Formation of a Chloride-Conducting State in the Maltose ATP-Binding Cassette (ABC) Transporter

ABC transporters use an alternating access mechanism to move substrates across cellular membranes. This mode of transport ensures the selective passage of molecules while preserving membrane impermeability. The crystal structures of MalFGK 2 , inward- and outward-facing, show that the transporter is sealed against ions and small molecules. It has yet to be determined whether membrane impermeability is maintained when MalFGK 2 cycles between these two conformations. Through the use of a mutant that resides in intermediate conformations close to the transition state, we demonstrate that not only is chloride conductance occurring, but also to a degree large enough to compromise cell viability. Introduction these to ion-conducting states an ABC transporter into a channel.


Introduction
Most membrane transporters facilitate the movement of substrate molecules using an alternating access mechanism (1)(2)(3)(4)(5). This conserved mode of transport corresponds to a cycle of conformational changes coupled to the movement of gating elements or rigid-body structures on either side of the membrane (5)(6)(7). This movement results in the alternate exposure of the substrate-binding site to the extracellular and intracellular environments (2,6). Intermediate states, which normally occur only transiently during transport, lie between the inward-and outward-facing conformations (8,9).
The alternating access mechanism ensures that the free flow of ions and water molecules is restricted during transport (1,10). Accordingly, since only one gate is open at a given time, the transporter is switching conformations without ever producing a membrane channel (8). Spacefilling models derived from the crystal structures of inward-and outward-facing transporters indicate that the gates are sufficiently tight to prevent the passage of ions and water molecules (2,7,11). However, crystal structures represent static conformations, it is therefore unknown if the dynamics of the gates, especially during conformational transitions, is coordinated enough Ion conductance through MalFGK 2 to prevent the passage of ions and water molecules.
A recent molecular dynamics simulation on various alternating access transporters reported that water passage might be possible through short-lived intermediates states (9). It was proposed that these water-conducting states are inherent characteristic to the alternating access mechanism. Water conductance has also been reported in the family of ion co-transporter (12)(13)(14). In the ABC transporter family, with the exception of CFTR (15), there is however no such evidence. This may be due to the highly transient nature of the intermediate states, which makes detection inherently difficult (9).
In this study, we utilize the MalFGK 2 transporter to explore the gating mechanism. The substrate translocation pathway is comprised of two membrane proteins, MalF and MalG. The nucleotide-binding domain, which controls the conformation of MalFG, is composed of the homodimeric MalK subunit. Transport also requires the periplasmic binding protein MalE to traffic maltose to the transporter and to stimulate its ATPase activity (16)(17)(18). In this study, to facilitate the detection of ion conduction, we employed the mutant transporter MalF500 (19,20). This mutant hydrolyzes large amounts of ATP independent from MalE and maltose. This high ATPase activity is because the mutant rests in intermediate conformations near the transition state (20)(21)(22). Our results demonstrate that MalF500 forms an ion conducting channel which, when overproduced, is deleterious to the cell. In contrast, wild type MalFGK 2 rests in the inwardfacing conformation which is impermeable to ions.

Incorporation of proteins into proteoliposomes
Total E. coli lipids dissolved in chloroform were dried under a stream of nitrogen. The lipids were resuspended in TSG buffer (50 mM Tris-HCl, pH 7.8, 100 mM NaCl, 10% glycerol) containing 0.5% DDM. The lipids were mixed with the purified MalFGK 2 complex at protein: lipid ratio of 1:2000 in TSG buffer plus 0.1% DDM. Detergent was removed using BioBeads (one-half volume) and gentle shaking overnight at 4°C. Proteoliposomes were isolated by centrifugation (100,000 × g, 1h, 4°C) and resuspended in TSG buffer at a final concentration of 3μM MalFGK 2 . For incorporation of ergosterol into proteoliposomes, a mixture of DOPC:DOPG: ergosterol (ratio 60:20:20) was dissolved in chloroform, air dried, and resuspended in S buffer (50mM HEPES pH 7.4, NaCl 150mM, 0.5% DDM) before addition of MalFGK 2 . To incorporate nystatin, proteoliposomes were first Ion conductance through MalFGK 2 frozen in liquid N 2 after which they were thawed on ice in the presence of nystatin (75μg/mL) and then briefly sonicated (15 sec, three times) (26).
Cells were harvested 60 min after induction (3000 x g, 10 min), washed with 1 volume of 5% sucrose buffer (20mM Tris-SO 4 pH 7.8) and resuspended in 1/20 cell culture volume of 18% sucrose buffer. Cells were converted to spheroplasts by addition of 0.1mg/mL lysozyme and 2 mM EDTA for 10 min on ice. Conversion to spheroplasts was considered complete when cell lysis was total and immediate upon dilution into water. To measure membrane permeability, spheroplasts were diluted 20-fold in 500 µl of buffer L (293mM KCl, 20mM Tris-SO 4 pH 7.8) in the presence or absence of valinomycin (5μM). Cell lysis was measured every 5 seconds at 540nm. Tips with a large diameter were employed at all time to prevent pressure-induced lysis during pipetting.

Planar Lipid Bilayer experiments
The electrical currents were recorded on a Planar Lipid Bilayer Workstation (BLM; Warner instruments) composed of a Digidata 1440 Lownoise Data Acquisition System and a BC-535 Bilayer Clamp amplifier. Unless otherwise stated, the recorded data were sampled at 1kHz. The lipid bilayers were painted across a 150 μm hole aperture using a flame smoothed glass applicator stick dipped into a mixture of DOPC:DOPG (ratio 70:30 at 15 mg/mL) in hexadecane. Lipid bilayers were considered ready for protein insertion when their capacitance reached 70-90 pFa. The chambers on the cis-side and trans-side of the bilayer were adjusted to 650 mM KCl and 150 mM KCl, respectively. The two chambers were connected to Ag/AgCl 2 electrodes using an agarose salt bridge (2% agarose in 1M KCl). The electrical current across the lipid bilayer was stable for at least 10 min before addition of proteoliposomes in the cis-chamber. Fusion of the proteoliposomes to the lipid bilayer was facilitated using a stirring magnet. The bilayers containing channel activity were sometimes broken but could be repainted from the cis-chamber. All bilayer measurements were performed at room temperature. Data were recorded and analyzed using the Axoclamp pClamp software suite, version 10.2.

Other methods
The ATPase activity of the maltose transporter was determined by measuring the release of inorganic phosphate using the malachite green photocolorimetric method (27). Protein concentrations were determined using the Bradford assay (28).

MalF500 can readily access the transition state
In the absence of nucleotide, the conformation of the maltose transporter is inward-facing (6,18). The transporter becomes outward-facing upon binding of ATP (29). Its basal ATPase activity is also stimulated by MalE and maltose. In contrast, the ATPase activity of MalF500 (bearing mutations MalF N505I and MalG G338R ) is very high and independent of MalE and maltose (19,20,30). It is proposed that mutations in MalF500 destabilize the inward-facing state, thereby diminishing the energy barrier of the transition state (6). To assess the conformation of MalF500, we performed a crosslinking analysis using the cysteine pairs MalF 442C -MalG 230C and MalF 394C -MalG 182C (18,29). The cysteine pair MalF 442C -MalG 230C reports on the inward-facing conformation of the transporter (Fig. 1a). Without ATP, wild type MalFGK 2 is inward-facing and the crosslink efficiency is maximal (taken as 100%; Fig 1b; upper panel). In the presence of ATP, MalFGK 2 converts to outward-facing and the crosslink efficiency diminishes to ~43%. The same analysis shows that MalF500 does not depend on ATP because its conformation remains essentially unchanged. Also, the maximal crosslink efficiency is 4-fold lower than the wild type (i.e. ~20-25%). Thus, MalF500 is resting in a conformation different than inward-facing. We next employed the cysteine pair MalF 394C -MalG 182C, which reports on the outward-facing state of the transporter (Fig  Ion conductance through MalFGK 2 1a). With MalFGK 2 , the crosslink efficiency is low because the transporter is inward-facing (Fig 1b; lower panel). In the presence of ATP, the crosslink efficiency increases up to ~84% because the transporter becomes outward-facing. In contrast, the crosslink efficiency with MalF500 is maximal and this conformational change is independent from ATP. Together, these data show that i) MalF500 rests in a conformation away from the resting state and ii) MalF500 reaches the outwardfacing conformation independent from ATP. Thus, MalF500 access spontaneously the transition state and hydrolyze larger amounts of ATP in the absence of MalE and maltose.

MalF500 is deleterious to the cell
Overproduction of MalF500 in the membrane results in a significant growth delay (Fig. 2a). This may be caused by the high basal ATPase activity of MalF500. It is also possible that the inherent conformational flexibility of MalF500 may increase membrane permeability and compromise cell viability. In an attempt to differentiate the possibilities, we introduced the mutation E159Q into the Walker A motif of MalK (30). This mutation almost fully abolishes ATP hydrolysis (Fig. 2b), yet the growth of MalF500 E159Q is still impaired compared to the wild type (Fig. 2a). We then introduced a his6-tag at the N-terminus of MalF. The his6-tag decreases the ATPase activity of the wild-type complex through stabilization of the inward facing conformation (18). Addition of the his6-tag on MalF500 reduces its ATPase activity to wild type levels (Fig. 2b), yet the mutant still displays a significant growth defect. These results suggest that the conformation of MalF500, not just its high basal ATPase activity, is detrimental to the cell.

MalF500 is permeable to chloride
We employed a spheroplast lysis assay to determine whether MalF500 affects cell membrane permeability. This method has been developed to study the ion conductance of the SecYEG protein translocation channel (31). Briefly, spheroplasts were diluted into an iso-osmotic solution of KCl in the presence of the K + ionophore, valinomycin. If the membrane is permeable to Cl -, the spheroplasts swell and eventually lyse due to the rapid influx of water. As expected, a low degree of lysis was observed in the absence of valinomycin (Fig. 3a).
In the presence of valinomycin, spheroplasts containing the MalF500 complex lysed immediately (Fig. 3b, red trace). The initial lysis rate is very high, more than 7-fold higher than spheroplasts containing wild type MalFGK 2 complex and near half of spheroplasts containing the SecYEG channel with an altered plug domain (SecEY Δ61-70 G; purple trace). When the experiment is performed with the mutant MalF500 his6 , the rate of lysis is diminished by ~50% (Fig. 3b, orange trace). This is consistent with the observation above that a his6-tag at the N-terminus of MalF only partially restores cell viability.

Ion channel activity of MalF500
The MalF500 complex was purified and incorporated into planar lipid bilayers in order to determine its ion conductance properties (Fig. 4a). The data show that MalF500 has a channel-like activity, and distinct single channel opening and closing events can be resolved. Quantitative analysis of the recordings (Fig. 4b and Fig. 4c) further show that (i) MalF500 creates a membrane channel that is voltage insensitive given the linear voltage-current relationship; (ii) is anion selective, with a reversal potential (-38mV) that is close to the calculated Nernst potential for chloride (-37mV); and (iii) has rapid gating kinetics, with an average dwell time for the opening state of only ~2-20mS. Such quick opening and closing kinetics suggests that the transporter is conformationally flexible. This is in contrast to wild type MalFGK 2 which, throughout the experiment, remained stably closed (Fig. 4a). Although bilayers often contained more than one channel, single channel opening events were easily identified and were used to calculate the ion conductance of MalF500. The histogram of current amplitudes and conductance magnitude were plotted as a current-voltage curve; the slope representing the single channel conductance in pS (Fig. 4c).

The periplasmic gate seals Mal500
The resting maltose transporter (i.e. inwardfacing) is sealed by gating loops on the periplasmic side on the membrane (6). The amino acyl side chain MalF V442 , MalG T228 and MalG V230 Ion conductance through MalFGK 2 form this interface (Fig. 5a). These residues were replaced by the short chain amino acid glycine. According to the space-filling model (Fig. 5a), such mutations produce a wide-open pore in lieu of the gate. Surprisingly, results from the cell growth (Fig. 5b), ATPase activity (Fig. 5c), and cell membrane permeability assays (Fig. 5d) reveal that MalFGK GGG behaves much like the wild type. The glycine residues were then introduced into MalF500 to produce the mutant MalF500 GGG . In this case, we observed an immediate and strong growth defect upon protein production (Fig. 5b) together with a dramatic increase in cell membrane permeability (Fig. 5d). We note that the ATPase activity of MalF500 GGG is ~ 4-fold lesser than MalF500 (Fig. 5c), suggesting that ATP consumption is not the primary reason for the higher growth defect of MalF500 GGG . Apparently, the nature of the gating residues is particularly important when the transporter rests in a conformation near to the transition state.

MalF500 GGG forms a quasi-permanently open membrane channel.
The mutant MalF500 GGG was purified and inserted into planar lipid bilayers (Fig. 6a). In contrast to MalF500, which is equally distributed between open and closed states, MalF500 GGG behaves like an open-state channel (Fig. 6b). The recordings (Fig. 6a) show that the frequency of a gating event is very slow (every ~1-2 sec, compared ~20 msec for to MalF500). This increased open pore duration is consistent with both the dramatic increase in cell membrane permeability (Fig. 5d) and the very strong effect on bacterial growth (Fig. 5b). Single channel recordings could not be captured with the MalF500 GGG mutant; however, isolated closing events could be detected in bilayers containing multiple copies of the mutant (Fig. 6a). The rate of these closing events was such that only one channel out of ~ 25 was observed closing at a time.

Discussion
The two major conformational states of the maltose transporter, inward-facing and outwardfacing, appear sealed against ions and water molecules (6,9,32). Yet, whether the impermeability is maintained when the transporter cycles between these two states is unknown. These intermediate conformations are structurally difficult to characterize and, in the case of MalF500, also difficult to crystallize (19). Recent computer simulations suggest that water molecules can permeate through the substrate passageway when the transporter pass through intermediate conformations; however the pore is very narrow and it is open for only a few nanoseconds (9).
Here, we have augmented access to the intermediate conformations using the mutant MalF500, which carries the mutations MalF G338R and MalF N505I . These mutations destabilize the inward-facing conformation, thereby decreasing the energy barrier for the transition (6,20,22). As a result, the MalF500 mutant is capable of spontaneously adopting the outward-facing conformation independently of ATP (Fig. 1). Through the use of this mutant, we show that increased access to the intermediate conformations allows for a significant degree of ion conduction (Fig. 4). The currents measured are high (>10 7 ions/sec), indicative of a movement of ions through a channel-like structure (12,33). In addition, the frequency of channel closing and opening is very fast: every few milliseconds. This rapid cycling is consistent with the decreased energy barrier between the inward-and outwardfacing conformations, and with the exceptional ability of this mutant to hydrolyze a large amount of ATP. The results also show mutations in MalF500 that stabilize its conformation, and therefore diminish its basal ATP activity, also diminish ion conduction (Fig. 3). Interestingly, the conductance and spheroplast assays indicate that MalF500 is selective for Clover K + , although there is no obvious structural characteristic in MalFGK 2 to explain this selectivity. Notwithstanding, our data show that MalFGK 2 must rest away from the conformations adopted by MalF500 in order to preserve the membrane barrier. This membrane barrier is particularly important in bacteria as ion gradients represent a main energy source (34). It is therefore not surprising that overproduction of MalF500, and especially MalF500 GGG , results in a significant bacterial growth defect (Fig. 5b). This growth defect may reflect the energetic cost associated with the use of counter-acting pumps required to compensate for the leakage of chloride ions.

Ion conductance through MalFGK 2
The periplasmic gate on MalFGK 2 is formed at the interface of fourhelices (6). Surprisingly, we find that alteration of this interface (i.e. introduction of the mutation MalF V442G , MalG T228G , and MalG V230G ) does not alter the function of the gate: the mutant is viable and we do not detect increased ion conductance (Fig. 5d). This result was rather unexpected since a space-filling computer model suggests that the glycine residues leave an open pore over the transport pathway (Fig. 5a). It is therefore possible that an energetic pressure is forcing the mutated gate structure to adopt a conformation that re-creates this essential membrane seal. This phenomenon has been reported in the case of the SecYEG translocon after the deletion of the plug domain (35). In contrast, introduction of the glycine residues into MalF500 leads to a quasi-permanently open pore (Fig. 6). The residues forming the periplasmic gate are therefore critical when the transporter rests in a conformation close to the transition state.
Active transporters and ion channels function by different mechanisms (10), yet our results show that the maltose transporter can acquire a channel-like activity after just a few pertinently located mutations. In the case of CFTR, it is proposed that the stabilization of an intermediate conformation is at the origin of the conversion of this transporter into a chloride channel (36). Specifically, comparisons made with the closely related ABCA4 transporter suggest that the mutation of a salt bridge positioned far away from the gate region might have been crucial for conversion of the ancestral transporter into a channel-like state. After this initial conformational shift, the modern pore and gate regions in CFTR were acquired through additional mutations (36). Consistent with this hypothesis, our results reveal that a simple disruption to the gate region, as in the case of MalFGK GGG , does not suffice to create an ion channel. It is rather the mutations that promote access of the transporter to intermediate conformations, like in MalF500, that are responsible for the ion channel activity.     (a) The electrical currents were recorded across a planar lipid bilayer (70% DOPC, 30% DOPG) at a holding membrane potential of +50mV. Current traces were filtered at 500Hz. The traces for the wild type MalFGK 2 were recorded after >15 fusion events using nystatin/ergosterol as a reporter system indicating protein delivery.

(b)
Histogram of current amplitudes. The number of channel events obtained at +50mV was determined using the Clampfit analysis program and the single channel search function. The currents were plotted as a function of their intensity.
(c) Current-voltage curve for MalF500. Current amplitudes (pA) were plotted according to the applied holding voltage (mV). The slope of the curve represents the channel conductance in pS. The reversal potential is -38mV as indicated by the x-intercept of the curve.