Structure Changes of a Membrane Polypeptide under an Applied Voltage Observed with Surface-Enhanced 2D IR Spectroscopy

Abstract

The structures of many membrane-bound proteins and polypeptides depend on the membrane potential. However, spectroscopically studying their structures under an applied field is challenging, because a potential is difficult to generate across more than a few bilayers. We study the voltage-dependent structures of the membrane-bound polypeptide, alamethicin, using a spectroelectrochemical cell coated with a rough, gold film to create surface plasmons. The plasmons sufficiently enhance the 2D IR signal to measure a single bilayer. The film is also thick enough to conduct current and thereby apply a potential. The 2D IR spectra resolve features from both 3₁₀- and α-helical structures and cross-peaks connecting the two. We observe changes in the peak intensity, not their frequencies, upon applying a voltage. A similar change occurs with pH, which is known to alter the angle of alamethicin relative to the surface normal. The spectra are modeled using a vibrational exciton Hamiltonian, and the voltage-dependent spectra are consistent with a change in angle of the 3₁₀- and α-helices in the membrane from 55 to 44°and from 31 to 60°, respectively. The 3₁₀- and α-helices are coupled by approximately 10 cm⁻¹. These experiments provide new structural information about alamethicin under a potential difference and demonstrate a technique that might be applied to voltage-gated membrane proteins and compared to molecular dynamics structures.

Graphical Abstract

Studying the structures of membrane proteins under an applied voltage is experimentally difficult. The crux of the problem is one of sensitivity. Most spectroscopies and scattering experiments require thick samples composed of many layers of proteins to produce enough signal. When a field is applied across a thick sample, the field strength is strongest at the surface and rapidly diminishes in the interior. As a result, extremely high voltages are required to elicit a structural response for those proteins near the interior of a thick sample. Indeed, there is only one example of an X-ray scattering experiment performed on a protein crystal under an applied voltage.¹ Techniques with high sensitivity that do not need thick samples but can be applied to monolayers, single bilayers, or single molecules can be used to study voltage-dependent proteins. For example, fluorescence spectroscopy has been used in conjunction with a patch clamp to study membrane proteins under an applied voltage,^2,3 and surface-enhanced infrared absorption spectroscopy (SEIRA) has been used to measure the IR spectra of a protein monolayer with an applied voltage.^4,5

Using mid-IR plasmonics, analogous to SEIRA, the sensitivity of 2D IR spectroscopy has recently been improved to the point where a single monolayer of molecules can be measured.^6–13 Rezus and co-workers in 2015 used gold nanoantennas to create plasmons to measure thin organic films with 2D IR spectroscopy.¹⁰ Hamm and co-workers used rough metallic thin films to create localized surface plasmons to enhance the electric field strengths of the laser pulses sufficiently for measuring 2D IR spectra of carbon monoxide adsorbed on a surface.¹⁴ They also applied a voltage to the thin metal film and observed and quantified the Stark shift of the immobilized carbon monoxide.⁷ In a similar approach, Rubtsov and co-workers have used nanoantennas to study monolayers of ester containing molecules with 2D IR spectroscopy.⁸ In this report, we use thin metallic films to both measure 2D IR signals of alamethicin bound to a single membrane bilayer while also utilizing the film to apply a voltage. Using 6 μm light in the mid-IR allows the amide I mode, which is the vibrational mode created by the carbonyl stretches of the backbone and is well-known to be sensitive to polypeptide secondary structure, to be measured.¹⁵

Alamethicin is a 20 residue antimicrobial peptaibol from the fungus Trichoderma viridae.¹⁶ The monomer itself is comprised of an α-helical region and 3₁₀-helical segments, on the N- and C-termini, respectively.^17–19 The α-helix spans about 13 residues, while the 3₁₀-helix is 4 to 5 residues. The two helical segments are separated by a kink formed from the proline at position 14. The structure of alamethicin has been characterized many times, both in organic solutions and in membranes by experimental techniques as well as molecular dynamics simulations.^19–23 The aforementioned 3₁₀-helix was identified by solid-state NMR, sum-frequency generation, and a variety of molecular dynamics simulations.^18,24–26 Using molecular dynamics simulations and sum-frequency generation, it has also been established that a potential difference or a basic pH causes the alamethicin monomer to form bundles that create pores, though the number of monomers that contribute to the pore is yet to be determined.^19,27–30 Most evidence points to pores of 4–8 monomers, though pores of up to 12 monomers have been observed in simulations.^17,23,31,32 Another reason alamethicin was chosen in this study is because of the previous work using surface-enhanced infrared absorption spectroscopy (SEIRA) and polarization modulation–infrared reflection–adsorption spectroscopy to study the tilt angle of alamethicin.^33,34 Knowing that similar structural changes occur upon a potential difference as basic pH provides a good structural control to develop our technique.

Our approach is schematically shown in Figure 1 and can be found in detail in the Supporting Information. Briefly, a sample cell is constructed with two CaF₂ windows. For the first window, gold is thermally evaporated onto a CaF₂ surface to form a rough surface that is plasmonically active. For the pH studies, the gold thickness is 3 nm. For the voltage studies, the thickness is 10 nm so that the gold is conductive and can thereby act as the working electrode. For the voltage experiment, the counter electrode is made of conductive aluminum tape on the second CaF₂ window. The bilayer is created by soaking the gold coated window with 1,2-dipalmitoyl-sn-glycero-3-phosphothioethanol lipids, which have a headgroup modified with a thiol group. The sulfur bonds to the gold, making a sparse lipid monolayer. The bilayer is created by adding vesicles of POPC (1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine) lipids that fuse to the monolayer. The substrate is then washed to make sure that only the tethered bilayer remains and the alamethicin is deposited in potassium buffer solution on the POPC side of the bilayer at a peptide to lipid ratio of 1:20. 2D IR spectra are collected using a pump–probe beam geometry with the two colinear pump pulses separated by t₁ and a probe pulse delayed by t₂ (see Figure 1). In the two experiments, the same sample was modulated either by changing the pH of the buffer between pH 6.8 and pH 12 or changing the potential across the sample from 0 V applied to −900 mV relative to the gold electrode. Since our setup has no reference electrode, the absolute potential is not defined. Nonetheless, the structure of the alamethicin will be determined by the potential difference across the membrane.

Figure 1.

Schematic of the voltage-dependent experiment. Two colinear pump pulses and a probe pulse are directed into the sample cell. The working electrode consists of a rough, 10 nm thick layer of gold, while the counter electrode is made of conductive aluminum tape. The circuit is connected to a potentiostat to control the potential difference.

Figure 2A shows the spectrum of membrane-bound alamethicin at pH 6.8. There are three principal peaks along the diagonal at ~1630, 1640, and 1660 cm⁻¹ as measured by their frequency along the pump axis (labeled i, ii, and iii, respectively). The most intense peak is at 1630 cm⁻¹. They have relative intensities of 1:0.5:0.2, for the 1630, 1640, and 1660 cm⁻¹ features, respectively. Figure 2B shows the spectrum of alamethicin at pH 12. The same three peak frequencies are observed, albeit with different relative intensities, which are 1:1.1:1.7, respectively. To verify that the peaks shown are indeed from the peptide and not from the membrane bilayer, spectra were also collected of the membrane itself (Figure S1). The membrane-only spectra contained features near 1750 cm⁻¹ from the ester carbonyl stretches of the lipid headgroups, but no peaks in the amide I region were observed. Thus, the spectra in Figure 2 are solely created by alamethicin. To further verify that the signal arises from the lipid-bound peptide, we also considered that the solvation free energy of alamethicin to POPC is about −115 kcal/mol (compared to −20 kcal/mol in water), and so, most of the polypeptide is bound to the bilayer.³⁵ Nonetheless, free-floating alamethicin would not contribute to the signal, because the plasmonic enhancement of the signals only occurs within 50 nm of the gold surface.³⁶ Cross-peaks also appear that correlate the 1660 cm⁻¹ peak with the 1630 and 1640 cm⁻¹ peaks. We label them X1–X3 in Figure 2B, but they appear in all three spectra of Figure 2.

Figure 2.

Spectra of the pH-dependence of alamethicin. Alamethicin at (A) pH 6.8 and at (B) pH 12 in buffer. (C) Difference spectrum generated by subtraction of the pH 6.8 from the pH 12 spectrum. The spectra were normalized to their maxima, and the subtraction was weighted 1:2, respectively. Cross-peaks appear in all of the spectra and are labeled X1–X3 in (B).

Based on prior work, peaks i and ii are assigned to the 3₁₀-helical segment of the peptide, which has two peaks due to the splitting of the A and the E mode of the helix at approximately 1640 and 1630 cm⁻¹ respectively. The A and E modes are the modes along the axis of the helix and perpendicular to the axis of the helix, respectively. We use the terms A and E modes qualitatively. A and E modes are only rigorously correct in a perfectly ordered and infinitely long helix. In a finite helix and one with structural and environmental disorder, additional eigenstates gain oscillator strength. The frequency difference between the A and E modes depends on the number of residues per turn in the helix.¹⁵ There are 3.6 residues per turn in an α-helix but only 3.2 residues per turn in a 3₁₀-helix. In α-helices, the A and E modes are separated by less than 5 cm⁻¹ and so are not usually resolved from one another, but in 3₁₀-helices, the peaks are separated by about 10 cm⁻¹.³⁷ Peak iii at 1660 cm⁻¹ is assigned to the α-helical section of alamethicin. This value is higher than an α-helix in buffer but typical for membrane associated helices that have better defined structures and are protected from solvent, both of which result in a higher frequency.³⁸ Based on that assignment, cross-peaks X1 and X3 correlate the A mode of the α-helix with the A mode of the 3₁₀-helix, above and below the diagonal, while cross-peak X2 correlates the α-helix and the E mode of the 3₁₀-helix. The A mode of the α-helix is largely responsible for this peak due to the fact that the E modes are much weaker than the A modes in the α-helix, because the local mode transition dipoles lie nearly parallel to the helix axis. E modes for α-helices are rarely observed in 2D IR experiments, so we assume that the A mode contributes mostly to this cross-peak. Supporting that conclusion, our calculations predict that that the E mode intensity in our modeled spectra is 8.5% of the A mode in the α-helix. Cross-peaks in Figure 2B appear either entirely positive (blue) or negative (red). However, they are theoretically a pair of positive and negative peaks, separated by the off-diagonal anharmonicity.¹⁵ Due to overlaps in the experimental spectra, they are partially obscured. This anharmonicity leads to positively signed cross-peaks that do not appear directly above or below the transition as in the case of X2, as illustrated in the calculated spectrum in Figure 6.

Figure 6.

Calculated 2D IR spectrum of alamethicin including 10 cm⁻¹ coupling between the α- and 3₁₀-helices. Cross-peaks are labeled X1–X4 as in Figure 2B. Peak centers are labeled as i–iii and are consistent with the peak frequencies in experiment.

The 2D IR spectra were also simulated using COSMOSS, a program for modeling infrared spectra using vibrational excitons that is available on our Web site.³⁹ The transition dipole directions of the amide I modes were generated using the PDB file of alamethicin (1AMT).¹⁶ Published values for the 3₁₀- and α-helices coupling constants were used. The nearest neighbor couplings were held constant regardless of the precise dihedral angles from the PDB file. The coupling values for the 3₁₀-helix were 0.9 cm⁻¹ for nearest neighbor β_i,i+1, −2.5 cm⁻¹ for β_i,i+2, −2.8 cm⁻¹ for β_i,i+3, −0.8 cm⁻¹ for β_i,i+4, and neglected for more distant residues. The coupling values for the α-helix were 8 cm⁻¹ for β_i,i+1, −2 cm⁻¹ for β_i,i+2, −6 cm⁻¹ for β_i,i+3, −1 cm⁻¹ for β_i,i+4, −0.5 cm⁻¹ for β_i,i+5, and neglected for further residues.^15,40 Local mode frequencies are set at 1635 cm⁻¹ for the 3₁₀-helix and 1655 cm⁻¹ for α-helix portions based on previously published experimental data.^27,38,41 Each transition in the 2D IR stick spectrum is scaled with respect to its orientation in the bilayer where the electric fields of the laser are normal to the bilayer, followed by convolution with a 10 cm⁻¹ fwhm 2D Lorentzian. The spectra were calculated separately for the α- and 3₁₀-helices, and the couplings between the helices were neglected (see Figure 4 below for simulations that include couplings) so that the tilt of the α- and 3₁₀-helices in the membrane could be independently adjusted and their respective 2D IR spectra summed to best match the experiment.

Figure 4.

Spectra of the voltage-dependence of alamethicin. (A) Alamethicin in a bilayer on a 10 nm Au surface with 0 V applied. (B) Alamethicin in a bilayer on a 10 nm Au surface with −900 mV applied. (C) Difference spectrum of alamethicin at −900 and 0 mV. Real peaks corresponding to pH spectral peaks indicated with dashed lines i–iii, corresponding to those peaks in Figure 2.

Diagonal slices of simulated of 2D IR spectra are shown in Figure 3A for comparison to Figure 2. The calculated 3₁₀-helix peaks at 1627 and 1635 cm⁻¹ correspond with peaks i and ii in Figure 2. The calculated α-helical peak at 1655 cm⁻¹ corresponds to peak iii. To match the peak intensities in the pH 6.8 spectrum (Figure 2A), the 3₁₀-helix was set to 55° and the α-helix was set to 33° (relative to the membrane normal). Since the A and E modes are perpendicular to each other, their peak intensities depend inversely with one another as a function of the tilt angle of the peptide relative to the plane of the membrane. Thus, these two peaks give the tilt angle of the 3₁₀-helix. With those fixed, the relative ratio to the peak at 1655 cm⁻¹ gives the tilt of the α-helix. The error in these angles is tied to approximations in the model, such as the lack of disorder, but by systematically varying the model parameters, we estimate the accuracy of these angles within our approximations to be ±10°. Amide I frequencies will also depend on helix length.¹⁵ Since the frequencies are unchanged, the lengths of the helices most likely do not appreciably differ when inserted. We also note that the convoluted peaks represent the traditional A and E modes of helices, but the stick spectra contain additional transitions. A and E modes are based on symmetry in the approximation of infinitely long helices. The vibrational exciton simulations presented here take into account the finite nature of the helices, which puts oscillator strength into eigenstate modes that are IR-forbidden in the symmetry approximation and, thus, are better for simulating the experimental peak intensities and extracting helix angles.

Figure 3.

Calculated 2D IR spectra by summing simulations for the α- and 3₁₀-helices. (A) Diagonal slices of the 2D IR spectra with stick spectra (blue) for the α-helix and (red) for the 3₁₀-helix. (B) Modeled 2D IR spectra with angles of 31 and 55° for the α- and 3₁₀-helices that reproduce the relative diagonal peak intensities in the neutral pH experiment (Figure 2A). (C) Modeled 2D IR spectra with angles of 60 and 44° for the α- and 3₁₀-helices that reproduce the relative diagonal peak intensities in the basic pH experiment (Figure 2B).

Shown in Figure 4 are spectra of alamethicin on 10 nm gold with (A) no potential, (B) a −900 mV potential difference, and (C) a difference spectrum of (A) subtracted from (B). In (A) and (B), a background signal is observed along the diagonal that was not present in the samples prepared for the pH studies. “Potential difference” refers to the voltage set on the power meter. In addition to changing the conductivity, the gold thickness also shifts the resonant plasmon frequency to longer wavelengths (Figure S3), which we attribute to the presence of the background. The background signal partially obscures the expected molecular features but is reduced by approximately 90% in the difference spectrum of Figure 4C as determined by the subtraction of the control sample seen in Figure S2. Though the distance between the electrodes is 50 μM, the double layer thickness should be sufficient to create a potential across the bilayer, because the bilayer is tethered to the gold. Previous Stark shift and SEIRA experiments have been able to generate a potential in this manner.^33,42–44 In Figure 4C, three diagonal peaks are observed at 1630, 1640, and 1660 cm⁻¹ that correspond to the same peaks i, ii, and iii in the pH spectra of Figure 2. The relative intensities in the difference spectrum are 0.5:0.5:1, for i:ii:iii, respectively.

As outlined above, previous work has found that the alamethicin membrane-bound structure is similar at pH 12, as it is under a potential difference.²⁶ Thus, we subtracted the pH 6.8 (Figure 2A) from the pH 12 spectrum (Figure 2B) to create a difference spectrum (Figure 2C) for comparison to the difference voltage spectrum (Figure 4C). Relative intensities of the pH spectra were adjusted for best agreement to Figure 4C. The pH difference spectrum contains the same three peaks as the voltage difference spectrum with similar relative intensities. We note that in both the pH and the voltage subtraction spectra (Figures 2C and 4C, respectively, the overtone intensities are diminished due to destructive cancellation upon subtraction further validating the subtraction. Thus, we conclude that the structure adopted by alamethicin under an potential difference is similar to that at pH 12, in agreement with prior studies.^26,27

Shown in Figure 5 is an illustration of the change in structure that is consistent with the data presented here. We see no evidence in the spectra that there is a significant change in secondary structure, because the peak frequencies and line widths are unchanged when the potential is modulated (and at pH 12). A reduction in the α-helix length, for example, would shift its frequency higher.¹⁵ Instead, the data is consistent with a change in the tilt of the helices that alter the peak intensities, as the simulations in Figure 3 show.^23,45,46 We estimate that the tilt of the 3₁₀-helix decreases by 10°, and the α-helix tilt increases by 29°.

Figure 5.

Graphical depiction of alamethicin corresponding to the tilt angles at (A) 0 V and (B) −900 mV applied. Red represents the 3₁₀-helix segment, and blue represents the α-helical segment. Angles are defined in relation to the plane of the membrane.

The relative angles of 16 and 24° between the α- and 3₁₀-helices before and during insertion is consistent with previous molecular dynamics simulations,^45,46 which found the kink angle to be 24 ± 8°.⁴⁵ Indeed, those simulations found little change in the kink angle before and after insertion, also in agreement with our work. However, the simulations put the insertion angle, which is defined as the angle between the membrane plane and the line joining the centers of the first and last seven α-carbons of a helix,⁴⁶ for alamethicin at 70 to 90°, which is much larger than the steepest angle (60°) in our experiments. Those simulations may not have converged to the equilibrated molecular structure, because the angle of alamethicin in the membrane was still decreasing at the end of the 100 ns trajectories. The insertion angles do not appear consistent with the barrel stave model proposed previously for alamethicin, although these experiments do not explicitly test the bundle structure. To investigate the structure of the bundle, isotope labeling experiments looking at residue-specific couplings and 2D line shapes would be necessary.^41,47,48

We also simulated spectra that included couplings between the 3₁₀- and α-helices in order to explore the origins of the cross-peaks. To do so, 10 cm⁻¹ couplings were added between the top four residues of the 3₁₀-helix and the bottom four residues of the α-helix; 10 cm⁻¹ is consistent with couplings between helix turns in helices. The Hamiltonian is found in the SI. The resulting spectrum (Figure 6) contain cross-peaks between the A mode of the α-helix and the A and E modes of the 3₁₀-helix, generating the experimental peaks X1–X3. The simulations also resolve a peak that we label X4, which correlates the E mode of the 3₁₀-helix to the A mode of the α-helix, analogous to peak X2 on the opposite side of the diagonal. Presumably, the structure of the polypeptide as it transitions from the 3₁₀- to the α-helix is complicated, with a kink and structural flexibility,⁴⁶ which would give rise to a distribution of couplings as well as disorder in the Hamiltonian. Thus, these simulations confirm our spectral assignment based on the frequencies of the cross-peaks but should be considered illustrative. 2D IR spectra can now be calculated from molecular dynamics simulations quite accurately and used to test atomic structures.^15,49,50 It is now also established that vibrational energy transfers over large distances, which might also contribute to the cross-peak intensities. Future studies might simulate 2D IR spectra from voltage-applied molecular dynamics simulations.

The experiments reported here demonstrate 2D IR spectroscopy applied to a membrane-bound polypeptide under a potential difference. They build upon previous work using SEIRA to study the voltage-dependence of membrane proteins^5,33,51–53 while providing additional information through couplings, 2D line shapes, and other observables intrinsic to multidimensional spectroscopy. The signal enhancement that results from mid-IR plasmonics makes it possible to measure the 2D IR spectra of a single bilayer, which in turn enables the application of an external voltage. It should be possible to generate fields with other plasmonic surfaces, such as islands, that might reduce the background.⁵⁴ Further work is also needed to characterize the magnitude of the potential difference; −900 mV is most certainly not the potential at the bilayer, because the bilayers we created have imperfections that conduct and thereby decrease the local potential. Nonetheless, for alamethicin, we find that the potential difference causes the polypeptide tilt to increase in the bilayer but does not measurably change the relative populations or lengths of the 3₁₀- and α-helices. We suspect that the potential difference induces structural changes that occur at the transition between the 3₁₀- and α-helix that are not detected in our experiments, but they do create cross-peaks in the spectra, as confirmed by simulations. More detailed structural information might be obtained with isotope labeled peptides to spectroscopically identify those residues.⁵⁵ As mentioned above, it is now relatively straightforward to accurately compute 2D IR spectra from molecular dynamics trajectories and thereby test atomic structures against experiment. That approach has been applied to membrane proteins.^{49,50,56–58} Molecular dynamics simulations can also be used to study structures of membrane proteins under an applied potential. The methodology and 2D IR experiments detailed in this publication are early steps toward structural and dynamical studies of voltage-gated membrane proteins.