Mechanism and Kinetics of Copper Complexes Binding to the Influenza A M2 S31N and S31N/G34E Channels

Abstract

Copper(II) is known to bind in the influenza virus His37 cluster in the homotetrameric M2 proton channel and block the proton current needed for uncoating. Copper complexes based on iminodiacetate also block the M2 proton channel and show reduced cytotoxicity and zebrafish-embryo toxicity. In voltage-clamp oocyte studies using the ubiquitous amantadine-insensitive M2 S31N variant, the current block showed fast and slow phases, in contrast to the single phase found for amantadine block of wild-type M2. Here, we evaluate the mechanism of block by copper adamantyl iminodiacitate and copper cyclooctyl iminodiacitate complexes and address whether the complexes can coordinate with one or more of the His37 imidazoles. The current traces were fitted to parametrized master equations. The energetics of binding and the rate constants suggest that the first step is copper complex binding within the channel, and the slow step in the current block is the formation of a Cu-histidine coordination complex. Solution-phase isothermal titration calorimetry and density functional theory (DFT) calculations indicate that imidazole binds to the copper complexes. Structural optimization using DFT reveals that the complexes fit inside the channel and project the Cu(II) toward the His37 cluster, allowing one imidazole to form a coordination complex with Cu(II). Electrophysiology and DFT studies also show that the complexes block the G34E amantadine-resistant mutant despite some crowding in the binding site by the glutamates.

Significance

Copper complexes are known to block amantadine-resistant influenza infections of cell cultures through inhibition of proton transport by M2. The kinetics of current block in transfected oocytes reveal that the binding affinity of Cu(II) iminodiacetates of amantadine and cyclooctylamine are consistent with Cu coordination by a channel lumen residue. Density functional theory calculations reveal plausible imidazole coordination configurations in aqueous solution and with His37 in the channel lumen, consistent with isothermal titration calorimetry measurements of the heat of imidazole binding to the complexes.

Introduction

Influenza A virus pandemics of zoonotic origin occur every few decades and cause respiratory disease globally. The influenza A M2 channel is responsible for the acidification of the influenza virus interior, which leads to the separation of ribonuclear proteins from the virus, at which point replication begins. Protons are transported to the viral interior through the protonation of imidazole nitrogens in the His37 cluster (Fig. 1). Amantadine and rimantadine block the M2 wild-type (WT) variant, but they do not block the ubiquitous M2 S31N variant nor other naturally occurring variants including V27A, A30T, and G34E (1) and are no longer FDA approved as influenza A antiviral drugs. Numerous studies have been carried out in search of organic compounds to block these mutants (2, 3, 4, 5, 6, 7, 8, 9). None has been shown to be effective against all naturally occurring influenza A M2s (10), although a few compounds have recently been identified that can block some of the most common single and double mutants (11).

Figure 1.

M2 S31N ssNMR Structure (2KQT). The model shows the transmembrane domain (residues 22–46). The left image is a top-down view (N-terminus to C-terminus). The right image is a side view with two of the peptide backbones and the side chains hidden.

The His37 cluster is an excellent target for an antiviral because it is almost perfectly conserved in nature (12), probably because of its key role in selecting (13) protons for transport into virions after endocytosis. The His37 cluster was shown to be a viable binding site for copper complexes (12) based on the knowledge that Cu²⁺ binds to imidazole (14) and blocks the M2 channel (15) from a position between the His37 and Trp41 clusters (16). Previous work using the electrophysiology two-electrode voltage-clamp (TEVC) method showed that cyclooctyl iminodiacitate (Cu(CO-IDA)) and copper adamantyl iminodiacitate (Cu(AMT-IDA)) block the M2 S31N variant at low (therapeutic) concentrations, similar to AMT block in M2 WT. A two-phase block was observed: an initial fast phase followed by a slow phase (12). By site-directed mutagenesis of the His37 to Ala37, the complexes were shown to interact with the His37 cluster. The mutagenesis resulted in elimination of the slow phase, leaving the initial fast phase of block followed by complete wash-out in complex-free perfusate (12).

In vitro, Cu(CO-IDA) and had low cytotoxicity and was potent at submicromolar concentrations against A/Calif/07/2009 H1N1, which contains M2 S31N (17). In an evaluation of M2 resistance development against these copper complexes, 10 passages were insufficient for resistance development, in contrast to resistance to amantadine, which develops after only a few passages (17). However, there is another functional (18) naturally occurring mutation, G34E, that could interfere with binding of the copper complexes because of the proximity to the binding site and added bulk of the glutamate side chains. Although this mutation did not appear spontaneously during virus passaging, we thought it prudent to test the copper complexes in the M2 S31N-G34E variant using TEVC and to explore its possible copper complex binding configurations with density functional theory (DFT).

This research explores the mechanism by which Cu(CO-IDA) and Cu(AMT-IDA) block the M2 channel. We explore the binding kinetics and configurations of the copper complexes with M2 S31N. A global fit of electrophysiology data allows the determination of rate constants, providing insight into the coordination and dissociation processes. Isothermal titration calorimetry (ITC) was used to evaluate the enthalpy of bonding and dissociation constant for the binding of imidazole to the copper complexes. Quantum mechanical DFT calculations provide a model for the configuration of the copper complexes bound to histidines in the channel.

Materials and Methods

A/Udorn/72 H3N2 M2 S31N-G34E mRNA synthesis

The preparation of messenger RNA (mRNA) for oocytes with the influenza A/Udorn/72 H3N2 M2 S31N protein were reported previously (12). Here, we modified that pGEM3 plasmid to add a G34E mutation. A 393 basepair DNA fragment containing full-length A/Udorn/72 H3N2 M2 with both the S31N and the G34E mutations was obtained from Twist Biosciences (San Francisco, CA). The fragment was digested using BamHI and HindIII restriction nucleases and cloned into the pGEM3 plasmid, digested with the same restriction nucleases, to create the A/Udorn/72 H3N2 M2 S31N-G34E gene-containing plasmid. The plasmid was transformed into chemically competent Escherichia coli by standard methods. The plasmid was harvested using the Zymo Miniprep Kit (Zymo Research, Irvine, CA). To confirm that no mutations were introduced, the M2 DNA segment was PCR amplified and Sanger sequenced (Fig. S1). After confirmation, the PCR product was transcribed using the mMESSAGE mMACHINE T7 ULTRA Transcription Kit (Thermo Fisher, Waltham, MA) to prepare mRNA for oocyte injections.

Electrophysiology

Here, we extend the experiments reported previously (12) by adding higher concentrations of metal complexes, normalizing each trace after a limited leak subtraction, and averaging together such traces from three separate experiments. For this, Cu(AMT-IDA)⋅5H₂O (418.93 g/mol) and Cu(CO-IDA)⋅3H₂O (358.88 g/mol) were synthesized according to previously published procedures (12). Oocytes from Xenopus laevis (Ecocyte, Austin, TX) were maintained in ND-96²⁺ solution (96 mM NaCl, 2 mM KCl, 1.8 mM CaCl₂, 1 mM MgCl₂, 2.5 mM sodium pyruvate, 5 mM HEPES-NaOH (pH 7.4)) at 17°C until injection of ∼40 ng of A/Udorn/72 H3N2 M2 S31N or A/Udorn/72 H3N2 M2 S31N-G34E mRNA using a Nanoject II (Drummond Scientific, Broomall, PA). After injection, the oocytes were maintained at 4°C in ND96²⁺ (pH 7.4) until electrophysiological recording. 72 h after mRNA injection, whole-cell currents were recorded with a TEVC apparatus at a membrane potential Vm = −20 mV, room temperature, in Barth’s solution (0.3 mM NaNO₃, 0.71 mM CaCl₂, 0.82 mM MgSO₄, 1.0 mM KCl, 2.4 mM NaHCO₃, 88 mM NaCl, 15.0 mM HEPES (pH 7.5)). Inward current was induced by perfusion with Barth’s pH 5.3 (15.0 mM MES instead of 15.0 mM HEPES). MES was chosen as a Good buffer for noninteraction with Cu(II) (19). The oocytes were then perfused by Barth’s pH 5.3 with Cu²⁺(aq), Cu(CO-IDA), or Cu(AMT-IDA) at concentrations 0.1, 0.5, or 1 mM. A wash-out was done using Barth’s pH 5.3 without complex. Noninjected oocytes were also tested with the same acid perfusion protocols and concentrations of complex to assess the possible native acid-activated channel and Cu²⁺(aq), Cu(CO-IDA), or Cu(AMT-IDA)-induced leak current in the oocytes. Leak-current-subtracted current traces were obtained from each of three oocytes for each of the concentrations, normalized to run from zero to 1.0 by dividing each point in the trace by the maximal inward current, and averaged.

Global nonlinear least-squares curve fit

The two-phase block in the current traces (i.e., fast and slow phase) produced by perfusion with Cu²⁺(aq), Cu(CO-IDA), or Cu(AMT-IDA) were analyzed using the serial two-site binding model:

$$O\underset{k_4}{\overset{k_1}{\leftrightarrow }}C1\underset{k_3}{\overset{k_2}{\leftrightarrow }}C2.$$ (1)

The open state, O, is the unblocked M2 channel where proton conductance is unimpeded before perfusion with blocking agents. C₁ represents the initial bound state. Because the fast block never reaches completeness before the slow block begins, the permeability of C1 cannot be directly ascertained. We assume that the complex only partially blocks proton current in this state and we represent the fraction of current transported in this state with the fitting parameter f. Occupancy of the complex in the second binding site is represented by state C2, which is assumed to be a fully blocked M2 state because at high concentrations and long exposure, full block is achieved. To obtain the rate constants, k_i, they and f were used as parameters in Eq. 2 to fit the compound wash-in and wash-out current traces, similar to the method used to fit adamantanamine block in M2 WT and S31N currents (20).

$$I\left(t\right)=iN\left(P_O\left(t\right)+f{P}_{C1}\left(t\right)\right).$$ (2)

The single-channel current at the applied membrane potential is i, and the number of channels in the cell membrane is N. During drug wash-in, I(t), the M2 S31N oocyte current, starts at a maximal level because the probability of being in the open state, P_O(0) is 1 and declines with time as the sum of two exponentials toward a steady state level of block as the drug binding in the two states reaches equilibrium. The probabilities of the other two occupancy states are also functions of time, starting at no occupancy (P_C1(0) = P_C2(0) = 0) and approaching their equilibrium levels with a difference that also falls as the sum of two exponentials. During wash-out, the process is reversed, and all three state probabilities relax back to their predrug occupancies as the sum of two exponentials. The underlying differential equations governing the relaxation of the probabilities after a change in boundary condition (bath drug concentration) and their solutions are given for both wash-in and wash-out in Supporting Materials and Methods. The rate constants and coefficients of the exponentials differ with drug concentration, and for the wash-out, the coefficients also depend on the state probabilities achieved during the finite wash-in. Nonlinear least-squares curve fitting was carried out in MATLAB R2018a (The MathWorks, Natick, MA) using the Levenberg-Marquardt algorithm. Uncertainties in the parameters were derived from the error matrix, i.e., the square roots of the diagonal elements (21).

To understand the strength of the interaction between M2 and the copper complexes, the effective equilibrium dissociation constant (dissociation from the second binding site to the open state, O) was calculated as the product of the outer and inner dissociation constants, K_d1 and K_d2, using Eq. 3:

$$K_{d_{eff}}=K_{d1}{K}_{d2}=\frac{k_4}{k_1}\frac{k_3}{k_2}.$$ (3)

ITC

For ITC, Tris-buffered solutions of imidazole were prepared by adding imidazole to 8.0 mL of Milli-Q water, adding 1.0 mL of 20 mM Tris, adjusting the pH by adding either 10.0 or 1.0 M NaOH or HCl, and adding water to reach 10.0 mL of solution. Solutions of complexes were made by adding 41.5 mg Cu(AMT-IDA) or 27.3 mg Cu(CO-IDA) to 8 mL of Milli-Q water, adding 1.0 mL of 20 mM Tris, and sonicating and/or heating, until the complex dissolved (20–60 min). The pH was then adjusted to the target pH with 10.0 or 1.0 M HCl or NaOH and water added to reach 10 mL. Concentrations of copper complexes were verified by measuring ultraviolet absorbances of the complexes and using the Beer-Lambert law and by inductively coupled plasma mass spectrometry analysis. The molar extinction coefficient for Cu(AMT-IDA) is 5300 M⁻¹ cm⁻¹ and for Cu(CO-IDA) is 3700 M⁻¹ cm⁻¹ at 252 nm (12).

ITC experiments were performed using a Nano-ITC low-volume calorimeter (TA Instruments, Layton, UT) equipped with gold reference and sample cells of 170 μL, having a minimal detectable heat of 0.05 μJ and a baseline stability of 0.02 μW/h. All titration runs were carried out with a 50 μL injection syringe (minimal injection volume 0.06 μL) at 25°C and a stirring rate of 350 rpm. Both the syringe and well solutions were adjusted to the desired pH of 6, 7, or 8 using HCl or NaOH. 1 or 2 μL volumes of imidazole solution were injected into ∼1 μM copper complex solutions (Cu(AMT-IDA) or Cu(CO-IDA)). Areas under the heat-of-injection curve for each experiment were corrected for heat of dilution and fitted using the “independent fit” in the NanoAnalyze software (TA Instruments, New Castle, DE), and fit parameters were averaged. The number of experiments for each complex ranged from two to six. K_d-values depend primarily on the slope of the titration curve, ΔH on the intercept, and n, the number of imidazole binding sites on the complexes, modulated both. The value of n showed no pattern with complex or pH and was therefore averaged over all fits, giving n = 1.01 ± 0.37. On the grounds that variations in n probably reflected experimental fluctuations in complex integrity or solubilization, the modulating effects of n on ΔH and K_d were incorporated in their averages. The first heat-of-injection peak in an experiment was routinely discarded because of dilution of the injectant by diffusion. Additionally, some experiments were rejected because of drifting baseline temperature or evidence of solution contamination due to incomplete cleaning of the calorimeter cell.

Quantum chemical modeling of Cu(AMT-IDA) and Cu(CO-IDA)

As a quantum chemical model for M2 channel binding, DFT calculations were used to examine the coordination enthalpy of neutral imidazole to Cu(AMT-IDA), structure previously determined (12,22), and Cu(CO-IDA). These calculations were performed using Terachem (v1.93P; PetaChem, Los Altos Hills, CA) using the ωB97X-D3 functional (23). Calculations were carried out with the default Terachem COSMO continuum solvent model (24). All structures were optimized to stationary points and confirmed as minima by vibrational frequency analysis using ωB97X-D3/6-31G^∗∗ with the LANL2DZ used for copper. A subsequent self-consistent field energy evaluation was performed with ωB97X-D3/6-311+G^∗∗, using LANL2DZ for copper, so that final enthalpy values, which contain the ΔG of solvation estimate, are ωB97X-D3/6-31G^∗∗[LANL2DZ Cu]//ωB97X-D3/6-311+G^∗∗[LANL2DZ Cu].

Quantum chemical modeling: copper complexes binding His37 in S31N and S31N-G34E M2

Previously, molecular dynamics simulations illustrated the accessibility of the channel lumen to these metal complexes (17). Here, we explore possible binding configurations to a His37 side chain to determine whether complexation is feasible. Geometry optimization of the copper complexes binding to one histidine in the His37 cluster of the M2 S31N or S31N-G34E structure were done using Gaussian 16 (Gaussian, Wallingford, CT). Using the VMD Mutator plugin, homology models for M2 S31N and S31N-G34E were created from the 2KQT NMR structure, relaxed first in a dimyristoylphosphatidylcholine bilayer with molecular dynamics simulation. Each glutamate side chain was negatively charged. The M2-copper complex system was optimized in gas phase. ONIOM (25) was set up with the high-level QM region including all four histidines in the His37 cluster, Cu(AMT-IDA) or Cu(CO-IDA), and all four glycine or glutamate side chains at position 34. The low-level region included all other regions of the M2 channel. The ωB97X-D DFT method and 6-31G^∗∗ basis set were used to optimize the QM region, except for the copper atom, for which the LANL2DZ basis was applied. The PM6 semiempirical method was used to optimize the low-level regions.

Results and Discussion

Binding kinetics: electrophysiology with M2 S31N

Binding kinetics were determined from the time course of inward current reduction in transfected oocytes, which are voltage clamped to carry proton current from the extracellular N-terminus of the full-length channel to the intracellular C-terminus. Normalized current data and theoretical curve fits for M2 S31N perfused with three concentrations of the three blockers are shown in Fig. 2. Block is slow to develop, so to expedite its measurement on the timescale of oocyte viability, supraclinical concentrations were used. The normalized current traces are pointwise averages from three different X. laevis cells with leak current subtracted from each trace. The simultaneous optimization in the global fit of the parameters from the block and wash-out electrophysiology data at varying concentrations constrained the parameters sufficiently to extract the rate constants (k₁, k₂, k₃, and k₄) and the fractional current, f, of the first binding site. For all three blockers, f was optimized at ≤0.02, i.e., complete block in the C1 state.

Figure 2.

Cu²⁺(aq) (a and b), Cu(AMT-IDA) (c and d), and Cu(CO-IDA) (e and f) in M2 S31N tested at 0.10, 0.50, and 1.0 mM (top to bottom). (a, c, and e) Perfusion of drug solution (wash-in) is shown. (b, d, and f) Drug-free perfusion (wash-out) representing wash-in concentrations of 0.10, 0.50, and 1.0 mM top to bottom is shown. Solid lines represent measured currents (average of three traces). Dotted lines represent two-exponential relaxation to the equilibrium levels, which are <0.0013 in all wash-in cases. Fast exponential time constants in seconds, low concentration (top) to high (bottom) for wash-ins: (a) 0.89, 0.71, 0.56; (c) 1.7, 0.58, 0.32; and (e) 3.9, 2.1, 1.3. For wash-outs (concentration 0.0 mM): (b) 0.95, (d) 3.1, and (f) 5.0. Slow exponential time constants (same units and pattern): (a) 181, 46, 29; (c) 71, 42, 38; (e) 133, 50, 41; (b) 290,000; (d) 11,000; and (f) 13,000.

At 1.0 mM Cu²⁺(aq), the M2 S31N current is almost completely blocked in 5 min (Fig. 2 a). No significant wash-out was observed for Cu²⁺(aq) after 4 min (Fig. 2 b). The rate constants obtained in the slightly over-constrained global fit from the blocking and wash-out traces are 754 M⁻¹ s⁻¹ (k₁) and 0.0824 s⁻¹ (k₂) for the binding to the first and second sites, respectively, and 3.6 × 10⁻⁴ s⁻¹ (k₃) and 0.97 s⁻¹ (k₄) for the dissociation from the second and first binding sites, respectively (Table 1). The corresponding effective equilibrium constant for the dissociation reaction of Cu²⁺(aq) from M2 S31N, K_d,eff, is 56 nM (Table 2). This very low dissociation constant is primarily due to the strong binding at the second site, but it is incomplete on the experimental timescale at all three concentrations because of the low association rate constants, k₁ and k₂. We propose that the strong binding is due to the formation of a coordination complex between Cu(II) and a histidine residue in the channel and that the first binding site is located at or near the well-known amantadine binding site in the WT channel (26).

Table 1.

Association and Dissociation Rate Constants for the First Binding Site (k₁ and k₄) and the Second Binding Site (k₂ and k₃) from the Global Nonlinear Least-Squares Fit for Cu²⁺(aq), Cu(CO-IDA), and Cu(AMT-IDA) in M2 S31N

Global Fit Results
Compound	k1 (M−1 s−1)	k2 (s−1)	k3 (s−1)	k4 (s−1)	Weighted Reduced χ2
Cu2+(aq)	754 ± 5	0.082 ± 0.012	0.36 × 10−5 ± 0.08 × 10−5	0.97 ± 0.09	0.00174
Cu(CO-IDA)	2876 ± 3	0.029 ± 0.004	2.9 × 10−5 ± 0.4 × 10−5	0.29 ± 0.01	0.00192
Cu(AMT-IDA)	594 ± 4	0.032 ± 0.003	1.5 × 10−5 ± 0.1 × 10−5	0.17 ± 0.01	0.00104

Table 2.

Dissociation Constants for the First and Second Binding Steps and for the Holistic Binding Process, i.e., from Bulk to Complexed State

Compound	Kd1 (M) = k4/k1	Kd2 = k3/k2	Kd,eff (M) = Kd1 × Kd2
Cu2+(aq)	1.29 × 10−3 ± 1.06 × 10−4	4.37 × 10−5 ± 1.13 × 10−5	5.62 × 10−8 ± 1.53 × 10−8
Cu(CO-IDA)	1.01 × 10−4 ± 3.83 × 10−6	1.00 × 10−3 ± 1.98 × 10−4	1.01 × 10−7 ± 2.03 × 10−8
Cu(AMT-IDA)	2.86 × 10−4 ± 2.36 × 10−5	4.69 × 10−4 ± 5.67 × 10−5	1.34 × 10−7 ± 1.96 × 10−8

Errors are propagated from those in Table 1 assuming no correlations.

Cu(AMT-IDA) blocks M2 S31N (Fig. 2 c) and produces blocking kinetics similar to Cu²⁺(aq). Almost complete block was achieved for 0.50 and 1.0 mM solutions by 60 min (Fig. 2 c), and no significant wash-out was observed by 5 min (Fig. 2 d). However, there are some significant differences in the fitted parameters, with k₁ and k₂ lower, whereas k₃ is higher (Table 2), corresponding to a larger yet still high-affinity value for K_d,eff, 134 nM (Table 3). The increase in k₃ is consistent with binding being destabilized by configurational limitations imposed by the AMT-IDA ligand. From the low effective dissociation constant, and more particularly from the low value for the rate of dissociation from C2 (k₃ = 1.5 × 10⁻⁵ s⁻¹), we propose that the copper in this complex simultaneously forms a bound coordination complex with the channel.

Table 3.

The Heats of Complexation ΔH and Binding Constants K_d for Cu(AMT-IDA) and Cu(CO-IDA) at Various pH Values

ITC Results
Copper Complex	pH	ΔH (kcal/mol)	ΔH SD	Kd	Kd SD
1 mM Cu(Amt-IDA)	8	−5.8	0.3	0.87 × 10−4	3.7 × 10−5
	7	−6.5	1.1	2.4 × 10−4	8.3 × 10−5
	6	−5.5	1.1	1.5 × 10−4	3.5 × 10−5
1 mM Cu(CO-IDA)	8	−8.4	0.4	1.8 × 10−4	2.4 × 10−5
	7	−6.6	0.4	1.1 × 10−4	1.2 × 10−5
	6	−4.8	0.5	2.8 × 10−4	1.9 × 10−5

The measurements for each complex are averages from analyses of two to six experiments.

Cu(CO-IDA) also had block kinetics similar to Cu²⁺(aq) and Cu(AMT-IDA), with nearly complete block at 0.50 and 1.0 mM by 25 min (Fig. 2 e) and no significant wash-out after 5 min (Fig. 2 f). It should be noted that the upper traces in Fig. 2, c and e show more rapid block than the representative traces previously published (12) for the same concentrations of these compounds, probably because the concentrations in the previous work were lower than thought because of insoluble material associated with the previous complexes. Like Cu(AMT-IDA), unbinding from C2 (k₃) is higher than for Cu²⁺(aq), as expected from ligand-induced destabilization. Also, entry (k1) is remarkably higher than for Cu(II) or Cu(AMT-IDA), suggesting that the flexible ligand enhances entry into the channel. But these variations are compensated by other changes, such that K_d,eff is 101 nM, intermediate between the values for Cu(AMT-IDA) and Cu²⁺(aq). Again, the value suggests formation of a coordination complex between the CO-IDA complexed Cu(II) and the channel.

The outer dissociation constant (K_d1) for Cu(AMT-IDA) bound to the initial site, i.e., dissociating from state C1, is 2.86 × 10⁻⁴ M⁻¹, on the same order as the dissociation constant for AMT from M2 S31N, which can be calculated from the rate constants as 1.1 × 10⁻⁴ M⁻¹ (20). The association rate constant, k₁, is ∼5× higher for AMT than for Cu(AMT-IDA) and the dissociation rate constant, k₄, ∼2 higher, suggesting that the Cu-diacetate adduct inhibits passage through the Val27 stricture at the channel entry compared to AMT without the adduct. Similarly, cyclooctylamine, which is known to inhibit influenza A M2 somewhat better than AMT (27), gives faster entry and exit rate constants for Cu(CO-IDA) than AMT with a net lower K_d1 (Table 2), as would be expected because of the flexibility of its alkyl ring. The failure of Cu(IDA), with no hydrophobic adduct, to block the channel (Fig. S4) demonstrates that the hydrophobic moiety in the ligand is necessary. Therefore, although alternative double-exponential mechanisms such as multiple blocking sites with different degrees of block cannot be ruled out, the most parsimonious mechanism is the serial two-site blocking model (Eq. 1). Specifically, we suppose that the first binding step is entry of the copper complex into the channel near the established amantadine binding site (26), facilitated by the hydrophobic moiety associating with the hydrophobic Val27 cluster at the entryway. The second step is coordination complex formation with one of the His37 side chains, as previously suggested by the lack of strong binding in the H37A mutant (12).

Electrophysiology: S31N-G34E

The S31N-G34E variant has glutamates in the place of glycines in the channel, which could potentially create steric hindrance preventing the copper complexes from reaching the His37 cluster. Blocking and wash-out traces for M2 S31N-G34E exposed to 100 μM of Cu²⁺(aq), Cu(AMT-IDA), or Cu(CO-IDA) were obtained using TEVC (Fig. 3). Here, we give exemplary raw data traces for a qualitative evaluation of the question: do the copper complexes also block the S31N channel when the putative binding site is crowded with glutamate side chains at position 34? Cu²⁺(aq) (Fig. 3 a) shows blocking kinetics similar in the M2 S31N-G34E to those in the M2 S31N (Fig. 2, a and b), with an initial fast phase in the wash-in and modest recovery in the wash-out. Cu(CO-IDA) blocks 86% of the M2 current after 12 min of perfusion (Fig. 3 b), more than double the time to block M2 by Cu²⁺(aq). In the double mutant, Cu(CO-IDA) shows more wash-out compared to Cu²⁺(aq) (Fig. 3 a), though the appearance is exaggerated because of timescale contraction (20 min vs. 11 min). This implies that the complex does not bind as tightly as the aquated cation, although it still binds much more tightly than copper-free AMT (see below). Similarly, Cu(AMT-IDA) blocks 87% of the M2 current after 10 min of perfusion (Fig. 3 c). Again, the wash-out is modest. The block kinetics of M2 S31N-G34E by AMT contrast sharply to those of the metal complexes. Maximal block by 100 μM AMT is reached quickly, but the total block is only 30% (Fig. 3 d). This result is similar to previous results with AMT in M2 S31N (20), for which maximal block is achieved within 20 s. Interestingly, total block in M2 S31N was only ∼5%, suggesting that the Glu34 side chains in the double mutant used here retain the AMT better, perhaps through ionic interactions with the AMT amine. The wash-out is nearly complete immediately on perfusion with Barth’s pH 5.3 minus AMT. Thus, AMT is not effective in blocking this double AMT-resistance mutant, as would be expected. This may not prove that the metal complexes are immune to all possible mutations (28), but they do show that the metal complexes block M2 current even when glutamates replace the glycines in the channel, complementing the previously observed lack of resistance formation during extensive passaging (17).

Figure 3.

TEVC current traces for oocytes transfected with M2 S31N-G34E. Representative traces show whole-cell currents, with inward (negative) current beginning when perfusion solution is changed to pH 5.3 (shortly after t = 0 min). Then, block is tested by perfusion in 0.1 mM solutions of (a) Cu²⁺(aq), (b) Cu(CO-IDA), (c) Cu(AMT-IDA), and (d) AMT, which starts at 1–2.5 min at the time of onset of block of inward current. Wash-out with Barth’s pH 5.3 without blocker begins at the red arrow. Perfusion is returned to the starting solution, Barth’s pH 7.4, at the end of the experiment. To see this figure in color, go online.

ITC

To further assess the binding of histidine to the copper complexes, as a comparison for the apparent binding affinity in the channel (above) and the quantum chemical calculations (below), the heat and free energy of complexation by imidazole to an open site on the Cu(II) in the copper complexes were assessed using ITC. In this method, small volumes of imidazole solution were repeatedly injected into a temperature-controlled reservoir containing a known concentration of metal complex, and the heat required to maintain the temperature was measured for each injection. The concentrations of the reactants in the injectate and the reservoir were optimized beforehand to ensure that initial injections minimally affected the metal complex concentration and later injections were sufficient to nearly saturate the binding sites. Representative temperature plots and curve fits to the area-under-the-curve data are shown in Figs. S2–S4. When imidazole was added to solutions of Cu(AMT-IDA) or Cu(CO-IDA), the enthalpy of complexation ranged from −4.8 to −8.4 kcal/mol with a mean (over the two complexes and three pH levels tested) of −6.2 kcal/mol. The pH did not greatly affect the enthalpy of complexation, even though the concentration of protonated imidazole would change with pH. This could be due to imidazole exchanging with water on the Cu complexes, and as the imidazole is deprotonated, the water is protonated. Both copper complexes have K_d-values of the same magnitude (∼10⁻⁴ M, Table 3), suggesting strong coordination by imidazole. The enthalpy values and equilibrium constants for Cu(AMT-IDA) and Cu(CO-IDA) binding to imidazole are comparable with previous measurements of imidazole binding to aqueous Cu²⁺ with enthalpy = −7.2 kcal/mol and K_d = 4.9 × 10⁻⁵ M (29), demonstrating that the IDA ligation of the copper atom has only a small impact on imidazole coordination affinity. These values are slightly lower than the K_d2 values estimated from electrophysiology (4.69 × 10⁻⁴ and 1.0 × 10⁻³ M, Table 2). We expect that this small difference is due to the different environments in bulk solution and inside the channel. The similarity of K_d from ITC to K_d2 from electrophysiology further implies that the second binding state in the kinetic model used to fit electrophysiology data is coordination of a His37 side chain to the copper atom in the copper complexes.

Quantum chemical modeling of imidazole coordination to copper complexes

Table 4 reports the ωB97X-D3 calculated imidazole coordination energies to Cu(AMT-IDA)(OH₂)₂ and Cu(CO-IDA)(OH₂)₂ for the reactions in Fig. 4. This model represents the estimate (with inclusion of solvation free energy) of coordination enthalpy. The imidazole N replaced the water O in the equatorial plane for both complexes, with the plane of the imidazole optimally perpendicular to the equatorial plane (Fig. 4). The calculated ΔH-value (298 K, 1 atm, no concentration correction) for Cu(AMT-IDA) is −9.4 kcal/mol (Table 4), which is more exothermic than the ITC measurement of −5.8 kcal/mol (Table 3). Here, the DFT calculation for replacement of one water molecule with a neutral imidazole is simplified from the experimental conditions in that 1) the rest of the water is treated as implicit solvent parameterized against pure water (pH 7), 2) there is no buffer or salt present, and 3) the simulation does not account for the fraction of imidazole molecules that are protonated, which increases as pH is decreased. We compared to the enthalpy of binding measured at pH 8, at which the imidazole is almost exclusively neutral. Buffer and OH⁻ complexation, ionic strength, and implicit solvation errors—particularly neglect of the hydrogen bond between the deprotonated imidazole N and water—may accumulate to explain the differences observed. The calculated Cu-imidazole nitrogen bond length is 2.03 Å, which is in the range of 1.97–2.16 Å found for equatorial complexation of Cu(II) in crystal structures (30, 31, 32).

Table 4.

Copper Complexes Binding Imidazole N_δ in Implicit Water Solvent

Quantum Chemical Modeling Results
Reaction	ΔH (kcal/mol)	Cu-Imidazole Nδ Bond Length (Å)
Imidazole Nδ + Cu-(AMT-IDA) -2H2O -> Imidazole Nδ - Cu-(AMT-IDA)-H2O + H2O	−9.4	2.03
Imidazole Nδ + Cu-(CO-IDA)-2H2O -> Imidazole Nδ - Cu-(CO-IDA)-H2O + H2O	−12.5	2.03

Calculated enthalpy of reaction (difference between heat of formation of the reactants and products) using the ωB97X-D3 DFT functional in Terachem.

Figure 4.

Structural optimizations using DFT for (a) Cu(AMT-IDA) and (b) Cu(CO-IDA). Computational model reaction with unprotonated imidazole in implicit water is shown.

The calculated imidazole coordination enthalpy for Cu(CO-IDA) is −12.5 kcal/mol (Table 4), again more exothermic than the measured enthalpy value of −8.4 kcal/mol (Table 3). The calculated Cu-imidazole nitrogen bond length is 2.03 Å. These DFT coordination energies suggest that the ITC values are measuring imidazole N substitution for water O at Cu in the equatorial position, i.e. in the plane of the IDA ligand. Enthalpy calculations were done for a model with a second imidazole binding in the axial position on the copper for both complexes (Fig. S6; Table S1). The calculated enthalpy for two-imidazole coordination with Cu(AMT-IDA) was −15.8 and −18.4 kcal/mol for Cu(CO-IDA), an increase of ∼−6 kcal/mol in each case. Although this could suggest the possibility of coordination of two imidazoles, it is likely that the second imidazole coordination requires a significant entropy penalty that offsets the enthalpy gain, rendering it an endergonic second coordination. This is consistent with no second plateau in the injection heats in ITC. The results confirm that imidazole would coordinate with the copper complexes, optimally in an equatorial position.

Quantum chemical model: copper complexes binding His37 in S31N M2 and its G34E homolog

To explore the fit of the metal complexes in the central cavity (near G34) and the potential to orient amenable to complexation with a His37 imidazole N, we carried out simulations based on a representative solid state NMR structure of the transmembrane domain in a dimyristoylphosphatidylcholine environment, after relaxing it in the same lipid using molecular dynamics. Although the imidazoles are hydrogen bonded to each other at neutral pH (33, 34, 35), we assumed that it would be feasible for them to detach and rotate through torsion angle changes such that the imidazole N faces the central cavity, as is found in 2KQT and proposed for the proton transport cycle at low pH (33) and observed for the drug-bound state (35). The geometry optimization of Cu(CO-IDA) in the M2 S31N variant channel (i.e., the homolog of 2KQT) shows the nitrogen is bonded in an equatorial position on the copper, while the copper remains complexed with the CO-IDA (Fig. 5, a and b). The bond length between the copper and imidazole nitrogen is 1.99 Å. The Cu-N_Imid bond directionality is in the copper’s remaining equatorial position because of the equatorial position providing a stronger bond than an axial position. Indeed, it appears that tridentate complexation of Cu(II) is optimal for this situation (see “Summary of related projects” in the Supporting Materials and Methods). The binding of two imidazoles to the copper (one equatorial, the second axial) was attempted in these simulations. However, no proper starting geometries were achievable without extreme rotation of the copper complex and distortion of the M2 backbone. For more than one imidazole to bind the copper, it appears that the copper would need to dissociate from the ligand. The geometry optimization structure for Cu(AMT-IDA) in the M2 S31N variant is shown in Fig. 5, c and d. The optimized structure result is similar to Cu(CO-IDA). The adamantyl group is not as flexible as the cyclooctyl, but it still fits down near the His37 cluster, allowing the copper to bind to an imidazole nitrogen. The imidazole nitrogen binds in an open equatorial position of the copper, while the copper remains complexed with the AMT-IDA. The Cu-N_Imid bond length is 2.07 Å. The geometry optimization structure for Cu(CO-IDA) in the M2 S31N-G34E variant shows the glutamates at position 34 do not prevent Cu(CO-IDA) from reaching a position to bind with a His37 imidazole (Fig. 6, a and b). This agrees with the electrophysiology results that Cu(CO-IDA) still blocks M2 S31N-G34E current. The geometry optimization of Cu(AMT-IDA) in the M2 S31N-G34E variant also showed that the complex fits near the His37 cluster and the copper bonds to an imidazole nitrogen (Fig. 6, c and d). These results with the transmembrane domain are expected to be representative of the structure in the full-length channel, but variations in the structure of this binding site should be expected with peptide elongation (33,36) and lipid bilayer structure (37).

Figure 5.

Geometry optimization of Cu(CO-IDA) (a and b) and Cu(AMT-IDA) (c and d) in M2 S31N. The calculation was done in gas phase. The ribbons represent the backbone of M2. Cyan atoms are carbons, white atoms are hydrogens, red atoms are oxygens, blue atoms are nitrogens, and the orange atom is copper. (a and c) Face-on view. The front and back ribbons are not displayed for easier visualization of the imidazole binding with the copper complex, but the histidine is still connected to the backbone, as shown in the side view. (b and d) Side view. Magenta is used to highlight imidazole carbons and nitrogens in (d).

Figure 6.

Geometry optimization of Cu(CO-IDA) (a and b) and Cu(AMT-IDA) (c and d) in M2 S31N-G34E. The calculation was done in the gas phase. The ribbons represent the backbone of M2. Cyan atoms are carbons, white atoms are hydrogens, red atoms are oxygens, blue atoms are nitrogens, and the orange atom is copper. (a and c) Face-on view. The front and back ribbons are not displayed for easier visualization of the imidazole binding with the copper complex, but the histidine is still connected to the backbone, as shown in the side view. (b and d) Side view. Magenta is used to highlight imidazole carbons and nitrogens in (d).

In summary, the ITC and computational chemistry results provide insight into the ability of the copper complexes to fit in the M2 channel near the His37 cluster and project the copper towards the imidazole nitrogen. The DFT calculations indicate that imidazole would optimally bind with its plane perpendicular to the Cu(II) equatorial plane, but that in the channel, steric limitations prevent this binding configuration, explaining why the K_d2 measured with electrophysiology is higher than K_d measured with ITC. Nevertheless, coordination of the complex by a His37 imidazole in the central cavity of the channel is shown with DFT calculations of the complex in the channel to be structurally feasible, and the energetics of binding in the channel are reasonably consistent with those measured with ITC. This provides an explanation for the two-stage M2 current block in the electrophysiology results and the invulnerability of the complexes to resistance formation (17). Considering that IDA ligation of the copper substantially reduces its toxicity in zebrafish embryos (17), it would be interesting to examine toxicity and absorption, distribution, metabolism, and excretion properties of the complexes in higher animal models.

Conclusions

Cu(CO-IDA) and Cu(AMT-IDA) bind to the M2 channel in S31N and S31N-G34E. A two-site mathematical model used in a global fit with the electrophysiology traces varying in copper complex concentration provided rate constants for Cu(CO-IDA) and Cu(AMT-IDA) binding in M2 S31N. The copper complexes get into the His37 binding site at nearly the same rate as Cu²⁺(aq). The Cu²⁺(aq) and copper complexes are slow to leave the His37 binding site as shown by the dissociation rate constants being much smaller than the association rate constants, which implies strong binding to the His37 site. The copper complexes reside longer in the M2 channel than Cu²⁺(aq). The independent fit from ITC shows that the copper complexes are capable of binding one imidazole strongly, which agrees with the quantum chemical model. The quantum chemical model of the copper complexes in the channel show that the complexed copper binds to one imidazole in both the S31N and S31N-G34E M2 variants.

Editor: Vasanthi Jayaraman.