Probing the Orientation of Reconstituted Maltoporin Channels at the Single-protein Level*

Recently we have shown that maltoporin channels reconstituted into black lipid membranes have pronounced asymmetric properties in both ion conduction and sugar binding. This asymmetry revealed also that maltoporin insertion is directional. However, the orientation in the lipid bilayer remained an open question. To elucidate the orientation, we performed point mutations at each side of the channel and analyzed the ion current fluctuation caused by an asymmetric maltohexaose addition. In a second series we used a chemi-cally modified maltohexaose sugar molecule with inhibited entry possibility from the periplasmic side. In contrast to the natural outer cell wall of bacteria, we found that the maltoporin inserts in artificial lipid bilayer in such a way that the long extracellular loops are exposed to the same side of the membrane than protein addition. Based on this orientation, the directional properties of sugar binding were correlated to physiological conditions. We found that nature has optimized maltoporin channels by lowering the activation barriers at each extremity of the pore to trap sugar molecules from the external medium and eject them most effi-ciently to the periplasmic side. The outer membrane of Gram-negative bacteria inner the

Recently we have shown that maltoporin channels reconstituted into black lipid membranes have pronounced asymmetric properties in both ion conduction and sugar binding. This asymmetry revealed also that maltoporin insertion is directional. However, the orientation in the lipid bilayer remained an open question. To elucidate the orientation, we performed point mutations at each side of the channel and analyzed the ion current fluctuation caused by an asymmetric maltohexaose addition. In a second series we used a chemically modified maltohexaose sugar molecule with inhibited entry possibility from the periplasmic side. In contrast to the natural outer cell wall of bacteria, we found that the maltoporin inserts in artificial lipid bilayer in such a way that the long extracellular loops are exposed to the same side of the membrane than protein addition. Based on this orientation, the directional properties of sugar binding were correlated to physiological conditions. We found that nature has optimized maltoporin channels by lowering the activation barriers at each extremity of the pore to trap sugar molecules from the external medium and eject them most efficiently to the periplasmic side.
The outer membrane of Gram-negative bacteria consists of an asymmetrical bilayer with phospholipids in the inner leaflet and lipopolysaccharides in the outer leaflet. The lipopolysaccharide headgroups are cross-linked by divalent cations, thereby providing an impermeable network for hydrophilic solutes and protecting the cell from damaging agents such as bile salts, lipases, and proteases. Uptake of nutrients or secretion of proteins through this barrier is accomplished by several channel forming proteins called porins (1). With a lack of other carbohydrate sources, the entry of maltodextrins has been optimized by the bacteria through the specific channel maltoporin, also called LamB (2)(3)(4).
The high resolution x-ray structure of maltoporin has been solved to 3.1 Å (5). The protein forms a homotrimer presenting three water-filled channels. A monomer consists of an 18stranded antiparallel ␤-barrel embedded in the membrane, with large extracellular loops and short turns at the periplasmic side. The third loop, L3, folds inside the pore, contributing to a considerable constriction at the middle of the channel formed by the residue Tyr 118 . Interestingly, a line of six aromatic residues composed by Trp 74* of an adjacent subunit, Tyr 41 , Tyr 6 , Trp 420 , Trp 358 , and Phe 227 , called the "greasy slide," extends from the entrance vestibule to the periplasmic outlet (see Fig. 1A) and follows the left-handed helical shape of the longer maltodextrins, conferring a screw-like character to the translocation process. Crystal structures of maltodextrins bound to maltoporin reveal that the sugar molecule is oriented with its nonreducing end pointing toward the periplasmic exit of the pore (6,7). The complex also reveals three glucosylbinding subsites where the apolar pyranose rings are in van der Waals' interaction with the residues Tyr 41 , Tyr 6 , and Trp 420 at the central part of the greasy slide, whereas the moieties at either end of the sugar curl away from the residues at the periphery (Trp 74* at the vestibule and Trp 358 and Phe 227 at the periplasmic end). Simultaneously, sugar-hydroxyl groups are engaged in hydrogen bonds with two "polar tracks." All of these interactions are supposed to provide a specific sugar translocation pathway. Specific binding of maltodextrins increases the local concentration inside the channel and thus facilitates translocation at low substrates concentrations.
Maltoporin has been extensively studied using the black lipid membrane technique. The maltoporin proteins are incorporated into planar lipid bilayers and characterized by their conductance (8 -11). Based on the observation that sugar addition reduces channel conductance in a concentration-dependent manner, conductance measurements can be used as a probe to reveal the thermodynamic parameter of the sugar binding. Information about the rate constants of sugar translocation can be obtained using the spectral analysis of the ion current fluctuations (10 -13). In previous studies conducted on a large ensemble of porins, site-directed mutagenesis was employed to probe the functional role of the greasy slide (14 -16). The mutations affecting the aromatic residues at the channel center considerably reduces the stability constants as suggested by the crystallographic data. The effect of the mutation is less pronounced for residues located at either end of the slide (W74*A, W358A and F227A) but demonstrates that they are also involved in the facilitated substrate transport through maltoporin channels.
Recent conductance measurements performed on single maltoporin trimers show that the temporary binding of a maltodextrin molecule inside the channel can be observed as ion current fluctuation as penetrating sugar molecules constrain the passage of ions (17)(18)(19). Reconstitution experiments at the single-channel level allowed unambiguous exploration of the directional properties of maltoporin channels. These studies clearly showed that the channel is highly asymmetric with respect to the sugar entrance side and applied voltages and revealed that the maltoporin insertion in artificial membranes is always unidirectional (17)(18)(19). However, the real orientation of maltoporin remained an open question.
In multichannel reconstitution experiments, the direction of maltoporin incorporation has been a matter of debate. Based on the binding asymmetry of a pseudooligosaccharide modified at its nonreducing end (20), on mutants deleted in the large extracellular loops, and on asymmetrical pH-induced closure of the channel proteins, Benz and co-workers (21,22) propose that maltoporin preferentially inserts with the short periplasmic turns moving through the membrane. In contrast, Van Gelder et al. (23) used bacteriophageand observed the opposite orientation. The skeleton of maltoporin is in a ribbon representation shown in violet. The six aromatic residues of the greasy slide and Tyr 118 at the central constriction are highlighted using stick representation. The two outermost residues Trp 74 and Phe 227 , which have been mutated in alanine in this study, are colored orange. The bound maltohexaose molecule is represented as a Cory-Pauling-Koltun model. Green spheres are carbon atoms, and red spheres are oxygen atoms. The figure was carried out using InsightII, version 1998 software (Molecular Simulation Inc., San Diego, CA). We used the structure of the complex obtained by Dutzler et al. (6). B, structures of the natural maltohexaose (M6) and the maltohexaose analogue (M6-ANDS) modified at the reducing end with ANDS. The abbreviation M6-ANDS stands for a covalent link between a functional maltohexaose moiety (M6) and the ANDS substituent.
In the present study, we address the question of maltoporin orientation at the single channel level. First we investigate the influence of the mutations W74A at the extracellular vestibule and F227A at the periplasmic exit on the kinetics of maltohexaose binding. A second method to probe the orientation is based on the obvious asymmetric binding process of maltodextrin to maltoporin requiring an orientated penetration. We synthesized an ANDS 1 -maltoheptaose derivative (noted M6-ANDS in the following), a maltohexaose analogue with a bulky ANDS cap at its reducing end (see Fig. 1B). This modification prevents sugar molecules from penetrating the channel with the reducing end. The knowledge of maltoporin orientation associated with the functional role of the outermost residues of the greasy slide allowed us for the first time to provide a quantitative description of carbohydrates transport involving an asymmetric energy profile for a sugar molecule permeating the channel.

MATERIALS AND METHODS
Preparation and Characterization of M6-ANDS-Maltoheptaose was purchased from Senn Chemicals (Basel, Switzerland), and ANDS was from Interchim (Paris, France). 60 mg of maltoheptaose (around 50 mol) were labeled as a result of reductive amination by the adding of 250 l of 0.2 M ANDS in 15% acetic acid and the same volume of NaBH 3 CN 1 M in tetrahydrofuran. The reaction was incubated for 3 h at 55°C.
The choice of ANDS-maltodextrin cap to probe maltoporin orientation was motivated by the fact that synthesis and characterization of ANDS-oligosaccharide derivatives have already been described (24). Maltoheptaose ANDS-derivative, called here M6-ANDS (see Fig. 1B), was first analyzed by capillary electrophoresis, and the major peak of M6-ANDS was identified. Characterization of this peak was performed by capillary electrophoresis coupled with electrospray ionization mass spectrometry under the conditions described below. The negative mass spectrum obtained was dominated by peaks at m/z 1438 and 718.9 assigned to single and doubly charged deprotonated molecular ions (M Ϫ H) Ϫ and (M Ϫ 2H) 2Ϫ of M6-ANDS.
Capillary electrophoresis-mass spectroscopy analyses were carried out on a CE system P/ACE TM MDQ (Beckman Coulter, Inc.) with a 75-m ϫ 80-cm fused silica capillary. The outlet of the capillary was integrated into the ESI spray needle that was directly coupled to an ion trap mass spectrometry system (LCQ TM DUO; Thermofinnigan, Inc.). The separations were monitored with a Beckman laser-induced fluorescence detection system using a 4 mW argon ion laser with an excitation wavelength of 488 nm and an emission filter of 520 nm. During analysis, the temperature was maintained constant (25°C) along the capillary, and a voltage of 4 kV was applied at the outlet end of this capillary. The sheath liquid (water/isopropanol, 20/80, v/v) at the rate of 5 l/min and the sheath gas (nitrogen, 20 units) were infused coaxially to the CE capillary. For measurements, the negative mode was used, and all of the data were collected on X Calibur software (see Ref. 25 for details).
Planar Lipid Bilayer Experiments-Planar lipid bilayers have been prepared of diphytanoyl phosphatidylcholine (Avanti Polar Lipids Inc.) according to the technique of Montal and Mueller (26). They are formed across a 60-m-diameter hole in a 25-m-thick Teflon film (Goodfellow, Cambridge, UK) being sandwiched between two Delrine chambers, each containing 2 ml of an aqueous solution (1 M KCl, 1 mM CaCl 2 , 10 mM Tris buffered to pH 7.4). The whole set-up was shielded from external electromagnetic fields as well as from vibrations to minimize extra noise from other sources. The Delrine cell was enclosed in a double isolated Faraday cage connected with the signal ground and also with a homemade acoustically isolating closet placed on a piezo-electric vibration isolating table (model "Elite 3"; Newport Corp., Irvine, CA). The quality of the bilayer membranes was checked by capacitance and residual conductance measurements. The capacitance of the whole system proved to be about 150 picofarads. The residual membrane conductance (less than 7 picosiemens) was subtracted from the overall conductance.
The apparatus had been connected with the external circuit through a pair of homemade Ag/AgCl electrodes encased in 200-l pipette tips filled with 5% agarose soaked with 1 M KCl during the fabrication process. The electrode on the so called cis-side of the measuring cell was grounded, whereas the other (on trans) was connected with the head stage of an Axopatch 200B amplifier (Axon Instruments, Foster City, CA) in the voltage clamp mode.
The mutagenesis and purification of maltoporin have previously been described in detail (15). Small amounts of wild type maltoporin from a 0.1 g/ml buffer solution with 1% octyl-POE detergent (Alexis, Lauchringen, Switzerland) were injected into the cis-side compartment. Incorporation of maltoporin into the bilayer was promoted by applying a transmembrane voltage of 100 -200 mV to tentatively disturb the membrane and by stirring for a few seconds after addition. We have optimized the protocol to observe single porin insertion in less than 10 min and to inhibit further insertion. A single maltoporin molecule in the bilayer could be kept stable for several hours without any significant change of its physical properties. The concentration of maltohexaose (Senn Chemicals) was adjusted by adding appropriate small amounts of a concentrated stock solution. The exact sugar concentration was then determined by means of optical polarization measurements (Perkin-Elmer 241). After sugar addition the aqueous solution was homogenized by stirring during a few seconds. Then signals were recorded 20 min later. Titrations for the whole range of applied voltages were carried out with the same maltoporin molecule to avoid possible divergences between different individual single protein molecules. All of the measurements were performed at room temperature. The applied transmembrane voltage refers to the potential on the cis-side relative to the trans-side.
The data were filtered with the low pass Bessel filter of the amplifier at 2-5 kHz and then monitored with a Lecroy (Geneva, Switzerland) LT342 digital storage oscilloscope. The entire experiment was recorded on video tape with a digital type recorder (DTR-1204; Biologic, Claix, France). The average power spectrum of the current noise was recorded using the fast Fourier transform module of the oscilloscope. To overcome the additional noise sources mentioned above, the background spectrum of the membrane without sugar was subtracted from each individual spectrum. The curve fitting was carried out using the Marquart-Levenberg method. The total current recording was transferred to a personal computer via a GPIB card using the graphical program LabVIEW 4.01 from National Instruments. Statistical analysis of the blockade events was performed by means of the BioPatch Analysis (Science Instruments) and homemade software.
Derivation of the Individual Rate Constants of Sugar Binding-The binding of carbohydrates to the maltoporin channels was assumed for a long time to be symmetric (9,12), so that once a sugar molecule is bound inside the pore it would have equal probability to translocate or to exit to the same side. Only recent measurements performed on individual maltoporin trimers demonstrated that the association rate is 3-5-fold different depending on the sugar entrance side (18,19). The reaction for maltodextrin (M) binding has been proposed to be of first order. The internal binding site, P, is accessible from either side of the membrane (cis as well as trans). A bound substrate molecule, PM, closes the where k on cis and k on trans refer to the association rate constants for a sugar molecule entering the channel from the cis-and the trans-sides, respectively; and k off cis and k off trans stand for the dissociation rate constants of a sugar molecule exiting from the binding site, respectively, to the cisand the trans-sides. The two respective pairs of on-off rate constants associated with the cis-and trans-sides are related to the single thermodynamic equilibrium constant, K, according to the detailed balance principle.
The on rate constants k on cis and k on trans can individually be found out from the respective simple cases where sugar is only added to the cis-side or the trans-side of the channel. We cannot directly access the off rate constants k off cis and k off trans from the current fluctuation measurements because they are not distinguishable. Nevertheless, they can easily be derived from the apparent k off ϭ k off cis ϩ k off trans and Equation 1. The relaxation time constant of the binding process, , that modulates the ion current depends on the rate constants and sugar concentration, [M], as follows.
The parameter f c can be determined by fitting the power spectral density of the sugar-induced ion fluctuations with the Lorentzian form (12), where S 0 is the plateau power density at frequencies f Ͻ Ͻ f c , and f c is the corner frequency at S(f) ϭ S 0 /2.

RESULTS
Effect of the Mutations W74A and F227A on Ion Conduction-As seen in Fig. 1A, the residue Trp 74 is located at the entrance, and Phe 227 is located at the periplasmic exit, and both were changed into alanine. To study the influence of these mutations on the directional insertion of maltoporin channels, we first recorded the ion current through single mutant W74A at different applied voltages and calculated the corresponding conductance values. In a second series of measurements, we repeated the recording with the F227A mutant. The results of both recordings are shown in Fig. 2. We observe that the conductance of the wild type and mutated channels increases with the applied voltage and depends on the sign of the external potential. Repeated recordings on porin insertion revealed in all the cases (more than 200 attempts for the wild type FIG. 3. Sugar-induced current transitions through wild type and modified channels. Current recordings of single wild type (A) and mutated maltoporin trimers W74A (B) and F227A (C) in presence of 60 M of maltohexaose to the cisor trans-side of the membrane. Maltohexaose binds to maltoporin from either side of the membrane and closes channels for 1.1, 0.7, and 1.7 ms on average for wild type, W74A, and F227A proteins, respectively. It is seen that sugar addition to the cis-side causes fewer blockages than sugar addition to the trans-side. This asymmetry is more pronounced for sugar binding to the W74A mutant. In comparison with W74A, the longer sugar residence time inside F227A channels leads to higher probability to find two or three monomers closed at the same time. The dashed lines correspond to zero current. The applied voltage was ϩ150 mV. maltoporin and more than 50 attempts for each mutated proteins) an orientation having the high conductance under smaller potentials on the side of channel addition, i.e. at negative voltages according to our sign convention. At Ϯ190 mV, the asymmetry is 14% for both mutant channels. It is noteworthy that larger conductance at negative voltages has been obtained for all insertions independently on the sign of the potential applied during channel incorporation. However, it depends on the side of protein addition. Conductance of the mutated channels is slightly higher than the wild type maltoporin. Apparently the mutagenesis did not modify the asymmetry in ion conduction, suggesting that the mutations W74A and F227A do not affect the unidirectional insertion of the maltoporin channels into artificial membranes.
Effect of the Mutations W74A and F227A on Maltodextrin Binding-After showing the directional incorporation through the asymmetry in conductance, we titrated maltohexaose molecules on one side only to probe the orientation of reconstituted maltoporin channels. Typical current recordings obtained from single wild type, W74A, and F227A maltoporin channels in the presence of one-side addition of maltohexaose are shown in A, B, and C of Fig. 3, respectively. We observed transient blockade events whatever the side of sugar addition was. Kinetic analysis shows that the power spectral density of the fluctuations in the current through individual maltoporin mutants have a Lorentzian form (Fig. 4). Similar behavior has already been reported for the wild type maltoporin channels (17). The rate constants governing the open-close sequences experienced by a given channel can then be determined using Equation 2 by means of a linear plot of the reciprocal relaxation times versus maltohexaose concentrations. The on and off rate constants are reported in Table I. The first observation is that the apparent dissociation rate constant does not depend on the side of sugar addition. The mutation W74A leads to an increase of the off rate comparing with the wild type channels, whereas a significant decrease is found for the modification F227A. This result is directly related to the longer average sugar residence time in F227A channels as observed in Fig. 3. At ϩ150 mV, maltohexaose stays 1.1, 0.7, and 1.7 ms on average inside wild type, W74A, and F227A maltoporin channels, respectively. The second observation is that the mutation W74A introduces a pronounced on rate asymmetry with the side of sugar addition, whereas the asymmetry tends to be reduced in F227A. k on cis is smaller than k on trans by factors of 9.4 and 2.3 for W74A and F227A channels, respectively. A comparison of the on rate constants with wild type maltoporin channels reveals a higher contribution of Trp 74 than Phe 227 . The replacement F227A displays a small reduction of the on rate at the trans-side, whereas a small increase is observed at the cis-side. Mutant W74A shows a major reduction of the on rate constant at the cis-side compared with wild type by a factor of 3.9 at ϩ150 mV.
The kinetics of maltohexaose binding through wild type maltoporin channels has been shown to be sensitive to the external voltage with higher influence on the apparent dissociation rate of bound substrates (17)(18)(19). This binding property has been also used to detect the directional insertion of wild type maltoporin. We studied whether the mutations W74A and F227A influence this voltage dependence by determining the individual off rate constants, k off cis and k off trans , at Ϯ75 and Ϯ150 mV from one-sided sugar addition experiments (see "Materials and Methods"). The results are presented in Fig. 5. Elementary dissociation rate constants of maltohexaose binding to wild type and mutated channels exhibit the same voltage dependence. Off rates are always higher at negative voltages. This voltage-induced asymmetry is more pronounced at high applied potentials and predominantly concerns k off trans . About 50 -60% difference in k off trans is seen at Ϫ150 mV versus ϩ150 mV for all three channels studied.
Asymmetrical Addition of M6-ANDS-In a second series of measurements we probe the orientation using an asymmetric substrate. Covalent linkage of a bulky ANDS substituent at the reducing end of maltooligosaccharides strongly increases the inherent asymmetry of substrates. The mechanism of M6-ANDS transport through maltoporin channels was investigated at the single protein level by reconstituting one wild type maltoporin trimer into lipid bilayers. Further evidence of the asymmetry of the channel is observed after one-sided sugar addition (18,19). Typical recordings of maltohexaose-mediated current fluctuations are shown in Fig. 3A. It is interesting to note that the natural substrate closes the channels independently of which side it enters. In clear contrast to maltohexaose, Fig. 6A shows that one-sided addition of M6-ANDS generates conductance interruptions by multiples of one-third of the initial value only for substrate molecules entering from the cisside, i.e. the side of maltoporin addition. On the other hand, no modification occurs in the current through maltoporin channels when M6-ANDS is added to the trans-side of the membrane. When maltoporin is injected to the trans-side, open channel conductance asymmetry is inverted, and the side of M6-ANDS addition that generates blockade events becomes FIG. 4. Spectral analysis of maltohexaose-induced fluctuations in the current. The power spectral densities of the current noise through W74A channels (A) and F227A channels (B) show a single Lorentzian form, implying that sugar binding to W74A and F227A channels is still a simple two-state model with no memory effect. They also indicate that W74A channel is highly asymmetric with the side of sugar addition. The applied voltage was ϩ150 mV, and the maltohexaose concentration was 60 M.
the trans-side (data not shown). Importantly, the side of blockade events is independent of the sign of the external voltage, demonstrating that the introduction of two negative charges on the substrate molecule is not at the origin of the directional binding. Interestingly, the averaged residence time of a bound maltohexaose molecule, r , is 1.1 ms, whereas M6-ANDS closes channels for about 5.0 ms.
We verified that the presence of M6-ANDS to either side of the membrane does not modify channel permeability for natural substrates. Three consecutive experimental conditions were conducted on the same inserted maltoporin: 1) trans-addition of M6-ANDS leads to unaffected ions flowing through maltoporin channels (Fig. 6A, left panel), 2) trans-addition of maltohexaose generates current interruptions characteristic of sugar binding (Fig. 6B, left panel), demonstrating that only substrates with a unmodified reducing end can penetrate the channels from trans, and 3) addition of M6-ANDS to the cis-compartment leads to the superimposition of long blockade events (Fig. 6B, right panel) originating from sugar analogue binding (Fig. 6A,  right panel). Maltohexaose and M6-ANDS presenting different characteristic time of binding, power spectral density of the current in a mixture of both substrates exhibit low frequency and high frequency components induced by M6-ANDS and maltohexaose, respectively, resulting in the sum of two separated Lorentzian curves (Fig. 7). DISCUSSION Maltoporin Channels Insert with the Periplasmic Turns-We used two strategies to probe the orientation of maltoporin channels reconstituted into planar lipid bilayers. First, we have tested the asymmetry of maltoporin by site-directed mutagenesis at the exit and entrance of the channel (Fig. 1A). Second, we prepared a strongly asymmetrical maltodextrin analogue modified at its reducing end with an ANDS group (Fig. 1B).
We have chosen a mutation on the two outermost aromatic residues of the greasy slide, Trp 74 at the extracellular vestibule and Phe 227 at the periplasmic outlet (Fig. 1A), involved in maltodextrin translocation (14,15). One would expect that the on rate is more affected for sugar molecules entering the protein channels from the side of the modification. This assumption is supported by the recent crystal structure of W74A that revealed only minor structural changes induced by the mutation (15). The mutation W74A shows a strong asymmetrical influence on the on rate because reductions by factors of 3.9 and 1.4 are observed at the cis-and trans-sides, respectively, at ϩ150 mV ( Fig. 3B and Table I). The maltoporin orientation with the extracellular exposed surface at the cis-side provides a good explanation for the major contribution of Trp 74 to the association rate for sugars penetrating the channel from the cis-compartment. The minor on rate decrease associated with sugar entry from the periplasmic vestibule may derive from small mutation-induced rearrangements inside the pore that in turn modify the subtle position of the affinity determinants. It is noteworthy that the increase of channel conductance generated by the mutations W74A and F227A (Fig. 2) may come from larger vestibule diameters induced by the replacement of the bulky aromatic residues that permits higher ions flow.
In a previous study (19), we attributed voltage effects to changes in the overall dipole moment of the protein in the course of field-induced structural transitions that, in turn, alter the binding affinity. The neutral amino acids Trp 74 and Phe 227 are expected not to contribute significantly to the dipole moment of the maltoporin channel. The unaffected features of Maltohexaose was added either to the cis-or to the trans-side of the membrane. The apparent on and off rate constants were obtained from sugar-induced fluctuations analysis at ϩ150 mV. Each value is the mean of three independent titration experiments. The standard deviation is Ͻ20% of the mean value. The apparent binding constant, K, is given by k on /k off ⅐k on refers to the elementary association rate constant k on cis or k on trans when sugar is added to the cis-or to the trans-side of the membrane, respectively. The apparent dissociation rate constant, k off , is the sum of the individual k off cis and k off trans values.

Side of sugar addition
Wild type W74A F227A voltage-induced asymmetry in sugar transport through the W74A and F227A mutant channels compared with the wild type (Fig. 5) are in agreement with our previous model.
We studied the binding properties of the maltodextrin analogue M6-ANDS to confirm the orientation of maltoporin channels proposed by the asymmetric influence of the mutations W74A and F227A on sugar translocation. The drops in the conductance to multiples of one-third of the initial value can clearly be assigned to reversible and complete occlusions of the individual maltoporin channels among the trimeric protein by M6-ANDS molecules (Fig. 6A). In contrast to natural maltodextrins that reach the central binding site from both sides of the membrane (Fig. 3A), the highly asymmetric maltohexaose derivative, M6-ANDS, binds to the pore only when added to the same compartment as maltoporin.
The energy-minimized structure of M6-ANDS illustrated in Fig. 8 shows that the six glucopyranosyl rings keep a lefthanded helical shape close to the natural substrate maltohexaose conformation. This finding shows that M6-ANDS molecules entering maltoporin channels with the nonreducing end first can be recognized as natural maltodextrins despite the modification at the reducing extremity. This is in agreement with the crystallization of maltoporin ligated with nitrophenylmaltotrioside (7). The helical arrangement defined by the aromatic lateral chains lining the pore allows the translocation of the left-handed sugar helix in a screw-like manner (6,7). The orientation of sugar molecules entering into the pore is dictated by its helical conformation and the asymmetry of the molecules (28). Only sugars with the good configuration are translocated. To be in agreement with the orientation of bound sugars obtained by crystal structures, maltodextrins have to enter with the nonreducing end first from the extracellular side, whereas they have to present the reducing end when coming from the periplasmic side (Fig. 1). Penetration of sugars (natural sub-strates as well as M6-ANDS analogue) presenting the nonreducing end from the periplasmic extremity of the pore, thus resulting in bound molecules oriented in the opposite direction  Fig. 6B (left). The spectrum has a single Lorentzian form (solid curve through the data) characteristic of maltodextrins binding (17). The corner frequency is 290 Hz, and the plateau value is 1.8 ϫ 10 Ϫ25 A 2 /Hz. B, M6-ANDS to the cis-side leads to an additional low frequency Lorentzian form (dashed curve through the data) reflecting M6-ANDS noise. The corner frequency is 65 Hz, and the plateau value is 1.9 ϫ 10 Ϫ24 A 2 /Hz. The total spectral density is described by two Lorentzian forms separated in frequency corresponding to different binding kinetics of the two substrates in the aqueous solution. The applied voltage was Ϫ150 mV. than observed in the crystallized complex, is improbable because it would strongly modify the network of hydrogen bonds between the sugar hydroxyl groups and the residues of the polar tracks. Therefore, one would expect that M6-ANDS binds to the channel only when coming from the side of the membrane corresponding to the extracellular side of the bacterial outer membrane. Stepwise transitions of the current through one maltoporin trimer resulting from reversible bindings of M6-ANDS entering the channels with the nonreducing end first appear after it has been added to the same side of the membrane than maltoporin proteins (Fig. 6A). In agreement with the orientation of maltoporin channels suggested by the kinetics of maltohexaose transport through the mutants W74A and F227A, we conclude that maltoporin inserts with the periplasmic turns leaving the extracellular loops oriented at the cis-side of the membrane (Fig. 8), confirming the proposal of Benz and co-workers (21,22). It is noteworthy that Van Gelder et al. (23) found an opposite orientation of maltoporin in lipid bilayers. They used phagebinding, which binds only to the extracellular loops, avoiding sugar penetration from this side. The origin for this discrepancy is not clear. Unlike the results presented here, the latter were done on multichannels recording, which does not provide a clear yes-no answer. The conclusions were drawn on the inhibition of sugar binding after phageadsorption. One reason may be that LamB-phage interaction leads to structural rearrangement of the pore that may modify sugar binding properties differently when approaching the channel from different sides. Moreover, some contaminents not eliminated during the purification protocol of bacteriophages may bind to nonspecific sites on the porin. However, these are speculations, and we currently investigate in more detail the blocking events on a single molecular level.
The net translocation of M6-ANDS through the channel is clearly inhibited for sterical reasons generated by the bulky ANDS substituent. Structural accommodations associated to the passage of the ANDS cap through the pore would be energetically too expensive. We propose that all the time-resolved transitions in the ion current induced by M6-ANDS originate from penetrating M6-ANDS molecules that are forced to exit to the same side. Examination of the kinetics of M6-ANDS binding also provides an elegant way to probe the elementary dissociation rate constant of maltohexaose at the cis-side; the activation energy to exit to the trans-side is infinite, which simplifies, k off apparent ϭ k off cis ϭ r Ϫ1 ϭ ϳ200 s Ϫ1 . This is in excellent After soaking the M6-ANDS in a 5 Å layer of water molecules, molecular dynamics at 300 K and energy minimization using the conjugate gradient were carried out to optimize the conformation of M6-ANDS. All of the calculations were performed using InsightII software with Discover module, and the consistent valence forcefield was used. Orientation of maltoporin channels with the extracellular loops facing the cis-side of the membrane is consistent with the asymmetrical binding of M6-ANDS (Fig. 6A). The sugar analogue binds to the pore with the nonreducing end first only when coming from the extracellular medium, which is the cis-side in our experimental procedures.
FIG. 9. Schematic drawing of the sugar transport energy profile along the channel axis. We consider that sugar (S) transport via the maltoporin-binding site (P) is described by a one-site, two-barrier model. The access to the potential well (PS) from either the cis-or trans-side of the channel is determined by an energy barrier (called E a cis at the cis-side and E a trans at the trans-side) resulting from the formation of an activation state (or transition state) mentioned [P⅐⅐⅐⅐S cis ] and [P⅐⅐⅐⅐S trans ], respectively. The binding free energy, ⌬G 0 , can be calculated from the equilibrium constant, K, with ⌬G 0 ϭ ϪRT⅐lnK. The height of the activation barriers have been approximated as explained in the text. We used the rate constants determined at ϩ150 mV for wild type, W74A, and F227A channels ( Table I). The association rate constants are higher for sugar molecules coming from the trans-side indicating that E a trans Ͻ E a cis . We choose the situation where sugar is present only to the cis-side of the membrane. The ground state at the cis surrounding medium (P ϩ S cis ) is taken at 0. The thermodynamically lower state of the system corresponds to the ground state at the trans-side (P ϩ S trans ) leading to a directional flow of sugar from cis to trans. agreement with the value of k off cis for maltohexaose binding to wild type channels at ϩ150 mV reported in Fig. 5. In comparison with M6-ANDS, the shorter residence time for maltohexaose is due to another issue to escape with a lower energy barrier.
Physiological Implication of Trp 74 and Phe 227 -Now that we have solid evidence of maltoporin channel orientation in our reconstitution protocol, it is interesting to link the directional properties of maltohexaose binding with the asymmetric topology of the maltoporin channels. Even if the aromatic residues Trp 74 and Phe 227 have symmetrical position into the pore, our single maltoporin study revealed that they play totally different roles in the kinetics of sugar binding. The strong decrease of the on rate introduced at the extracellular side by the mutation W74A demonstrates that this residue helps the association step probably by properly positioning maltodextrins before they diffuse deeper into the channel. At the other end of the slide, Phe 227 facilitates the sugar release to the periplasm because its mutation in alanine slows down the dissociation rate.
According to the one-site two-barrier model associated with sugar transport through wild type, W74A, and F227A mutant channels, the corresponding free enthalpy energy diagram is illustrated in Fig. 9. The kinetic association constants depend on the free enthalpy of activation, E a , according to Eyring equation: k on ϭ k max ⅐exp(ϪE a /RT), where R is the gas constant, T is the temperature, and k max is a temperature-dependent factor. In classical transition state theory, k max ϭ 6 ϫ 10 12 s Ϫ1 at 20°C. In the present system, where the sugar transport involves motions in solution, this value is inappropriate. The pre-exponential factor, which reflects the diffusion-limited rate of encounter of sugar molecules and channel extremities in aqueous buffer, has been approximated for sugar binding to periplasmic proteins as 10 9 M Ϫ1 ⅐s Ϫ1 (29). We chose this value as an estimate for the pre-exponential factor of the association rate constants k on cis and k on trans that permits evaluation of the corresponding E a cis and E a trans reported in Fig. 9. Importantly, the relative onset activation energies between the cis-and trans-sides or between mutated and wild type channels can directly be calculated from the individual rate constants for sugar binding without any estimation of k max . Trp 74 participates in the reduction of the extracellular activation barrier, E a cis , because the mutation W74A induces an increase of 3.3 kJ/mol at ϩ150 mV. The activation barrier that has to overcome a bound substrate molecule to exit the pore by the periplasmic outlet equals ⌬G 0 ϩ E a trans . In the wild type channel, Phe 227 helps the sugar release to the periplasmic side by decreasing this activation barrier by 1.3 kJ/mol compared with the mutant F227A.
The net flux of substrate molecules, , across asymmetrical one-site two-barrier model under one-sided sugar addition at concentration [M] is given by the following equation,  Fig. 10 shows the flux of maltohexaose through a single wild type, W74A, and F227A channel at ϩ150 mV for translocation from cis to trans (thick lines) according to the free enthalpy diagram of Fig. 9 and from trans to cis (thin lines). Two main points have to be noted. First, the number of translocated substrate molecules through the three types of studied channels is higher in the cis to trans direction. The effect is more pronounced at concentrations higher than 100 M. Interestingly, although the activation barrier is higher on the cisside (resulting in a smaller number of binding events for sugar added to the cis-side; Fig. 3), the channel permeability is more efficient from the cis-side, because a substrate molecule will have a higher probability to translocate (Ͼ0.5) once it reached the binding site than to exit to the same side. The opposite is valid for sugars coming from the trans-compartment. This finding demonstrates that the orientation of maltoporin channels into the bacterial outer membrane is important to enhance sugar permeability from the extracellular medium (the cis-side in our protocol) to the periplasmic side. Second, the wild type channel is already optimized for long linear maltodextrins translocation because the mutations W74A and F227A reduce the maltohexaose flux.
Structural considerations showed that the periodic energy profiles of the hydrogen bonds and hydrophobic contacts are shown to be shifted resulting in a "smooth" total energy profile (30). This permeation type involves significant energy steps only on initial association and final dissociation of the channel. Furthermore, no significant rearrangements leading to energy reorganization had to occur along the translocation pathway of a sugar molecule (30,31). This suggests that maltoporin can reduce the onset activation barrier by having its affinity site prearranged with an hydrophobic environment and a charge FIG. 10. Flux of sugar through one maltoporin channel. The translocation rates (molecules/channel/s) from cis to trans (thick lines) and from trans to cis (thin lines) versus maltohexaose concentration for wild type (solid lines), F227A (dashed lines), and W74A (dotted lines) channels have been calculated with Equation 4 using the pertinent parameters as they have been evaluated from the experimental results at ϩ150 mV (Table I). The highest permeability is seen for wild type channels for directional substrates translocation from cis to trans, i.e. from the extracellular side to the periplasm. distribution complementary to the permeant sugar. This mechanism of transition state stabilization by specific solvation substitution constitutes the basis of enzymatic catalysis process in a polar solvent (32). In this sense, maltoporin channels catalyze the flux of maltodextrins the same way that enzymes catalyze chemical reactions. Protein channels and enzymes seem to employ the identical molecular mechanism of transition state stabilization to ensure their functions. In the particular case of maltoporin channels, we showed here that residues Trp 74 and Phe 227 are crucial for directional transport by reducing the two encountered activation barriers.
Concluding Remarks-To our great surprise maltoporin channels insert in the opposite direction than in vivo. This observation remains to be investigated in further detail. The in vivo mechanism of porins folding and insertion is not well established, but the implication of several patterns like lipopolysaccharides, periplasmic chaperones, and catalysts has been proposed (33)(34)(35)(36)(37). Black lipid membrane experiments represent a highly simplified model system where folded porins spontaneously insert into symmetric and planar lipid bilayers. In a future experiment the effect of lipopolysaccharides should be investigated.
On the other hand, the orientation of maltoporin with the external loops facing to the side of protein addition corresponds nicely to the orientation with an optimal turnover number of sugar molecules. Based on the presented data, we may easily quantify the energy barrier and turnover number of maltooligosaccharides across maltoporin.
We use the asymmetric voltage effects as a quick tool to demonstrate the directional insertion of maltoporin in an artificial membrane. At 100 mV applied voltage, the signal allows a clear distinction between the two orientations. It is interesting to note that the voltage across the outer membrane is much lower (38). Therefore, variations of the transmembrane voltage probably have a minor influence on the uptake of substrates in vivo (39).