Imaging Cu^2$+$ binding to charged phospholipid membranes by high-throughput second harmonic wide-field microscopy

Divalent cations are known to interact with cell membranes in numerous physiological functions, such as signaling through ion-specific channels in synapses,^1–3 membrane fusion,^4–7 and the induction of cell death.^8,9 Because of the importance of ion–membrane interactions, they have been studied extensively using fluorescent methods.^10–15 While these methods can generate important insights, the fluorophore itself, or the way in which the measurement is conducted/interpreted, might obscure or influence the molecular level mechanisms that are at play. For example, it was shown only recently that divalent cations such as Ca²⁺, Mg²⁺, and Ba²⁺ do not homogeneously bind to freely suspended charged lipid membranes, but instead form transient domains of ion–lipid complexes.¹⁶ Cu²⁺ is another divalent ion that plays an important role in numerous physiological functions of cells. For instance, the Cu²⁺ dependent enzyme, cytochrome c oxidase, is used in the synthesis of adenosine triphosphate (ATP) by translocating protons and generating electrochemical potential across the cell membrane,^17,18 and tyrosinase, a Cu²⁺-containing transmembrane protein, controls the production of melanin in the skin.^19,20 It is, therefore, important to study Cu²⁺–membrane interactions by taking into account the spatial distribution of ion binding across the membrane surface and membrane hydration. Second-order nonlinear optical techniques are an ideal tool for probing membrane interfaces, since the intrinsic symmetry of the light–matter interaction ensures that only non-isotropic structures are measured. The recent invention of a wide-field second harmonic (SH) microscope demonstrated an increase in the throughput by a factor of 5000 over conventional multiphoton scanning confocal imaging systems.²¹ With this microscope, it became possible to dynamically SH image the non-resonant interfacial response of water. Using water as a probe for the interfacial structure combined with dynamic spatial imaging on sub-second time scales allows for a label-free investigation of membrane processes.²² Since hydration is involved directly in membrane–ion interactions, it serves as an ideal probe to measure the interaction of divalent ions with lipid membranes. Recently, spatiotemporal membrane hydration changes revealed that the interaction of divalent Ca²⁺, Ba²⁺, and Mg²⁺ with the membrane is a dynamic process, whereby short-lived and dynamic domains of ion–lipid complexes are formed at the aqueous membrane interface.¹⁶ 

Here, we use a similar approach to investigate the interaction of Cu²⁺ with freestanding lipid membranes composed of lipids having phosphatidic acid (PA) or phosphatidylserine (PS) head groups. We find that Cu²⁺–membrane interactions also occur via heterogeneously distributed transient domains of ordered interfacial water molecules. Analysis of these domains shows that the electrostatic binding constant is higher for PA lipids than for PS lipids. The same applies for the extracted electrostatic free energy difference and the local/dynamic binding constants.

1,2-diphytanoyl-sn-glycero-3-phosphocholine (DPhPC), 1,2-diphytanoyl-sn-glycero-3-phosphate (DPhPA), and 1,2-diphytanoyl-sn-glycero-3-phospho L-serine (DPhPS) in powder form (>99%) (Avanti Polar Lipids, AL, USA) and hexadecane (C₁₆H₃₄, 99.8%, Sigma-Aldrich), hexane (C₆H₁₄, >99%, Sigma-Aldrich), chloroform (>99.8%, Merck), hydrogen peroxide (30%, Reactolab SA), sulfuric acid (95%–97%, ISO, Merck), KCl (99.999%, Aros), and CuCl₂ (99.999%, Sigma-Aldrich) are used as received. All aqueous solutions are made with ultrapure water (H₂O, Milli-Q UF plus, Millipore, Inc., electrical resistance of 18.2 MΩ cm). All aqueous solutions are filtered with 0.1 µm Millex filters. The coverslips used in the imaging are pre-cleaned with piranha solution (1:3–30% H₂O₂: 95%–97% H₂SO₄) and thoroughly rinsed with ultrapure water. For lipid membrane formation, a chloroform solution containing the desired lipids is used.

Freestanding horizontal planar lipid bilayers are formed following the procedure of Montal and Müller.²³ Two separate lipid monolayers on an air/water interface are combined in a ∼100-µm aperture in a 25-µm thick Teflon film. The presence of a bilayer is confirmed with white-light imaging and electrical recordings with specific capacitance, C_m > 0.7 μF cm⁻², and specific resistance, R_m ∼ 10⁸ Ω cm². The composition of the leaflets and the aqueous solution, where the bilayer leaflets reside, are controllable in situ. Unless stated, all measurements are performed under pH-neutral conditions.

The imaging setup has been characterized in detail in Refs. 16, 21, 22, and 24 based on the principles of second harmonic scattering. Two crossed beams from a Yb:KGW femtosecond laser (Light Conversion Ltd.) delivering 190 fs pulses, 1030 nm with a 200 kHz repetition rate, are incident at 45° with respect to the membrane. Each beam is loosely focused using an f = 20 cm doublet lens (B coating, Thorlabs), and polarization is controlled using a linear polarizer (Glan–Taylor polarizer, GT10-B, Thorlabs) and zero-order λ/2 wave plates (WPH05M-1030, Thorlabs). The average power for each arm is set to ∼260 mW. The phase-matched SH photons are collected with a 50× objective lens (Mitutoyo Plan Apo NIR HR Infinity-Corrected Objective, 0.65 NA) in combination with a tube lens (Mitutoyo MT-L), a 900 nm short pass filter (FES0900, Thorlabs), a 515 nm bandpass filter (FL514.5-10), and an intensified electronically amplified CCD camera (EM-ICCD, Pi-Max 4, Princeton Instruments). A 400 mm meniscus lens is placed behind the objective lens to remove spherical aberrations induced by the cover slip. The transverse resolution and, thus, the pixel width are 430 nm. All images are recorded with the beams polarized parallel to the plane of incidence (P). The acquisition time of individual frames is 1 s.

For a lipid membrane with two interfaces (i = 1 or 2), the total emitted SH intensity I(2ω, x, y) is related to the surface potential (Φ₀) and can be expressed as^16,22,25

where ω is the frequency of the fundamental beams, I₁ and I₂, x and y are the spatial surface coordinates, $χs,i(2)$ (i = 1 or 2) are the second-order surface susceptibilities, Φ_0,i (i = 1 or 2) are the surface potentials of each leaflet of the membrane, $χs(3)′$ is the effective third-order surface susceptibility of the water, and f₃ is an interference term, where f₃ = 1 for the case of a transmission experiment. For the case of symmetric membranes, such as those presented in this work, $χs,1(2)$ is equal to $χs,2(2)$⁠. Due to the fact that $χs,i(2)$ does not change significantly upon the addition of ions²⁶ and $χs(3)′$ is two orders of magnitude larger than $χs,i(2)$⁠,²² the SH signal observed in our images can be attributed to the difference in membrane surface potential ΔΦ₀ = Φ_0,1(x, y) − Φ_0,2(x, y), and Eq. (1) can, therefore, be reduced to

Based on Eq. (2), SH intensity is converted to a surface potential difference ΔΦ₀ by recording SH images as a function of external electric voltage across the membrane (details can be found in supplementary material, Sec. S1). From ΔΦ₀, the difference in surface charge density (Δσ₀) is calculated using a parallel plate capacitor model in contact with the ionic solutions, which is found as Δσ₀ = CΔΦ₀, with C = ε₀ε/d, where ε is the dielectric constant of the hydrophobic core (ε = 2.1) and d is the thickness of the membrane (d = 4 nm).²⁷ The electrostatic free energy (ΔG) induced by Cu²⁺ binding is given by ΔG = 2eΔΦ₀, and the dissociation constant (K_D) between ion and lipid membranes is found as K_D = e⁻Δ^G/RT, where R is the gas constant and T is the temperature.

To investigate Cu²⁺–membrane binding, the freestanding lipid membranes are formed in a ∼100-µm sized circular aperture in a 25-µm thick Teflon film by the Montal–Müller method.²³ The lipid membrane is measured with white-light imaging and electrical characterization [Fig. 1(a)]. The specific capacitance (C_m) and specific resistance (R_m) of the membrane (C_m = 0.96 μF/cm², R_m = 5 × 10⁷ Ω cm²), which agree well with reference values,^23,28 ensure the formation of freestanding lipid membranes. The lipid membranes are imaged using a medium repetition rate, wide-field home built SH microscope.^16,21,22,24 Two crossed beams from a laser source (200 kHz, 190 fs, 1030 nm) are overlapped on the membrane plane at a 45° angle with an excitation area with a diameter of ∼150 µm. Phase-matched 515 nm SH photons are emitted in the direction of the surface normal. For the case of a symmetric membrane and a symmetric aqueous environment, no coherent SH photons are detected.²² For the case of lipid hydrated membrane bilayers that are rendered asymmetric by the presence of different ions on either side of the membrane, asymmetry in the water structure leads to SH contrast. The degree to which centrosymmetry is broken and, therefore, the magnitude of the SH signal are given by the strength of interaction between the negatively charged head groups and the divalent ion. This is illustrated in Fig. 1(b). This asymmetry can be converted to surface potential values as explained in Sec. II D.

Figures 1(c) and 1(d) show the SH images of a symmetric membrane taken with a total averaged acquisition time of 20 s. The bilayer membranes are composed of 70:30 mol. % DPhPC:DPhPS and DPhPC:DPhPA, respectively, and the top leaflet is in contact with the CuCl₂ solution, while the bottom leaflet is in contact with the KCl solution, with both sides being pH-neutral and of the same ionic strength (500 µM). The SH images are corrected by the subtraction of the average value of hyper-Rayleigh scattering (HRS).

Figures 1(c) and 1(d) show that the relative SH intensity from a DPhPC:DPhPA membrane is higher than that from a DPhPC:DPhPS membrane. K⁺ ions do not bind to the membrane but provide electric field screening in the aqueous phase, which is equal in magnitude to that of the Cu²⁺ ions added to the other side of the bilayer. Therefore, the observed SH signal can be entirely attributed to electrostatic binding of Cu²⁺. When Cu²⁺ interacts with PA lipids in a bilayer membrane, it binds to their phosphate groups.²⁹ This binding neutralizes the membrane surface charge, leading to a SH response. In the case of PS, Cu²⁺ can also bind to the phosphate in a similar manner. However, unlike the binding of other divalent ions such as Ca²⁺ and Mg²⁺ to PS,¹⁶ Cu²⁺ can also form a Cu(PS)₂ complex, without altering the net negative charge on the membrane due to the deprotonation of the amine group.¹⁰ While both of these binding mechanisms take place at a bilayer interface, only the first one contributes to the change in membrane surface charge and can, therefore, be observed in the SH images.

To obtain more insight into the properties of membrane–water–ion interactions, we next quantify the spatial membrane potential, surface charge, and electrostatic free energy distribution. Figure 2(a) shows the recorded intensity values [corrected for hyper-Rayleigh scattering (HRS) by image subtraction]. Figure 2(b) shows the extracted values of the surface potential difference between the leaflets (ΔΦ₀) and the corresponding surface charge density difference (Δσ₀). The related electrostatic free energy difference (ΔG) and the electrostatic binding dissociation constant (K_D) are shown in Fig. 2(c). Instead of observing a uniformly distributed binding of Cu²⁺ ions to lipid head groups, as would be expected from mean-field theory, the images in Fig. 2 show transient structures of high intensity. By locally breaking the symmetry of interfacial water, these bright domains report on electrostatic binding of Cu²⁺ ions to the membrane. In the dark areas of the images there is no such interaction. Similar to the observations with other divalent ions,¹⁶ the chemical structures where electrostatic binding occurs are short-lived and keep changing across the entire observed region of the membrane.

To better understand the membrane–water–Cu²⁺ interactions, we next perform a single domain analysis. Figure 3(a) shows the single frame images (1 s/frame) of a symmetric membrane composed of 70:30 mol. % DPhPC:DPhPA (left) and DPhPC:DPhPS (right), with the top leaflet in contact with the CuCl₂ solution and the bottom leaflet in contact with the KCl solution. Using consecutive time frames (20 frames total), we obtain the normalized temporal autocorrelation function (TACF) to study the dynamic behavior of the domains (details can be found in supplementary material, Sec. S3). The TACF decays faster than the recording time regardless of PC:PA or PC:PS. These results show that the lifetime of each domain is shorter than the recording time, resulting in no correlation between the domains on the time scale of acquisition. We examine the ion–membrane interaction in terms of the SH intensity and size of the domain. Domains are selected based on their intensity and their individual sizes and are determined by a Gaussian fit. Details can be found in supplementary material, Sec. S2. Figure 3(b) shows the distribution of domains based on SH intensity together with their Gaussian fits. The parameters of the fits show that the SH intensity per domain is higher for the PC:PA membrane than for the PC:PS membrane, which is in agreement with the difference in average intensity observed in Fig. 2. This indicates a higher strength of electrostatic interaction between Cu²⁺ and PA than between Cu²⁺ and PS head groups. Figure 3(c) shows the percentage of the total membrane area that is occupied by the bright domains of Cu²⁺ binding in the image. The total domain area where Cu²⁺ interacts through electrostatic interactions with the lipid head groups is bigger for the DPhPC:DPhPA membrane compared to the DPhPC:DPhPS membrane. Considering that the same mol. % of PA and PS lipids were used, this indicates that PA lipids participate more in electrostatic binding than PS lipids. The average radius of the domains is analyzed using the normalized spatial autocorrelation function (SACF), and the details can be found in supplementary material, Sec. S3.

Both the size and SH intensity of each domain were taken into account to obtain the average SH intensity/domain and convert that to the average surface potential per domain. Figure 3(d) shows a higher average SH intensity/domain for DPhPC:DPhPA than for the DPhPC:DPhPS membrane. This results in a higher surface potential/domain for the PC:PA (−98 ± 3 mV) membrane compared to the PC:PS membrane (−72 ± 6 mV), as shown in Fig. 3(e). By taking into account the area per lipid for each type of membrane,^30,31 consequently, 0.70% of PA head groups are electrostatically bound to Cu²⁺, whereas for PS, the fraction is only 0.55%. These results show that Cu²⁺ ions bind to a small fraction of the lipid head groups for both systems, but that for PS head groups, electrostatic interaction is somewhat less favorable compared to PA head groups.

For the case of PA lipids, using the average SH intensity/domain, we obtain an average K_D value of 5.0 · 10⁻⁴ M. Previous experimental studies using fluorescence quenching assays have reported a K_D value for the Cu²⁺–PA interaction to be 1.0 · 10⁻⁴ M.²⁹ These values both indicate that the local K_D value for Cu²⁺–PA binding is in the high micromolar range.

In the case of PS lipids, in addition to the electrostatic binding, there is a possibility of Cu²⁺ binding through the amine without altering the interfacial potential. However, as noted before, this binding would not be observed by SH imaging. This means that the K_D values converted by the average SH intensity/domain specifically describe electrostatic binding, which alters the surface charge. Indeed, previous experiments using fluorescence quenching assays suggest that the Cu²⁺–amine–carboxyl complex formation has a K_D in the picomolar range.¹⁰ In these studies, however, the Cu²⁺–phosphate electrostatic binding would not be observed, as the fluorophores were almost fully quenched by the deprotonation binding at the relevant concentrations.

When we convert the average SH intensity observed in the domain areas to a K_D value as we did for PA, we obtain an average K_D value of 3.8 · 10⁻³ M for the local domains. The reason why this number is significantly different from the one reported by fluorescence quenching assays is that it only characterizes the binding that alters surface charge, such as the Cu²⁺–phosphate binding, and does not take into account other types of bindings such as the Cu²⁺–amine–carboxyl complex formation.

In summary, we investigate the interactions of Cu²⁺ with freestanding (DPhPC:DPhPA or DPhPC:DPhPS) lipid membranes in an aqueous solution using SH wide-field microscopy. We observe that the domains of ordered interfacial water in the direct vicinity of lipid bilayers are influenced by the Cu²⁺–membrane binding events and can be utilized to locate areas of the membrane where electrostatic binding takes place. We quantify the membrane potential, surface charge density, membrane hydration free energy, and binding dissociation constant upon Cu²⁺ binding. Analysis of these domains also reveals a stronger electrostatic binding of Cu²⁺ to PA lipids than to PS lipids. Understanding ion binding in terms of membrane hydration and local specific interaction can bring more insight into the molecular level mechanisms that are important for various biochemical processes, such as the operation of ion channels in neurons and ion-induced membrane fusion.

See the supplementary material for the conversion of SH intensity to the surface potential difference (S1), how the domains were selected for single domain analysis (S2), and spatiotemporal analysis of the ion-induced ordered water domain (S3).

The authors thank Saranya Pullanchery, Maksim Eremchev, and Orly B. Tarun for useful discussions. This work was supported by the Julia Jacobi Foundation, the Swiss National Science Foundation (Grant No. 200021-182606-1), and the European Union’s Horizon 2020 research and innovation program under Marie Skłodowska-Curie grant (Agreement No. 860592) (H2020-MSCA-ITN, PROTON).

The authors have no conflicts to disclose.

S.L. and D.R. contributed equally to this work.

The data that support the findings of this study are available from the corresponding author upon reasonable request.