Cu(II) Inhibition of the Proton Translocation Machinery of the Influenza A Virus M2 Protein*

The homotetrameric M2 integral membrane protein of influenza virus forms a proton-selective ion channel. An essential histidine residue (His-37) in the M2 transmembrane domain is believed to play an important role in the conduction mechanism of this channel. Also, this residue is believed to form hydrogen-bonded interactions with the ammonium group of the anti-viral compound, amantadine. A molecular model of this channel suggests that the imidazole side chains of His-37 from symmetry-related monomers of the homotetrameric pore converge to form a coordination site for transition metals. Thus, membrane currents of oocytes of Xenopus laevis expressing the M2protein were recorded when the solution bathing the oocytes contained various transition metals. Membrane currents were strongly and reversibly inhibited by Cu2+ with biphasic reaction kinetics. The biphasic inhibition curves may be explained by a two-site model involving a fast-binding peripheral site with low specificity for divalent metal ions, as well as a high affinity site (K diss ∼2 μm) that lies deep within the pore and shows rather slow-binding kinetics (k on = 18.6 ± 0.9m −1 s−1). The pH dependence of the interaction with the high affinity Cu2+-binding site parallels the pH dependence of inhibition by amantadine, which has previously been ascribed to protonation of His-37. The voltage dependence of the inhibition at the high affinity site indicates that the binding site lies within the transmembrane region of the pore. Furthermore, the inhibition by Cu2+ could be prevented by prior application of the reversible blocker of M2 channel activity, BL-1743, providing further support for the location of the site within the pore region of M2. Finally, substitutions of His-37 by alanine or glycine eliminated the high affinity site and resulted in membrane currents that were only partially inhibited at millimolar concentrations of Cu2+. Binding of Cu2+ to the high affinity site resulted in an approximately equal inhibition of both inward and outward currents. The wild-type protein showed very high specificity for Cu2+ and was only partially inhibited by 1 mm Ni2+, Pt2+, and Zn2+. These data are discussed in terms of the functional role of His-37 in the mechanism of proton translocation through the channel.

The M 2 protein of influenza A virus is thought to function as an ion channel that permits protons to enter virus particles during uncoating of virions in endosomes. In addition, in influenza virus-infected cells the M 2 protein causes the equilibration of pH between the acidic lumen of the trans-Golgi network and the cytoplasm (reviewed in Refs. 1 and 2). The M 2 protein contains a 24-residue N-terminal extracellular domain, a single internal hydrophobic domain of 19 residues which acts as a transmembrane domain and forms the pore of the channel, and a 54-residue cytoplasmic tail (3). Chemical cross-linking studies showed the M 2 protein to be minimally a homotetramer (4 -6), and statistical analysis of the ion channel activity of mixed oligomers indicated that the minimal active oligomer is a homotetramer (7). Despite the small size of the active M 2 oligomer, several pieces of evidence indicate that the ion channel activity is intrinsic to the M 2 protein. First, ion channel activity has also been observed in two expression systems in addition to oocytes, mammalian cells (8) and yeast (9). Second, the activity has also been reconstituted in artificial lipid bilayers (10) from purified M 2 protein. The currents associated with the M 2 ion channel are inhibited by amantadine, its methyl derivative rimantadine, and the spirene-containing compound BL-1743 (Structure 1). Mutant viruses resistant to amantadine or BL-1743 have been isolated, and most have been found to map to the extracellular half of the transmembrane domain of the M 2 protein (11)(12)(13). These mutant M 2 proteins have ion channel activity that is insensitive to the compound used to generate the resistant mutant virus (8, 12, 14 -16). Inhibition by BL-1743 also has the useful property that, unlike inhibition by amantadine, it is reversible on the time scale of physiological experiments (15).
The ion selectivity of the M 2 ion channel has been examined with ion substitution and reversal voltage measurements. The channel possesses at least 10 5 -fold selectivity for H ϩ (8,17). This finding has been confirmed with intracellular pH measurements, but the former experiments indicate that other monovalent cations such as Na ϩ also permeate (17).
Recently, we proposed a model for the three-dimensional structure of the transmembrane region of M 2 , based on a mathematical analysis of the functional properties of a series of mutants (18). The predicted structure consists of four ␣-helices arranged with approximate 4-fold symmetry about a central channel, which spans the transmembrane region of the protein. This model is in very good agreement with spectroscopic data (19 -21) and with an independently proposed structure based on molecular dynamics calculations (22). Our model provides a rationale for the proton selectivity of the channel; the tetramer defines a continuous, water-filled pore, which is interrupted at only one position, His-37 (Fig. 1A) which may act as a proton shuttle similar to that of carbonic anhydrase (23,24). A con-* This research was supported by Public Health Service Research Grants AI-20201 (to R. A. L.) and AI-31882 (to L. H. P.) from the National Institutes of Health. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact. siderable body of evidence shows that His-37 is indeed essential for the activity of M 2 , and mutation of this residue leads to a channel that conducts ions in a pH-independent manner (25).
Amantadine inhibits M 2 with a dissociation constant of approximately 0.2 M and a relatively slow rate of association (16). This drug binds to the channel with a half-time on the order of 10 -20 min when the drug is present at low micromolar concentrations (the second order rate constant ranges from 150 to 1000 M Ϫ1 s Ϫ1 for M 2 from different amantadine-sensitive subtypes of virus). Interestingly, another positively charged, hydrophobic drug, BL-1743, is known to bind M 2 at a similar rate, although its rate of dissociation varies widely; the halftime for dissociation of BL-1743 from M 2 is approximately 3 min (15), while amantadine dissociates at a rate too slow to be experimentally measured. The relatively slow rates at which these drugs associate with the channel may relate to the mechanism by which they enter the pore. Both drugs are relatively bulky and may experience some steric hindrance as they enter the channel. In addition, the partial or full dehydration of positively charged groups as the drugs enter the channel may be energetically difficult, leading to a slow rate of penetration. We suggest that these drugs interact with the lining of the pore, which is composed primarily of hydrophobic groups that form a widening near the center of the bilayer similar to that found in the K ϩ channel (26).
To test further the role of His-37 in the function and inhibition of M 2 , we have examined the ability of Cu(II) to inhibit the channel. Examination of the model (Fig. 1) indicated that this ion could interact with the four histidine imidazole groups similarly to the role proposed for histidine as the chelating ligand for Cu(II) binding to prion protein (27). We demonstrate that Cu(II) indeed binds to a high affinity site within the protein in a slow, time-dependent process.

EXPERIMENTAL PROCEDURES
Site-specific mutagenesis of M 2 cDNA was performed as described previously (15,19). In vitro synthesis of mRNA was performed using the mMessage mMachine T7 Transcription Kit (Ambion, Austin, TX).
Microinjection and Culture of Oocytes-Ovarian lobules from individually identified Xenopus laevis females (Nasco, Fort Atkinson, WI) were surgically removed and treated with collagenase B (2 mg/ml; Boehringer Mannheim) in Ca 2ϩ -free OR-2 solution (82.5 mM NaCl, 2 mM KCl, 1 mM MgCl 2 , 5 mM HEPES-NaOH, pH 7.5) at 24°C for 30 -45 min to liberate oocytes from follicle cells. Defolliculated oocytes were washed in OR-2 and maintained in ND-96 solution (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl 2 , 1 mM MgCl 2 , 2.5 mM sodium pyruvate, 5 mM FIG. 1. Location of histidine residues and possible mode of interaction of the M 2 transmembrane region with amantadine. A, the previously described model of the transmembrane region of the M 2 protein (18). The fourth helix of the tetramer has been removed from the front of the structure to reveal the pore interior. A large cavity can be seen at Gly-34, followed by an occlusion formed by the His-37 side chains. The backbone atoms and histidine side chains are shown as stick and ball-and-stick representations, respectively. A surface generated using a 1.4-Å radius probe (32) is shown for each helix. The histidine side chains from each of the helices pack together in an arrangement that may greatly retard the flow of most ions through the channel. However, by alternately protonating/deprotonating the N ␦ and N ⑀ atoms of the histidine residues, it may be possible to shuttle protons through the channel in a mechanism that resembles the proton shuttle of carbonic anhydrase (23,24). The central pore in our model reaches its widest diameter (ϳ4 Å) near the center of the bilayer, just above His-37. The widening of the aqueous pore may be important for minimizing the dehydration energy of protons as they pass through the channel; charged species diffusing through an aqueous pore in an otherwise low dielectric environment encounter an unfavorable dehydration energy, which reaches a maximum near the center of the bilayer. This energy barrier may be lowered by widening the pore near the center of the bilayer (26). B, a modified model including amantadine within the pore showing possible interaction with the side chains of His-37. Amantadine is shown in stick representation with the amine group protonated. The model was generated by placing one molecule of amantadine within the M 2 pore, centered at the largest opening at Gly-34, doing one round of energy refinement (500 steps steepest descent/5000 steps conjugate gradient) with fixed M 2 atom positions, followed by a second round of energy minimization with no atoms fixed. The proposed cavity in M 2 is also believed to comprise the binding site for amantadine and BL-1743. Amantadine-resistant forms of the virus show mutations at sites that line the proposed central pore of the channel. In our model, the hydrophobic adamantyl group interacts with the lining of the pore, which is composed primarily of hydrophobic groups. Also, the ammonium group forms stabilizing H-bonded interactions with the electron lone pairs on the imidazole nitrogen atoms of His-37 (shown in A). Minimizations were performed in the absence of water using Discover (Biosym/Molecular Simulations Inc., San Diego, CA). The figure was generated using Insight95 (Biosym/MSI).

STRUCTURE 1
HEPES-NaOH, pH 7.5) until injection of ϳ40 ng of mRNA. After this time they were maintained in ND-96, pH 8.5, in order to minimize the proton flux due to the action of the M 2 ion channel. All solutions were equilibrated with room air except for solutions containing Cu(I), which were equilibrated with N 2 . In control experiments we found that the currents of oocytes expressing the M 2 protein incubated in Barth's solution equilibrated with N 2 (pH 7.5 or pH 6.2) for up to 20 min were indistinguishable from those of oocytes incubated at the same pH in Barth's solution equilibrated with room air. Metabolic labeling of oocytes and analysis of proteins by SDS-polyacrylamide gel electrophoresis was carried out as described previously (15).
Electrophysiological Recordings-24 -48 h after RNA injection, whole-cell currents were recorded with a two-electrode voltage-clamp apparatus consisting of a differential preamplifier (Nihon Kohden MEZ-7101, Tokyo, Japan) that recorded the voltage difference between a pipette (filled with 3 M KCl) located in the cell and another in the surrounding bath. A voltage-clamp amplifier (Nihon Kohden CEZ-1100) provided feedback current to the oocyte through a second intracellular pipette. Oocyte currents were recorded in standard Barth's solution (0.3 mM NaNO 3 , 0.71 mM CaCl 2 , 0.82 mM MgSO 4 , 1.0 mM KCl, 2.4 mM NaHCO 3 , 88 mM NaCl, 15.0 mM MES, 1 pH 6.2, or 15.0 mM HEPES, pH 7.5) or a modified Barth's solution as indicated. Amantadine hydrochloride (Sigma) (10 mM stock in Barth's solution), CuCl 2 , and BL-1743 (kindly provided by Dr. Mark Krystal, Squibb, Wallingford, CT) were diluted as indicated. To check that the oocytes did not develop nonspecific leakage currents during the recordings, we applied amantadine hydrochloride (100 M) for 2-5 min at the end of the measurements from each oocyte. Data from an oocyte were only used if the current in the presence of amantadine was less than 10% of the initial current.

RESULTS
Inhibition of Currents by Cu 2ϩ -Cu 2ϩ is a potent, time-dependent inhibitor of the channel activity of oocytes that express the M 2 protein ( Fig. 2 and Table I). In the absence of M 2 protein, Xenopus oocytes have very small currents in the range of pH 4.0-pH 9.0. Oocytes that express M 2 protein have currents that are dependent on pH of the bathing medium (14,25). For pH 7.5 the current is double the background current found in control oocytes. However, at pH 6.2 the current is about 8-fold higher than the current at pH 7.5 (14,25). Bathing oocytes that express wild-type M 2 protein in a solution that contains 100 M amantadine causes the currents to decrease to background levels within 2-3 min (14,16). Prolonged exposure of oocytes to Cu 2ϩ can be toxic, so we took a number of precautions to ensure that the observed inhibition was specific to the M 2 ion channel. Before applying media containing transition metals, we first confirmed that the whole-cell current of each oocyte was activated by low pH out . We also tested each cell for the inhibition of current by amantadine by applying the drug after washing out the transition metal. The data reported here are from oocytes whose currents were increased 4 -6-fold by decreasing pH out from pH 7.5 to pH 6.2 and whose currents were reduced to 100 -200 nA by amantadine (100 M for 2 min). These values for activation by low pH and inhibition by amantadine are within the range for M 2 -expressing oocytes in the absence of transition metals and thus provide assurance that the inhibition reported is specific to the M 2 ion channel and is not confounded with nonspecific leakage currents.
Addition of Cu 2ϩ to the bathing solution (pH out ϭ 6.2) gives rise to a time-dependent decrease in the amantadine-sensitive channel activity, approaching full inhibition at long times ( Fig.  2 and Table I). The time course of inhibition differs from that for amantadine or other hydrophobic drugs, which generally show simple, pseudo-order first order decay kinetics under conditions where the drugs are in large excess. Instead, the data for inhibition by Cu 2ϩ is biphasic, and the data can be fit by a sum of two exponentials ( Fig. 2 and Table I), representing a fast and a slow process with both relaxation times linearly proportional to the concentration of Cu 2ϩ between 100 and 1000 M. Indeed, we show below that the initial rapid rate (93 Ϯ 5 M Ϫ1 s Ϫ1 ) is associated with a nonspecific, partial block of the channel that is not dependent on His-37, whereas the slower process (18.6 Ϯ 0.9 M Ϫ1 s Ϫ1 ) requires the presence of histidine at position 37. The second order rate constant associated with this latter process is nearly 2 orders of magnitude slower than the value observed for inhibition of the M 2 channel activity (A/Udorn subtype) by amantadine. Fig. 3 and Table II illustrate the recovery of the amantadinesensitive channel activity following removal of Cu 2ϩ . Again, the data can be fit by two exponentials. Interestingly, the curves show a small initial burst, possibly reflecting dissociation from the low affinity site, followed by a slow recovery of channel activity representing dissociation from the high affinity site. Indeed, the fraction of the fast component of the recovery kinetics decreased with time of preincubation with Cu 2ϩ and was essentially unmeasurable after the long (over 20 min) incubation period with 100 M Cu 2ϩ (see Fig. 3 legend and Table II). The slow dissociation rate constant (1.8 ϫ 10 Ϫ4 s Ϫ1 ) corresponds to a half-life of approximately 2 h for copper bound to the high affinity site. The dissociation constant for binding of Cu 2ϩ to the low affinity site of M 2 can be estimated to be approximately 100 M from the ratio of the fast process on and off rate constants, assuming that these are kinetically independent (mechanistic analysis from Appendix A yields 400 M or 250 M, depending on the mechanism chosen). The corresponding dissociation constant for the high affinity site (using 1 The abbreviation used is: MES, 4-morpholineethanesulfonic acid. . Table I for parameters that were fitted to these data. the ratio of slow off to fast on rate constants) would be about 2 M (Appendix A yields 1.9 or 1.6 M), significantly tighter. pH Dependence of the Cu 2ϩ -binding Site-To help identify the nature of the binding sites, we first determined if they could be titrated within the range of pH we were able to test. This titration was done by measuring the time course of inhibition and recovery from inhibition in solutions of various pH values. These experiments paralleled previous experiments in which we showed that amantadine binds less strongly at low pH, and we showed that the pH dependence of the interaction required the presence of His-37. Similarly, the time course of inhibition was slower at acidic pH (Fig. 4), and the recovery was faster for the lower values of pH. We were able to test only the effects of pH in the range pH 5.2-6.2. Below pH 5.2 irreversible effects due to large proton currents flowing for many seconds were seen, and above pH 6.2 the current amplitude was too small to measure inhibition directly.
Voltage Dependence of Cu 2ϩ Binding-If the Cu 2ϩ -binding sites lie within the transmembrane region, then the rate of association of the positively charged copper ion may be accelerated at negative applied potentials. To test this possibility we measured the time course of inhibition by 250 M Cu 2ϩ at various holding voltages. We found that the rate of onset of inhibition was greater for more negative holding voltages ( Fig.  5 and Table III), consistent with the presence of at least one Cu 2ϩ -binding site that is located at least partially inside the electric field of the membrane.
Directionality of Cu 2ϩ Inhibition-Cationic, open-pore blocker molecules usually attenuate only the current originating from the side of the membrane to which they are applied. An example of this is the block of the outward, but not inward, K ϩ current of the squid axon by internally applied tetraethylammonium ϩ cation (28). We tested the directionality of the inhibition by Cu 2ϩ by measuring the current-voltage relationship of the oocyte (pH out ϭ 6.2) with slowly varying ramps of current before and at various times after the application of Cu 2ϩ . For these experiments, the leakage current in the presence of amantadine was subtracted in order to obtain an accurate estimate of the reversal potential, and the membrane voltage was restricted to ϩ50 mV to avoid activating endoge- . Table II for parameters that were fitted to these data.

TABLE II
Parameters fitted to the time course after exposure to Cu 2ϩ (see Fig. 3) Values represent means Ϯ S.D.; in three cases the standard deviation could not be calculated.  Lines are fits to double exponential functions as yielding the below tabulated fast and slow forward (k ff and k fs ) and backward (k bf and k bs ) rate constants. See Table III for parameters that were fitted to these data. nous channels. We found (in six cells) that inward and outward currents were attenuated equally (see Fig. 6) and that the reversal voltage remained constant with increasing percentage inhibition by Cu 2ϩ . Thus, inhibition by Cu 2ϩ , like inhibition by amantadine (see Fig. 6, inset), is bi-directional.
Binding of BL-1743 Prevents Binding of Cu 2ϩ -If Cu 2ϩ binds to an internal site, it should be possible to block binding by prior application of a compound that either competes for the same site or occupies the outer regions of the pore. The compound BL-1743 has been shown to be a reversible inhibitor of the currents of the M 2 ion channel (15). Mutations conferring resistance to BL-1743 map to the pore region immediately above His-37 (as viewed in Fig. 1), suggesting that BL-1743 penetrates deeply into the pore.
We tested the ability of BL-1743 to prevent inhibition by Cu 2ϩ . The test depended on the fact that recovery from inhibition by BL-1743 ( Fig. 7 and Ref. 15) is considerably faster than that from 0.5 mM Cu 2ϩ (4 min versus 4 h, respectively). To determine whether BL-1743 prevents inhibition by Cu 2ϩ , we performed the following steps: (i) inhibited the channel completely with BL-1743; (ii) while maintaining the concentration of BL-1743 constant, added 0.5 mM Cu 2ϩ to the solution for 5 min (such that the Cu 2ϩ would have completely inhibited the currents had it been applied alone); (iii) washed out free Cu 2ϩ from the recording chamber briefly (2 min) while BL-1743 was still maintained in the solution; (iv) washed out BL-1743 and measured the time course of recovery. Our earlier results demonstrated that the time course of recovery from inhibition by Cu 2ϩ alone is much slower than that from BL-1743. Thus, if Cu 2ϩ had gained access to an internal binding site in the presence of BL-1743, then during washout of BL-1743 the recovery would have been slow and incomplete. However, we found that the time course of the recovery during washout of BL-1743 did not differ from the time course that would have been measured had Cu 2ϩ not been applied (Fig. 7). Thus, BL-1743 prevented the binding of Cu 2ϩ to a presumably internal site.
Mutation of His-37 Eliminates the High Affinity Cu 2ϩ -binding Site-The above results strongly suggest that the high affinity Cu 2ϩ -binding site lies within the transmembrane pore of M 2 . To confirm these observations, we examined the ability of Cu 2ϩ to inhibit a number of variants of M 2 in which potential chelating groups were altered. The amino acid sequence of the transmembrane region of M 2 (residues 24 -43) is: Asp-Pro-Leu-Val-Val-Ala-Ala-Ser-Ile-Ile-Gly-Ile-Leu-His-Leu-Ile-Leu-Trp-Ile-Leu. The only titratable residue in the transmembrane  region is His-37, and this residue has been shown to be important for amantadine inhibition. Thus, we compared the ability of Cu 2ϩ to inhibit the mutant proteins M 2 -H37A and M 2 -H37G with its ability to inhibit the wild-type M 2 protein. In addition, we examined the role of the Asp-24, which lies near the N terminus of the ␣-helix, and Ser-31, which lines a portion of the predicted pore. Finally, wild-type M 2 protein has two cysteine residues at positions 17 and 19 in the extracellular domain. It appeared unlikely that they would be responsible for binding Cu 2ϩ because these residues participate in disulfide bonding to stabilize the homotetramer (4 -6). Nevertheless, we tested for this possibility by using a previously characterized "cysteineless" mutant in which each of the three cysteine residues of the M 2 protein, found at positions 17, 19, and 50, were mutated to serine. We examined the time course of inhibition and recovery from inhibition of the currents of the M 2 -H37A and M 2 -H37G mutant proteins by 0.1 and 1.0 mM Cu 2ϩ . The limiting fractional inhibition was less than for the wild-type protein, and the rates of inhibition and recovery for both of these mutant proteins were faster than for the wild-type M 2 protein (Fig. 8 and Table  IV). Unlike the data for the wild-type protein, that for both mutants could be well fit by single exponentials. The rate of inhibition was proportional too, and the rate of recovery from inhibition was independent of the Cu 2ϩ concentration. Both rates were also independent of membrane voltage. Taken together with the above kinetic data for the wild-type protein, these data strongly suggest that mutation of His-37 results in the removal of the high affinity site, while maintaining the integrity of the peripheral binding site.
To determine the possible location of the low affinity Cu 2ϩbinding site in M 2 -H37A and M 2 -H37G, we measured the voltage dependence and pH dependence of their inhibition. The voltage dependence for both proteins (see Fig. 8 and Table IV for M 2 -H37G) is similar and much less significant than for the wild-type protein ( Fig. 5 and Table III). Moreover, we found that the inhibition of neither mutant protein was strongly dependent on pH (see Fig. 9 for M 2 -H37G), in contrast to the greater pH dependence found for the wild-type protein (Fig. 4). These results suggest that replacement of His-37 with a residue incapable of coordinating Cu 2ϩ leaves the ion channel with a low affinity binding site that is located near the outside of the electric field.
We also tested the possibility that the cysteine residues located at positions 17 and 19 (29), the charged aspartate residue at position 24, or the polar serine residue at position 31 might provide a low affinity binding site for Cu 2ϩ that may be partly responsible for the rapid recovery from inhibition by 1 mM Cu 2ϩ (see Fig. 3, Table II, and "Discussion"). We applied Cu 2ϩ to oocytes expressing the cysteineless (29) mutant protein and the mutant proteins M 2 -D24A, M 2 -S31A, and M 2 -S31G. The only departure from the behavior of wild-type protein for any of these proteins in response to application of Cu 2ϩ (0.1 and 1.0 mM) was that the rate of onset of inhibition was slightly slower for the M 2 -D24A mutant protein than for the wild-type protein. Perhaps this difference was due to the reduced charge of the M 2 -D24A protein at the mouth of the presumed pore region. These results suggest that neither Cys-17, Cys-19, Asp-24, nor Ser-31 participate directly in the inhibition of the channel by Cu 2ϩ .
Metal Ion Specificity-To characterize further the ion specificity of the low and high affinity binding sites in M 2 , we evaluated a series of transition metal ions, including Cd 2ϩ , Mg 2ϩ , Mn 2ϩ , Ni 2ϩ , Pt 2ϩ , and Zn 2ϩ . As was expected for the FIG. 7. Prior exposure to BL-1743 prevents inhibition by Cu 2؉ . The oocyte currents were measured at pH 6.2 to determine the maximal current. The current was then completely inhibited by BL-1743 (0.1 mM). While inhibited, 0.5 mM Cu 2ϩ was applied to the oocyte for 5 min, a long enough time to achieve full inhibition from which recovery would have been very slow had the Cu 2ϩ been applied alone (see Fig. 2). After 2 min washout of Cu 2ϩ , the BL-1743 was finally washed out, and the time course of the fraction of the recovery was measured (q). This recovery was compared with the recovery from 0.1 mM BL-1743 alone (Ⅺ) and the recovery from 0.5 mM Cu 2ϩ alone (ࡗ). Note that the time course of recovery of current of oocytes treated with BL-1743 before Cu 2ϩ (q) was not distinguishable from that of oocytes inhibited by BL-1743 alone (Ⅺ; data from Ref. 15).

FIG. 8. Inhibition of the current of the M 2 H37-G mutant protein by 0.1 mM Cu 2؉ is rapid, incomplete, and reversible.
Inhibition is also relatively independent of membrane voltage. q, Ϫ40 mV; f, Ϫ20 mV; OE, 0 mV; , ϩ20 mV. See Table IV for parameters that were fitted to these data. The lines show single exponential fits with the following rate constants: inset, the time course of inhibition by 0.1 mM Cu 2ϩ of the M 2 -H37A mutant protein (average of four cells) is compared with that of the wild-type protein (data from Fig. 1 and Fig. 2); note the rapid onset and offset of inhibition; black bars show times of application of Cu 2ϩ . See Table IV for parameters that were fitted to these data. replacement of a relatively soft for a hard metal ion, the replacement of Cu 2ϩ with Mg 2ϩ led to essentially no inhibition, indicating that Mg 2ϩ does not interact with either site. Similarly, Mn 2ϩ , which has ligand preferences similar to Mg 2ϩ , but also shows some "soft" character, inhibited the channel by less than 10% at 1 mM concentration. We next examined Cu ϩ and Zn 2ϩ , which have ligand preferences similar to Cu 2ϩ but have a preference for octahedral or tetrahedral complexes and would be less likely to assume the distorted square pyramidal complex hypothesized for Cu 2ϩ . These metal ions showed partial inhibition of the channel (Table V), and recovery from inhibition by these metal ions was nearly complete within 2-5 min.
We therefore tentatively assign their effects to interactions with the low affinity site. In a similar manner, we examined Ni 2ϩ and Pt 2ϩ , which have a preference for forming square planar complexes; again these ions gave rise to only partial, rapidly reversible inhibition of the channel ( Table V). All of these data suggest that the high affinity metal ion-binding site is quite specific for Cu 2ϩ , whereas the remaining, low affinity site is less specific and able to interact with a variety of metal ions. To confirm this suggestion, we tested the ability of 1 mM Cd 2ϩ , Ni 2ϩ , and Zn 2ϩ to inhibit the currents of the M 2 -H37A and M 2 -H37G mutant proteins. Indeed, all three metal ions inhibited these mutants in a manner similar to that observed for the wild-type protein (data not shown).

DISCUSSION
The experiments described here demonstrate that Cu 2ϩ and other transition metals are capable of inhibiting the M 2 ion channel of influenza A virus. Inhibition and recovery from inhibition can be explained by a model with two binding sites, each capable of inhibiting the current upon binding a ligand. One site, of low affinity and low specificity for Cu 2ϩ , is located near the outside of the electric field of the membrane and only partially impedes ionic current through the channel. The second site has higher affinity for Cu 2ϩ , is located inside the applied electric field, and more completely blocks current flow. The high affinity site is probably formed by the association of imidazole side chains of His-37 from the transmembrane helices of the M 2 tetramer.
The strong voltage dependence of the level of inhibition of the current of the wild-type channel by Cu 2ϩ (Fig. 5 and Table III) suggests that the imidazole of His-37 is tightly coupled to the high affinity binding site for Cu 2ϩ . This is consistent with models showing His-37 to be located well inside the presumed transmembrane domain (18,22), which begins with Pro-25. Also, inhibition by low [Cu 2ϩ ] is dependent on pH out . The dependence of Cu 2ϩ inhibition on pH out occurs within a range of pH that modulates the currents of the wild-type ion channel. This modulation of the ionic current has been shown to depend on the presence of His-37. Since His-37 is the only residue in the transmembrane domain that is titratable in the range of pH values studied, it is likely that this residue is also responsible for the pH dependence of inhibition rate. A reasonable explanation for the slower onset of inhibition at low pH (Fig. 4) is that competition of H ϩ and Cu 2ϩ for a binding site on the imidazole of His-37 favors H ϩ at low pH. Finally, the presence of the reversible inhibitor BL-1743 prevents inhibition by low [Cu 2ϩ ] (Fig. 7). BL-1743 has been shown to inhibit the M 2 ion channel, and several mutants that are not affected by the compound have been identified (15). Most of these mutants map to residues of the transmembrane domain located between Pro-25 and His-37, suggesting that the compound occupies the outer region of the pore. It is probable that the compound prevented inhibition by restricting access of Cu 2ϩ to His-37. These results point to the predictive power of the current model for M 2 and also add considerably to the mechanistic understanding of this channel. We have recently proposed that His-37 lies within the conduction pathway of the channel, where it serves as a relay to selectively shuttle protons down their concentration gradient into the virus. In this mechanism, the channel conducts more efficiently at low pH primarily because of an increase in the concentration of the protonated histidine. His-37 allows efficient shuttling in vivo because it has a pK between pH 5 and 7 where the proton is not too strongly bound. The finding that Cu 2ϩ binds to a high affinity site associated with His-37 is consistent with this because the normal ligation would require the histidine to be in its unprotonated, basic form.
Although our experiments indicate that the inner, high affinity binding site is probably His-37, we were not able to identify a single low affinity, exterior site. We mutated each residue external to His-37 whose side chain might be thought to bind divalent metal ions (Cys-17, Cys-19, Asp-24, and Ser-31), and we found that inhibition by Cu 2ϩ was essentially the same as that found for the wild-type M 2 protein.
As many ligand-and voltage-gated ion channels are modified by transition metals, it would be of interest to compare the effects of transition metals on the M 2 ion channel with the

ion channel activity by transition metals
The inhibition of these metals was measured as described for Cu 2ϩ with the exception that Cu ϩ was dissolved as the tetrakisacetonitrile salt in a medium equilibrated with N 2 (see "Experimental Procedures"). Time is the elapsed time between the start of incubation with the metal ion and the time when the measurement was made. effects on other ion channels. Among the best-studied of these channels is the Na ϩ channel; however, the inhibition of the Na ϩ channel by transition metals differs greatly from the inhibition of the M 2 ion channel by Cu 2ϩ . 1) Cu 2ϩ inhibits both inward and outward currents of the M 2 ion channel, but transition metals inhibit the inward current of the cardiac Na ϩ channel.
2) The Cu 2ϩ -binding site in M 2 is highly specific for Cu 2ϩ , whereas Cd 2ϩ and Mn 2ϩ inhibit the cardiac Na ϩ channel but do not inhibit the M 2 ion channel.
3) The high affinity binding site differs; the imidazole of His-37 of the transmembrane domain of the M 2 ion channel is probably the high affinity binding site for Cu 2ϩ , but Cys-401 of the pore region of the cardiac Na ϩ channel is thought to be the high affinity binding site for Cd 2ϩ and Zn 2ϩ (30,31). Such differences in metal binding indicate that these ion channel proteins have stable, but different, structures that present coordinating ligands of different affinities to permeating metal ions. The results of our experiments have important implications for future work. First, Cu 2ϩ has an unpaired electron and thus could serve as a probe for structural studies of the M 2 molecule using either electron paramagnetic resonance or ENDOR methods. Second, the inhibition of the M 2 ion channel by low [Cu 2ϩ ] shares several important properties with the inhibition by amantadine; both are slowed by low pH, both inhibit currents in both inward and outward directions, and both depend on the presence of His-37. If Cu 2ϩ and amantadine interact with the imidazole of His-37 in a similar fashion, then information gained using Cu 2ϩ may be helpful in the design of inhibitors that are more useful than amantadine.

APPENDIX
This Appendix derives kinetic models for two-state binding of Cu 2ϩ to the M 2 ion channel protein. We envision an outer site located in the pore of the channel close to the outside of the membrane electric field and an inner site located near the inside and well within the transmembrane electric field. Binding of Cu 2ϩ to the outer site is assumed to be capable of inhibiting the current to a fraction of its initial value, whereas binding to the inner site inhibits completely. States are defined by their location and Cu 2ϩ occupancy: Sa is the state defined by unoccupied inner and outer sites, Sb by occupancy of only the outer site, Sc by only the inner site, and Sd by both sites occupied. Different, non-equivalent kinetic schemes can be devised to connect the states. We used two as follows: the first (Scheme 1) prohibiting Sa to Sc but allowing a direct, reversible Sa to Sd transition; and the second (Scheme 2) allowing Sa to Sc and prohibiting Sa to Sd. Note that no distinction is made in this mechanism between the two sides of the membrane so it is possible that the Sd to Sa transition involves Cu 2ϩ entering the cytoplasm of the oocyte.
The forward and reverse reaction rate constants for each transition are those defined in the following equations of Schemes 1 and 2. dSa dt ϭ Ϫ͑k 1f ͓Cu 2ϩ ͔ ϩ k 4f ͓Cu 2ϩ ͔ 2 ͒Sa ϩ k 1b Sb ϩ k 4b Sd dSb dt ϭ Ϫ͑k 2f ϩ k 1b ͒Sb ϩ k 1f ͓Cu 2ϩ ͔Sa ϩ k 2b Sc dSc dt ϭ Ϫ͑k 3f ͓Cu 2ϩ ͔ ϩ k 2b ͒Sc ϩ k 2f Sb ϩ k 3b Sd dSd dt ϭ Ϫ͑k 3b ϩ k 4b ͒Sd ϩ k 4f ͓Cu 2ϩ ͔ 2 Sa ϩ k 3f ͓Cu 2ϩ ͔Sc SCHEME 1 dSa dt ϭ Ϫ͑k 1f ϩ k 2f ͓͒Cu 2ϩ ͔Sa ϩ k 1b Sb ϩ k 2b Sc dSb dt ϭ Ϫ͑k 3f ͓Cu 2ϩ ͔ ϩ k 1b ͒Sb ϩ k 1f ͓Cu 2ϩ ͔Sa ϩ k 3b Sd dSc dt ϭ Ϫ͑k 4b ͓Cu 2ϩ ͔ ϩ k 2b ͒Sc ϩ k 2f ͓Cu 2ϩ ͔Sa ϩ k 4b Sd dSd dt ϭ Ϫ͑k 4b ϩ k 3b ͒Sd ϩ ͑k 4f Sc ϩ k 3f Sb͓͒Cu 2ϩ ͔ SCHEME 2 These equations, with the initial condition that Sa ϭ 1 and all other states are empty, were used (by the program MLAB; Civilized Software, Inc., Bethesda, which has a built-in differential equation solving routine (33)) to generate current-time records for the three different Cu 2ϩ concentrations (100, 500, and 1000 M). Copper concentrations were taken to be step functions with transitions corresponding to those applied in each experiment, and the rate constants were estimated by curve-fitting. The parameters fitted to the data using each of these model schemes are shown in Table IA. Standard errors, as reported by the MLAB program, are for qualitative comparison only because they depend on the unsupportable assumption that the curve-fitting error function varies linearly with the fitting parameters. The number of parameters in the models discourages quantitative interpretation of these results.
To compare the site affinities for Cu 2ϩ , the reactions Sa ϩ Cu 2ϩ 3 Sb and Sa ϩ Cu 2ϩ 3 Sc are the relevant equilibria. The "operational" site dissociation constants K outer and K inner corresponding to the above reactions can be calculated from the rate constant ratios. For Scheme 1, K outer ϭ k 1b /k 1f ϭ 0.4 mM and K INNER ϭ K OUTER k 2b /k 2f ϭ 1.9 M. For Scheme 2, K outer ϭ k 1b /k 1f ϭ 0.25 mM; K inner k 2b /k 2f ϭ 1.6 M.