Ultrasensitive Label-Free Detection of Protein-Membrane Interaction Exemplified by Toxin-Liposome Insertion.

Measuring the high-affinity binding of proteins to liposome membranes remains a challenge. Here, we show an ultrasensitive and direct detection of protein binding to liposome membranes using high throughput second harmonic scattering (SHS). Perfringolysin O (PFO), a pore-forming toxin, with a highly membrane selective insertion into cholesterol-rich membranes is used. PFO inserts only into liposomes with a cholesterol concentration >30%. Twenty mole-percent cholesterol results in neither SHS-signal deviation nor pore formation as seen by cryo-electron microscopy of PFO and liposomes. PFO inserts into cholesterol-rich membranes of large unilamellar vesicles in an aqueous solution with Kd = (1.5 ± 0.2) × 10-12 M. Our results demonstrate a promising approach to probe protein-membrane interactions below sub-picomolar concentrations in a label-free and noninvasive manner on 3D systems. More importantly, the volume of protein sample is ultrasmall (<10 μL). These findings enable the detection of low-abundance proteins and their interaction with membranes.

Lipid membranes serve as a hub for vital biochemical reactions and play an important role in chemical signaling.[1] Membrane–protein interactions are characterized by sparse, high-affinity, and highly specific binding interactions that lead to specific responses.[2] Membrane channels and pores are examples of the result of such interactions, keeping specific chemicals in- or outside of the cell or organelle to maintain the function of the organism. Pores are also used by bacteria for the opposite effect: to invade cells by effectively punching a hole in the membrane.[3] These pore-forming toxins (PFTs) have highly specific membrane interactions, often requiring specific receptors, certain membrane composition, and an aqueous environment. One example is perfringolysin O (PFO, depicted in Figure A), which is a representative of the cholesterol-dependent cytolysin family. PFO is a soluble elongated 4-domain protein that is rich in β sheets and is produced by the pathogenic Clostridium perfringens.[4] PFO has been demonstrated to selectively interact with cholesterol-rich membranes[5] where it self-assembles into a barrel-like 38-nm-wide structure of up to 50 monomers[6] (see Figure A). The monomeric structure of PFO was obtained by X-ray crystallography,[5] while its oligomer structure has been characterized by combining information about the monomeric form, cryo-electron microscopy (cryo-EM) data of pneumolysin,[7] and molecular dynamics (MD) simulations.[8]Figure B shows cryo-EM images of 120-nm-diameter-large unilamellar vesicles (LUVs) with membranes composed of sphingomyelin (SM) and cholesterol (Chol) at 50:50 mol % (1B i, iii, iv) and at 80:20 mol % (Figure B, ii). The PFO clearly displays strong interactions with the membrane in images i, iii, and iv, while no interaction is observed in ii. This is consistent with previous studies[7,9] that show fully inserted pores in the membranes that have 50:50 mol % of SM:Chol.

Figure 1

The PFO structure, the PFO-membrane interaction, and the high-throughput second harmonic scatteringsetup. (A) The structure of the PFO:PFO monomer with the undecapeptide, located on domain 4, that binds to cholesterol shown in red, PFO membrane insertion where one pore consists of 40 protomers, and a top view of such a pore. This figure is adapted from ref (3). (B) Cryo-EM images of SM:Chol 50:50 mol % (i) and 80:20 mol % (ii) LUVs measured with the same concentration of PFO in the solution (c = 9 × 10–6 M). 2D classifications of cryo-EM images of oligomerization on the SM:Chol 50:50 mol % membrane from the side (iii) and top (iv). The scale bar corresponds to 200 nm for i and ii, and to 20 nm for iii and iv, respectively. (C) Illustration of the SHS setup. All measurements were recorded with all beams polarized in the horizontal plane leading to a PPP polarization combination. For the single angle experiments, the scattering angle θ was set to 45° corresponding to the angle with maximum scattering intensity.

To understand the protein–membrane interaction beyond the crystallographic level, currently used methods are quartz crystal microbalance (QCM),[10−12] flotation assays,[13] surface plasmon resonance (SPR),[14−18] calorimetric methods,[19] and different fluorescence approaches.[20−22] Although these methods provide information about binding kinetics, they require a large amount of sample and the sensitivity is limited; namely, it does not allow the detection of low-abundance proteins whose concentration is below micromolarity. Besides the limited sensitivity, fluorimetry requires the labeling/binding of a fluorophore, which may alter the chemical and/or structural properties of the target membrane proteins, thereby perturbing the native binding affinity with the membrane. Likewise, QCM and SPR are indirect methods that require a specific substrate, which restricts the application mostly to planar-supported model membranes, and calorimetry lacks specificity to a binding event. Resonant reflection second harmonic generation (SHG) has recently been successfully employed and suggested as an alternative to immunoassays[23,24] since the SH photons emerge directly from an interface and thus facilitate direct detection of interfacial events. In these experiments, the SH photons were generated via a total internal reflection scheme from a UV resonant interaction (266 nm). Further improvements should allow for nonresonant detection, with high sensitivity, from any type of realistic membrane interface, such as that of a liposome.

Here, we show a method to detect protein liposome interaction using membrane interfacial water as a contrast agent and demonstrate sensitivity in the femtomolar range, using small volumes of protein solution. High throughput second harmonic scattering[25] (Figure C) has been shown to be a label-free method using the detection of interfacial water on buried interfaces in aqueous solution, e.g., liposomes[26,27] and oil droplets.[28] In a nonresonant SHS experiment, a pulsed femtosecond near-infrared laser beam interacts with the liposome solution. SH photons are emitted from all noncentrosymmetric molecules that are noncentrosymmetrically distributed. Since interfacial water is noncentrosymmetrically distributed while bulk liquid is not, SHS has an exquisite interfacial sensitivity. The interfacial water outnumbers lipids with a ratio of >1:100, and due to the nonresonant nature of the process both constituents respond with electromagnetic fields that have equal magnitudes. Because the SH intensity scales quadratically with the emitted electromagnetic field, SHS generally reports on membrane hydration, detecting the net orientational distribution of water molecules along the surface normal. The response of the water has been successfully used to quantify physio-chemical interfacial properties, such as the electrostatic potential (Φ0) and the degree of water ordering (quantified by the second-order susceptibility, χs,2(2)).[29,30] A recent study has also shown that changes in the membrane water can help understand the interaction of α-synuclein and the aqueous environment of liposomes.[31] Building onto these findings, we use high-throughput SHS to investigate the binding of a transmembrane protein, taking PFO as an example (Figure A as an illustration). We demonstrate sensitivity in the femtomolar range and determine a PFO-liposome binding rate of (1.5 ± 0.2) × 10–12 M.

Figure 2

PFO-vesicle binding. (A) Illustration of PFO insertion into 50:50 mol % ratio of SM:Chol LUVs surrounded by water. The pore insertion is expected to modulate the SHS intensity by removing a significant part of the membrane. Without this membrane, the number of interfacially oriented water molecules is reduced due to a lack of membrane induced anisotropy (blue color represents the aqueous medium, and the water molecules that contribute to the SHS signal are emphasized; for simplicity the molecularly complicated interaction is here shown as the formation of a hole). (B) Measured ζ-potential of the 50:50 mol % SM:Chol LUVs with 0 M and 10–9 M of PFO. (C) SHS patterns of 50:50 mol % SM:Chol LUVs with 0 M (red circles), 10–12 M (blue triangles), and 10–9 M (black squares) of PFO recorded using the PPP polarization combination. The inset shows normalized SHS patterns. The error bars were determined as a standard deviation from 20 measurements. (D) SH intensity difference (ΔS) at the angle with the maximum intensity (θmax = 45°) vs PFO concentration in logarithmic scale for two different LUV membrane compositions, SM:Chol 50:50 mol % (blue circles) and 80:20 mol % (red squares). ΔS is a difference between the coherent signal from the vesicles with a given concentration of PFO and the vesicles without PFO (with the highest number of interfacially oriented water molecules). The error bars were determined as a standard deviation from 100 measurements. The data of SM:Chol 50:50% are fitted using eq , giving the dissociation constant Kd = (1.5 ± 0.2) × 10–12 M.

To test the suitability of SHS measurements for ultrasensitive detection of transmembrane protein–membrane interactions, we employ PFO protein–lipid membrane interaction events, comparing the two membrane systems of Figure B, using a 50:50 mol % ratio of SM:Chol and an 80:20 mol % ratio of SM:Chol, whereby protein binding/insertion occurs only in the first case. Figure shows ζ-potential values (Figure B) and SHS scattering patterns (Figure C) obtained for dispersions of LUVs with a 120 nm radius and with a 50:50 mol % ratio of SM:Chol. Adding 1 pM and 1 nM of PFO to the solution does not change the ζ-potential but does change the intensity of the SHS patterns. The SH intensity arises from coherent light scattering by oriented interfacial water. This water is either oriented by chemical interfacial interactions (such as hydrogen bonding, dipole–dipole, or van der Waals interactions) or by interfacial electrostatic effects. Within the nonlinear light scattering theory that is used to describe this type of data, both interactions are captured by distinct physical parameters, namely the interfacial second-order susceptibility (χ(2)) and the surface potential (Φ0). Since the membrane interaction event of PFO is interpreted to result in the appearance of a hole, the effect on the interfacial structure of water is to reduce the number of interfacially oriented water molecules due to the lack of membrane induced anisotropy, as the interfacial area is reduced (see Figure A for an illustration). Therefore, the change in SHS response arises from a modification of the surface susceptibility rather than the surface potential. This is confirmed by the ζ-potential data, which shows no PFO-induced changes. Further verification is obtained from the shape of the scattering patterns for different PFO concentrations: since the χs,2(2) and Φ0 terms are weighted by different scattering angle-dependent form factor functions, the SHS patterns do not change shape as the PFO concentration is increased. It means that the intensity changes must either come from changes in χs,2(2) or from changes in Φ0 but not from simultaneous changes.

The coherent normalized SHS intensity patterns (S(2ω, θ)) of Figure C are thus the sum of the scattering patterns of individual bare LUVs (sL(2ω, θ)) and LUVs that have a PFO protein (sLP(2ω, θ)) inserted. With N0 being the original number of liposomes and N the number of liposomes with a bound protein, we obtain for the total intensity:The relation between the coherent SH intensity (I(2ω, θ)) and the plotted response (S(2ω, θ)) is given by eq S1 in the SI. Having confirmed that the SHS response arises purely from the degree of water ordering at the interface, we performed measurements at the scattering angle of maximum intensity (θmax, Figure C). Figure D displays the difference in SH response, ΔS= |S(2ω, θ) – S(2ω, θ)|, where SL(2ω, θ) is the coherent water normalized SHS intensity from liposomes without PFO, as a function of the PFO concentration in the solution. We employ two types of LUVs, composed of 80:20 mol % (red squares) and 50:50 mol % (blue circles) SM:Chol, respectively. The first LUVs with lower Chol contents show no apparent change in the SHS intensity, while those with 50 mol % Chol exert a gradual increase at ∼100 fM and a rapid rise around ∼1 pM of PFO. The SH intensity difference levels off at 10 nM, which represents the saturation of the system meaning that no more proteins can bind to the liposomes. This protein-induced change in the intensity is reminiscent of a Langmuir isotherm. We describe this interaction as an equilibrium, to retrieve N (the number of liposomes assumed to have a (pre)pore (Figure A)) to obtain an expression for the measured SHS intensity difference. The PFO (P) binding to liposomes (L) can simplistically be described byThis follows a kinetic equation relating the rate of insertion (k1) to the change in the number of liposomes with a PFO (N):At equilibrium conditions (dN/dt = 0), we then havewhere Kd = k–1/k1 is the equilibrium PFO binding/insertion constant. Connecting this to eq , the measured difference in the coherent SH intensity of Figure D then becomesApplying this expression to the data in Figure D, we obtain a pore insertion constant of Kd = (1.5 ± 0.2) × 10–12 M.

Using the high throughput angle-resolved SHS technique, we thus demonstrate ultrasensitive detection of protein binding to liposomes. This is of special interest since standard methods require working on a supported lipid bilayer or surface-attached vesicles. QCMD or other mass-based binding assays lose sensitivity at weights less than nanograms of protein. SPR-based technologies require kinetic data for recording and modeling, which limits their sensitivity in the micromolar range. In addition, SHS as presented here does not require any planar substrate and can utilize free-floating liposomes in an aqueous environment. Moreover, the obtained SHS data are quantifiable and could be extended in the following ways: Obtained values for the binding constant, the second-order surface susceptibility of water, and the electrostatic surface potential[23] could be connected with MD simulations as is currently done for X-ray crystallography and cryo-electron microscopy. In fact, if MD-predicted molecular conformations would have been used here, it would have been possible to explicitly compute the SHS response, rather than using a more simplified assumption as was done here. It is also expected that different interaction mechanisms will lead to different intensity or scattering pattern responses. For example, a drastic restructuring of the liposomes (e.g., transformation into bicelles or micelles) will change the second-order nonlinear scattering form factors and thus alter the intensity in a different manner than worked out here. Another example might be a protein or pore that undergoes a drastic change in the charge distribution as this will result in a different (but still quantifiable) change in the effective third-order form factors. Many of such structural and charging effects have been modeled previously.[32−35] Thus, the SHS technique advances the toolbox for quantifying structural biology approaches toward high throughput, smaller samples, and label-free and noninvasive approaches. This will lead to a better understanding of membrane–protein interactions.

In summary, besides QCM, SPR, and floatation assays that are commonly used to determine the protein–membrane binding affinity, we introduced a complementary method, high-throughput second harmonic scattering, that is overcoming many limitations of these techniques. The SHS method enables to mimic the native environment of a membrane protein since the membrane composition and curve could be simulated, bridging the current gap between in vitro and in vivo experiments. We have shown the ultrasensitive detection of protein liposome interaction in an aqueous solution with a sensitivity in the femtomolar to picomolar range, which translates to a single transmembrane protein being bound to a single liposome. Furthermore, high-throughput SHS can potentially be performed with a small footprint tabletop instrument that uses small sample volumes (<10 μL), instead of 800 μL used in the current study, and small amounts of protein (nM). More importantly, as shown here, it enables the detection of membrane–protein interactions in the low femtomolar regime in a direct, label-free, model-free, and noninvasive manner.