemical analyses of TM287/ 288 revealed that nucleotide binding stabilizes the cross-NBD contacts of the inward-facing transporter in an allosteric fashion involving the D-loops [11,12], thereby preventing full separation of the NBDs as seen in inward-facing ABC exporters featuring two consensus sites [13][14][15]. Mutations introduced at the degenerate site are generally better tolerated than analogous mutations at the consensus site [16,17]. This was also observed for Walker A lysine mutants of TAP1/2, CFTR, and MRP1 [8,10,18,19], suggesting that nucleotide binding to the degenerate site is not a strict requirement for NBD closure and substrate transport.
Recently, the heterodimeric ABC exporter EfrCD of Enterococcus faecalis was identified as an important drug efflux pump in this opportunistic human pathogen [20]. EfrCD contains a complete set of noncanonical substitutions in the degenerate site, namely an aspartate instead of a glutamate in the Walker B motif, a glutamine instead of a histidine in the switch loop and a threonine instead of a glycine in the middle position of the ABC signature motif (Fig. 2). In this study, we introduced 15 point mutations into the NBDs of EfrCD and studied their functional consequences with regard to transport and ATPase activity. Our results provide evidence for a regulatory role of the degenerate site in modulating the activity of the consensus site as well as for the functional necessity of ATP binding to the degenerate site.

Results
Point mutations were introduced into the coding sequence of the multidrug ABC exporter EfrCD ( Fig. 1) [20]. Taking advantage of the robust drug transport activity and high basal ATPase activity of the purified transporter, the mutants were studied in three different assays: (a) Transport of the fluorescent dyes ethidium, Hoechst 33342, and 2 0 ,7 0 -Bis(2-carboxyethyl)-5(6)-carboxyfluorescein acetoxymethyl ester (BCECF-AM) in Lactococcus lactis cells, (b) Hoechst 33342 transport in inside-out membrane vesicles (ISOVs), and (c) activity measurements with detergentpurified protein to determine apparent K m and v max of ATP hydrolysis and co-operativity between the two ATP-binding sites.

Walker B motif
The Walker B glutamate of the consensus site was exchanged for glutamine (E512Q EfrD ). The resulting mutant was incapable of transporting fluorescent dyes in cells and inactive in terms of ATPase activity of the purified protein (Fig. 3, Table 1). This finding is in full agreement with the literature on heterodimeric ABC exporters [5,8,16,17,[20][21][22]. In all experiments conducted in this study, wild-type and E512Q EfrD mutant were included as controls and represented the boundaries (fully active vs. inactive) in the functional assays.
When the equivalent noncanonical aspartate of the degenerate site was mutated to asparagine (D492N EfrC ), the functional consequences were less severe. The capability of transporting Hoechst 33342 and BCECF-AM in cells was only marginally reduced, while the mutant had a reduced capability of ethidium transport. Hoechst 33342 transport in ISOVs was clearly impaired, indicating that this assay is more sensitive to mutations than Hoechst 33342 transport in cells. The purified D492N EfrC mutant exhibited a 5fold decreased v max and a 2.6-fold increased K m of basal ATPase activity as compared to the wild-type transporter ( Table 1). The same aspartate residue was also substituted by glutamate (D492E EfrC ). The ATPase activity and apparent ATP affinity of the D492E EfrC mutant was almost identical to the D492N EfrC mutant, indicating that introduction of a  The D492E EfrC mutant therefore appears to run in an 'energy-saver' mode, exhibiting wild-type transport activity at 5-fold reduced ATPase consumption. The functional consequences of the Walker B mutations in the two NBSs are highly asymmetric; while they are comparatively well tolerated at the degenerate site, the equivalent mutation at the consensus site leads to a complete inactivation of the transporter. Nevertheless, a correctly positioned negative charge at the position of the noncanonical Walker B aspartate is of functional importance. Our results as well as previous studies [16] suggest that the noncanonical Walker B aspartate is not involved in ATP hydrolysis at the degenerate site. Therefore, the decreased ATPase activities observed for these mutants must be due to reduced ATP hydrolysis at the consensus site. The Walker B motif precedes the D-loop, which allosterically couples the two ATPbinding sites [11]. This coupling is likely compromised by mutations introduced at the degenerate site Walker B motif.

Switch loop
The switch loop histidine interacts with the Walker B glutamate and is part of the catalytic dyad at the consensus site. In EfrCD, the degenerate site contains a glutamine instead of the consensus histidine. In agreement with earlier studies on HlyB [23], the H543A EfrD substitution at the consensus site led to a complete inactivation of transport and ATPase activity ( Fig. 4, Table 1). The equivalent Q523A EfrC substitution at the degenerate site was comparatively well tolerated. The mutant was almost indistinguishable from the wild-type transporter in terms of ATPase activity, but had a clearly reduced transport activity, i.e., it is less efficient in coupling ATP hydrolysis with substrate transport. In the inward-facing TM287/288 structure (PDB: 4Q4A), this glutamine plays an important role in mediating cross-NBD hydrogen bonds, which are eliminated by the substitution to alanine. A weakened NBD-NBD interface of inwardfacing EfrCD as a consequence of the Q523A EfrC mutation may therefore explain the observed coupling defect.

ABC signature motif
The ABC signature motif plays a key role in the dimerization of the NBDs as it binds to the phosphate groups of ATP in the closed NBD dimer. The degenerate site ABC signature motif of EfrCD has the sequence FSTGQ, whereas the equivalent sequence of the consensus site is FSGGQ. The T489G EfrD substitution at the degenerate site resulted in a mutant carrying two canonical ABC signature motifs. This mutant was as active as the wild-type transporter in cellular transport assays using the dyes Hoechst 33342 and BCECF-AM (Fig. 5, Table 1). Surprisingly, the mutant exhibited an even stronger ethidium efflux phenotype than the wild-type transporter, while the v max was decreased by a factor of 1.8. The G469T EfrC substitution at the consensus site resulted in an EfrCD mutant carrying two noncanonical ABC signature motifs. The mutant showed a mildly impaired transport activity for all three dyes tested, and had a 5.6-fold reduced v max . Interestingly, the apparent affinity for ATP increased 2.4-fold, which is presumably caused by interactions of the hydroxyl group of the introduced threonine with the phosphates of ATP. In rat TAP1/2, the corresponding amino acid is a valine and it has been suggested that-in contrast to EfrCD-the noncanonical deviation causes a decrease of affinity, which permits the reopening of the degenerate site without the necessity to hydrolyze the bound ATP [18]. With regard to substrate transport, our data are in full agreement with a study on human TAP1/2, which showed increased peptide transport for a chimera carrying two consensus site ABC signature motifs and slightly reduced transport for a chimera containing two degenerate site ABC signature motifs [24]. The mutations in EfrCD lead to mildly asymmetric functional results. In general, alterations at the middle position of the ABC signature motif are comparatively well tolerated.

D-loop
The D-loop owes its name to a highly conserved aspartate at the end of the consensus sequence SALD. In TM287/288, the D-loops are structurally asymmetric. The degenerate site D-loop strongly interacts with the opposite NBD and thereby prevents complete NBD separation of the inward-facing transporter. In contrast, the consensus site D-loop is highly flexible and only contacts the Walker A motif of the opposite NBD when a nucleotide is bound to the degenerate site [11]. Introduction of the D-loop mutations into EfrCD had asymmetric functional consequences. The consensus site D-loop mutant D498A EfrC was completely inactive in terms of ATP hydrolysis and dye transport in cells and ISOVs (Fig. 6, Table 1). The degenerate site D-loop mutant D518A EfrD has a 36fold reduced v max of ATP hydrolysis and its apparent K m is 1.69 mM, which is much increased compared to the wild-type transporter. Nevertheless, the D518A EfrD mutant still exhibited clearly measurable transport activity in cells and ISOVs. In summary, the degenerate site D-loop aspartate is important for the proper function of EfrCD, while the consensus site D-loop aspartate is essential. The asymmetric result of these equivalent mutations can be explained by structural differences between the D-loops as observed in TM287/288.

Q-loop
The Q-loop contains an eponymous highly conserved glutamine residue. In high-resolution full-length ABC exporter structures, the oxygen atom of the glutamine side chain interacts with the catalytic magnesium ion and is in hydrogen bonding distance of the c-phosphate of ATP [5,[25][26][27]. The other residues of the Qloop are at the base of a groove in the NBDs, which serve as docking site accommodating the coupling helices of the TMDs [28]. A recent study on ABCB1 has demonstrated that the Q-loop plays a role in coupling drug binding and transport events at the TMDs with the catalytic cycle at the NBDs [29].  and its v max was reduced by a factor of 3.4 as compared to wild-type EfrCD (Fig. 7, Table 1). The phenotype was more pronounced when the equivalent mutation Q431A EfrD was introduced into the consensus site, which led to a 15-fold reduced v max value and a clearly reduced capacity to transport all three dyes. Both Q-loops proved to be of functional relevance in EfrCD. The consensus site Q-loop is in direct contact with the catalytic magnesium, which explains the observed loss of ATPase and transport activity. Interestingly, the equivalent mutation in the degenerate site as well resulted in a clear drop of activity. It is unlikely that the degenerate site Q-loop directly contributes to ATP hydrolysis. Rather, it is involved in an allosteric coupling to the consensus site. To investigate whether the Q-loops of EfrCD are responsible for cross-communication between the TMDs and NBDs as suggested for ABCB1 [29], we measured ATPase activity of the mutants in the presence of increasing concentrations of daunorubicin or Hoechst 33342 as performed previously [20]. However, the pattern of drug-modulated ATPase activity could not be distinguished from the wild-type transporter (not shown).

A-loop
Both A-loops of EfrCD contain a tyrosine residue which contributes to nucleotide binding by interacting with the adenine ring of ATP. To study its functional role, the tyrosines were replaced by alanines, thereby removing the aromatic ring needed for p-stacking.  Their v max values for ATP hydrolysis were around 100-fold decreased with respect to the wild-type transporter (Table 1). In addition, the K m values are 1.95 mM and 1.91 mM for Y338A EfrC and Y359A EfrD , respectively, indicating a much decreased apparent ATPase affinity as a consequence of these mutations. Interestingly, the mutation in the degenerate site (Y338A EfrC ) abrogated the positive co-operativity of ATP binding (Hill coefficient of 1.01 instead of 1.78 for wild-type EfrCD). Given the low ATPase activities of the mutants, the cellular transport activity remained surprisingly high for both A-loop mutants (Fig. 8), i.e., they were similar to the degenerate site Walker B mutant D492N EfrC . In contrast, Hoechst 33342 transport in ISOVs was heavily impaired, in particular for the degenerate site mutant Y338A EfrC . The A-loop mutants clearly show that nucleotide binding to both ATP-binding sites is of critical importance for EfrCD and that the mutations impair the function in a similar fashion regardless whether they were introduced at the degenerate or the consensus site.

Walker A motif
The Walker A motif contains a highly conserved lysine residue, which plays a role in ATP binding, NBD closure, and ATP hydrolysis by interacting with the band c-phosphates of ATP. In EfrCD, both NBSs contain the conserved Walker A lysine and its positive charge was removed by substituting with methionine. The resulting single mutants K369M EfrC and K389M EfrD were catalytically inactive and did not exhibit any measurable transport activity in cells and ISOVs (Fig. 9, Table 1). This finding reinforced the observation made for the Aloop mutants that regardless whether the mutation was  introduced at the degenerate or the consensus site, the functional outcome was symmetric. From these results, it was concluded that nucleotide binding at the degenerate site is a strict requirement for NBD closure and transport activity of EfrCD.

Discussion
Many heterodimeric ABC exporters contain asymmetric ATP-binding sites in which one site carries a characteristic set of noncanonical substitutions. In Type I ABC exporters sharing a TMD architecture as described for TM287/288 [5], including the prominent human examples TAP1/2, SUR1, and CFTR (Fig. 2)  of the central glycine by another amino acid), indicating that they play a dedicated function. Of note, there are asymmetric Type II ABC exporters with an inverse domain topology and different TMD architecture, such as Pdr5 [30], Cdr1 [31], and ABCG5/ABCG8 [32]. They contain a degenerate ATP-binding site as well, but the pattern and functional role of their noncanonical substitutions is different from classical Type I ABC exporters. In EfrCD and its closely related homologs, the deviations at the degenerate site lead to an inactivated catalytic dyad. This implies that the task of the consensus site is the one of a workhorse, which is responsible for ATP hydrolysis, while the functional role of the degenerate site is the one of a coachman. Thus, the degenerate site does not hydrolyze ATP, but rather senses nucleotide binding and regulates the events at the consensus site. Structurefunction studies and DEER measurements on TM287/ 288 revealed two roles of nucleotide binding at the degenerate site [11,12]: (a) it prevents full NBD dimer separation by establishing additional cross-NBD hydrogen bonds in an allosteric fashion and (b) communicates across the NBD to the consensus site via the flexible consensus site D-loop to promote ATP hydrolysis.
Here, we studied key residues in the NBDs of EfrCD with unmatched completeness. Of note, all EfrCD mutants yielded similar amounts of protein and eluted as monodisperse peaks from size exclusion chromatography (not shown). Remarkably, all point mutants had a decreased ATPase activity compared to the wild-type transporter, and with the exception of T489G EfrD and D492E EfrC , they were partially or fully defective in transport. Hence, the sequence motifs of both the degenerate and the consensus site form an intricate network, which operates in concert and requires all players for efficient ATP hydrolysis and substrate transport. To the best of our knowledge, Aloop and Q-loop mutants have never been investigated in other Type I ABC exporters containing a noncanonical NBS.
Mutations introduced into Walker B and switch loop, which vary between the two NBSs are comparatively well tolerated when introduced at the degenerate site, but inactivated the transporter when placed at the consensus site (Table 2). This finding is in accord with previous reports investigating the Walker B motifs of LmrCD [16], CFTR [19], TAP1/2 [8], and TmrAB [17], and the switch loop of TAP1/2 [8,18]. We also created EfrCD with two degenerate or two consensus ABC signature motifs, leading to a slight reduction or slight activation of transport activity, respectively. This finding was again in agreement with reports on TAP1/2 [18,24]. Substitutions of the A-loop tyrosine and the Walker A lysine, which are invariant in both NBSs, resulted in identical functional outcomes regardless where they were introduced. This finding stands in contrast to previous studies on TAP1/2 [33], MRP1 [10], and CFTR [19], in which Walker A mutants introduced at the degenerate site resulted only in a partial inactivation, while the equivalent mutations at the consensus site fully inactivated the transporters. At the consensus site, the Walker A lysine is involved in nucleotide binding and hydrolysis, which explains why the residue is crucial for function. At the degenerate site, however, the Walker A lysine plays only a role in nucleotide binding and NBD closure. Substitution of the Walker A lysine leads to a strong affinity decrease [7,10]. The residual nucleotide affinity at the degenerate site of MRP1 and CFTR may be sufficiently high for partial nucleotide binding, which then results in a partial transport activity, while in EfrCD, the affinity drop may be so severe that nucleotide binding vanishes completely. As an alternative explanation, Walker A   mutations may lead to a complete loss of nucleotide binding at the degenerate site in all transporters, but only leads to a full inactivation in EfrCD, while MRP1 and TAP1/2 may still be partially active with exclusive nucleotide binding to the consensus site. In any case, nucleotide binding to the degenerate site is important for proper functioning. This notion is further supported by the A-loop mutants, for which we observed a highly symmetric functional outcome. The mutations in the D-loops of EfrCD had a very strong impact on in vitro and in vivo activity, underscoring the functional importance of this motif. In TM287/288, by contrary, the ATPase activity of the consensus site D-loop mutant could still be determined, while it was found to be even increased for the equivalent mutation at the degenerate site [11]. The EfrCD data also differ from a D-loop study on TAP1/ 2, in which the degenerate site mutant remained fully active in terms of transport and the consensus site mutant turned the active transporter into a nucleotidedependent passive facilitator [34]. Of note, EfrCD mutated at the consensus site D-loop did not facilitate the influx of dyes into the L. lactis cell. Facilitated dye influx was previously described for uncoupled mutants of the major facilitator family drug transporter LmrP in the same organism [35].
By comparing basal ATPase activity and transport activity, we observed differences in coupling efficiencies. The A-loop mutants, degenerate site Walker B mutants and the consensus site ABC signature mutant, exhibited strongly reduced ATPase activities, but retained comparatively high transport activities. Hence, they seem to couple ATP hydrolysis more efficiently with substrate transport than the wild-type protein and appear to run in an 'energy-saver' mode. In contrast, the switch loop mutant Q523A EfrC retained a high ATPase activity while being severely impaired in transport, indicating a coupling defect. Interestingly, the apparent ATP affinity of most 'energy-saver' mutants was found to be decreased and their transport rates likely become strongly attenuated if the ATP concentration drops in the cytoplasm. Unfortunately, we were unable to measure EfrCD-mediated ATPase activities directly in cell-derived ISOVs, which would have permitted the determination of the ATPase activity in the native membrane. Therefore, conclusions regarding the coupling need to be taken with a grain of salt.
Overall, we observed asymmetric results for mutations introduced at motifs that differ among the two NBSs. Mutations placed at the noncanonical NBS were consistently better tolerated. Asymmetric functional results were as well obtained for mutations of the D-loop and the Q-loop, which play a role in interdomain communication. While the D-loop exhibits distinctive structural differences among the two NBSs, it remains to be shown whether this is also the case for the Q-loop, for example, in the still missing outward-facing structure of a heterodimeric ABC exporter. Finally, when mutating motifs which are invariant among the two NBSs, such as Walker A and A-loop, the observed loss of function was identical regardless where the mutations were introduced.
This in-depth study provides novel insights into the functional mechanism of the large family of heterodimeric ABC exporters and will pave the way for further studies on transporters of high medical relevance involved in multidrug resistance and hereditary diseases.
Cloning, expression, and purification of transporters in and from L. lactis The enterococcal ORF efrCD was amplified from the genomic DNA of E. faecalis V583 and cloned into the pREXNH3 shuttle vector as described previously [20,36]. The mutants were generated by QuikChange site-directed mutagenesis (primer sequences are given in Table 3 and were obtained from Microsynth, Balgach, Switzerland). Wild-type and mutant efrCD were then subcloned into the L. lactis expression vector pNZ8048 via vector-backbone exchange (VBEx) cloning [37]. EfrCD was expressed in L. lactis and purified using b-DDM as described previously [20].

Transport assay of fluorescent dyes in cells
Lactococcus lactis NZ9000 DlmrADlmrCD cells harboring plasmids encoding wild-type or mutant EfrCD were grown in M17 containing 5 lgÁmL À1 chloramphenicol and 0.5% glucose at 30°C. Expression was induced at an OD 600 of 0.4-0.6 with a nisin containing culture supernatant of L. lactis NZ9700 for 1 h energized by adding 0.5% glucose. Nigericin and valinomycin (1 lM each) were added prior to the addition of 2 0 ,7 0 -bis-(2-carboxyethyl)-5(6)-carboxyfluorescein acetoxymethyl ester (BCECF-AM) to avoid changes of BCECF fluorescence as a result of pH changes in the cytoplasm. The accumulation of 5 lM ethidium, 0.5 lM Hoechst 33342, or 0.2 lM BCECF-AM was followed at 25°C for 600 s using a Fluorescence Spectrometer LS-55 (Perkin Elmer, Schwerzenbach, Switzerland). Excitation and emission wavelengths and slit widths were set at 520 nm, 10 nm and 595 nm, 15 nm for ethidium, 355 nm, 5 nm and 457 nm, 5 nm for Hoechst 33342, and 502 nm, 2.5 nm and 525 nm, 4.0 nm for BCECF-AM, respectively. Transport assays for each mutant were conducted four times with two independent cell batches. For clarity of discussion, one representative dataset is shown.

Hoechst 33342 accumulation in inside-out membrane vesicles (ISOVs)
Inside-out membrane vesicles were prepared using a Microfluidizer (Microfluidics) at 30 000 psi as described previously [20] and collected by ultracentrifugation. The protein concentration in ISOVs was determined by the Micro bicinchoninic acid (BCA) Protein Assay Reagent Kit (Pierce, Reinach, Switzerland) and BSA as standard. ISOVs containing 400 lg protein were added to 2 mL of 50 mM KP i pH 7.0 and 10 mM MgSO 4 . About 0.5 lM Hoechst 33342 was added 10 s after starting the measurement. Transport was initiated by adding 5 mM ATP after 180 s. Hoechst 33342 fluorescence was monitored for 600 s at 25°C. Excitation and emission wavelengths (slit widths) were set at 355 nm (5 nm) and 457 nm (5 nm), respectively. Transport assays were conducted four times with two independent ISOV batches and one representative dataset is shown.

ATPase activity assay
ATPase activity was determined using detergent-purified protein in 50 mM K-HEPES pH 7.0, 10 mM MgSO 4 , and 0.03% b-DDM containing increasing concentrations of ATP (0.1-8 mM) for 15 min at 30°C. Hydrolyzed inorganic phosphate was detected colorimetrically using a malachite green-molybdate solution as described previously [11]. The data were fitted using the Hill equation (SIGMAPLOT 11.0; Scientific Solutions, Pully, Switzerland): In which v max corresponds to the maximum rate of ATP hydrolysis, K m to the apparent ATP affinity, and n to the Hill coefficient.