Spectroscopy of model-membrane liposome-protein systems: complementarity of linear dichroism, circular dichroism, fluorescence and SERS

Membrane proteins are estimated to comprise up to 30% of all protein types in biological systems [1]. However, this is not reflected in structural data repositories with membrane proteins typically being 1% of available types [2]. To bridge this inconsistency, methods that probe membrane proteins in environments that mimic a biological system are required.

Molecular spectroscopic methods involve measuring changes that occur when electromagnetic radiation (often generically referred to as light) interacts with a molecular system. The wavelength of light used ranges from radio waves to γ-rays, whilst species studied range in size from single molecules containing a few atoms to large molecular assemblies [3]. This review focuses on certain spectroscopic measurements of unilamellar liposomes — spherical model of membrane bilayer systems — with proteins or peptides on or in the membrane (Figure 1). Unilamellar liposomes [4] are artificial membrane system which mimic many properties of the biological system being composed of a single lipid bilayer, containing one or more types of lipids. Biomolecules can bind on the lipid surface or in the bilayer to enable us to study structure, kinetics, catalysis, membrane disruption, and membrane transport.

The first questions usually asked about protein–membrane interactions are:

• (i) has the protein bound to (on or in) the membrane?

• (ii) if so, what binding geometry has the protein adopted?

• (iii) has the interaction affected the protein or the membrane?

Rather simplistically, we can summarise the applications of different spectroscopies as follows. Fluorescence spectroscopy can be used to see whether environments around intrinsic and extrinsic fluorophores have changed. With targeted labelling, fluorescence can indicate the depth of a protein in the membrane, as well as giving an indication of protein and lipid dynamics [6]. Circular dichroism (CD) spectroscopy quickly indicates whether a peptide or protein has changed conformation [7] and linear dichroism (LD) [8] can be used to determine the orientation of protein secondary structures with respect to the membrane normal. These spectroscopies are usually implemented with UV-visible light. The lipids are only apparent with UV techniques due to the light scattering that occurs when the lipids assembled into particles (liposomes) comparable in size to the wavelength of the light. To probe the lipids per se, vibrational spectroscopies such as infra-red absorbance (IR) and Raman may be employed. Our focus here is on Surface Enhanced Raman Spectroscopy (SERS) [9]. Membrane–protein spectroscopy experiments are prone to artefacts resulting from light scattering and absorption flattening so care must be taken. The scattering component of any spectroscopy measurement can be minimised by reducing the distance from the sample to the detector. However, scattering is seldom completely removed and care must be taken while undertaking a baseline correction, especially when the spectroscopic signals are small relative to the scattering. It should be noted that scattering signals do contain useful information [10], but with liposomes little work has been done to develop methods to extract it [11].

Fluorescence spectroscopy involves measuring the amount and wavelength of light emitted after molecules have been excited to a higher energy level. There is a wide range of fluorescence techniques, but our focus is on ones that are most useful for membrane–protein systems [6]. Nearly all fluorescence occurs from the lowest electronic excited state, as solution-phase molecules usually find non-radiative pathways to lose the energy to get back to that level. Fluorescent chromophores then return to the ground state through the emission of light. The frequency of the emitted light is typically less than that required to excite the molecules as illustrated in Figure 2A. For biomolecules we usually choose an excitation wavelength and scan over the emission wavelength, because scanning the excitation wavelength produces a spectrum similar to its absorbance spectrum (though usually with the longest wavelength band enhanced in intensity relative to shorter wavelength bands). The simplest application of protein emission fluorescence is to monitor any shift of tryptophan fluorescence when a protein or peptide is added to a solution of liposomes. A hydrophobic environment causes the tryptophan fluorescence maximum to move towards 330 nm from values as high as 350 nm. The usual excitation wavelength is 280 nm, but this induces fluorescence from both tyrosines and tryptophans. The former is not environment sensitive, so exciting at 295 nm selectively for tryptophan gives clearer information about its surface environment. An example of the fluorescence of gramicidin A's tryptophans in water and in a membrane is shown in Figure 2B. It should be noted that merely dipping the tryptophans into the headgroup region of the bilayer is sufficient to move the fluorescence to 325 nm, and that no further change is observed when the backbone of gramicidin fold and inserts (see below).

FRET (fluorescence resonance energy transfer) and FRAP (fluorescence recovery after photobleaching) both provide useful information about protein–lipid systems. FRET involves illuminating a fluorophore and measuring the resulting emission from another (donor and acceptor in Figure 3) — so they have to be close enough to transfer energy, with the intensity following Förster's [13, 14]. FRET is especially useful in estimating intramolecular distances and dynamic conformational changes perhaps as a function of property change measurements (e.g. a pH change) [15]. The technique has limitations connected with the need for an absorbing chromophore/accepting fluorophore pair and low signal-to-noise ratio. These issues can be overcome in most cases [16].

For example, Gorbenko et al. [17] used FRET assay to estimate the penetration depth of 1–83 N-terminal fragment of the protein apoA-I (A83) into the lipid bilayer. The FRET efficiency between donor (tryptophan) and acceptor (4′-N,N-dimethylamino-4 methylacryloylamino chalcone, DMC) was determined. The results helped to estimate the penetration depth of apoA-I variants into the membrane, which proved to depend on the cholesterol concentration. The work also illustrates the value of molecular dynamics simulations to interpret fluorescence data. The same principle was used by Elderdfi and Sikorsk [18] to find the equilibrium dissociation constant value of the binding affinity of membrane palmitoylated protein-1 (MPP1) with DOPC/SM/CH liposomes. The fluorescence signal of the dansyl labelled liposomes (acceptor) increased in the presence of tryptophan residues of protein MPP1 (donor). Membrane–protein interactions have also been studied by Lasitza-Male et al. [19] by probing the dynamics of DtpA protein (a proton-transporter) in various lipid environments (Figure 4).

The existence of various lipid fluorophores allows the use of the FRET techique to study lipid layer dynamics. For instance, the phase separation behaviour of a lipid bilayer containing the photopolymerizable, diacetylene lipid (DiynePC) was investigated by Okuno et al. [20] using a FRET assay with NBD-PE and Rhod-PE as a pair of fluorescent lipids, distributed in the DOPC-rich phase in lipid bilayers. When the phase separation triggered by temperature change occurred between DiynePC and DOPC, fluorophores became concentrated in DOPC domains and FRET efficiency increased.

FRAP (fluorescence recovery after photobleaching) is less relevant in spectroscopy than in microscopy. In FRAP, the monitored fluorescent region is bleached by a high-intensity laser. The bleached region then recovers due to the diffusion of molecules. The recovery of fluorescence intensity is monitored and recorded into a recovery curve (Figure 5). The plotted data contains the information required to calculate lateral diffusion coefficients, which are in turn used in membrane dynamics study. FRAP has been applied for comparative study of new types of liposomes [21–23], fluidity of the lipid bilayers [24, 25] and proteins’ dynamics in the membrane [26, 27]. The main limitations of FRAP are photoswitching of analyte chemical or structural analytes and localised heating by the laser [28] which can change the sample.

CD, to a good approximation, may be deemed to be the sum of the signals of all the component parts, and there are several structure-fitting methods which can be used to give a reasonable estimate of the number of residues that are in different secondary structure motifs [32–40]. Most approaches select members of a reference set of protein CD spectra that most closely resemble an unknown text spectrum and then determine the secondary structure context of the unknown protein from a weighted sum of the secondary structure percentages of the best matching members of the reference spectra [36, 41–44]. The website DichroWeb enables registered users to apply SELCON, CONTIN, CDSSTR and K2D analysis programs with a range of reference sets to do this [45, 46].

When one is considering the CD of proteins in membrane environments [47, 48], it is important to recognise that the wavelengths of transitions and shapes of spectra may move relative to their positions in aqueous solution in part due to the dielectric constants of the medium and in part because secondary structure motifs in membranes tend to be longer than in globular proteins, thus having fewer residue in the less tightly held ends of e.g. helices. Red shifts of n-π* transitions (which occur at ∼220 nm in proteins) in hydrophobic environments are well-established, with the converse operative to a lesser extent for π-π* transitions (190–210 nm in proteins). Such effects are sometimes observed for membrane proteins [7]. It is almost always the case that the 208 nm signal for a helical membrane protein/peptide is greater than its 222 nm signal [49].

It is widely accepted that CD spectra down to 190 nm contain three independent pieces of information whereas data down into the vacuum UV region (say 175 nm) can provide more information [50]. However, we have recently shown, using Bayesian inference, that it is not possible to identify more than 3 distinct secondary structure classes from CD spectra above 175 nm, though more data do slightly improve the quality of the helix and sheet estimates [51]. The quality of the reference set has more impact on the reliability of a secondary structure estimate from CD. Therefore it is important that any reference set used for membrane protein structure fitting includes proteins in similar environments [52]. We have also shown the importance of being able to interrogate which members of a reference set are being selected, especially for the situation where random coil, polyproline II, and so-called β_II structures are present as they are structurally different but provide have essentially identical far UV CD spectra. So, if a best matching spectrum has been identified as having β_II structure (e.g. chymotrypsin, elastase, chymotrypsinogen, carbonic anhydrase [53] or indeed the β-structure of lysozyme [54]) care must be taken when attributing the secondary structure of an unknown to β, polyproline or random coil. For this purpose, our own self-organising map fitting approach SOMSpec is preferred because its output includes the map from the trained reference set and the position on the map of the unknown [44, 55].

In general, the key to interpreting a liposome flow LD spectrum is to know the transition polarisations of the molecules that are bound to the membrane [56]. Some key chromophores for peptides and proteins are illustrated in Figure 7B. The α-helix shows a clear n-π* transition at ∼220 nm in both CD and LD that is polarized across the short axis of the helix. So, if a helical peptide is inserted into the membrane parallel to the lipids, 220 nm has a positive LD signal. The other two transitions which can be access for an α-helix are two perpendicular exciton components of the first π-π* transition which occur at ∼210 nm and 190 nm. When inserted into the membrane, the 210 nm signal is expected to be negative, however, in practice looks like a positive minimum between two larger positive signals (Figure 7C). A β-sheet by way of contrast shows little net LD at 220 nm due to two cancelling transition components and the π-π* transition is a single band a bit below 200 nm.

Liposome flow LD experiments can be used to follow the kinetics of membrane binding if it happens slowly enough. Normal Couette flow cells have a deadtime of ∼30 s (the cell assembly time). Rapid injection [64] and channel flow systems [65] can be used to reduce the dead time to 500 µs and 25 µs respectively, though with the penalty of more sample and worse signal-to-noise relative to Couette flow.

One of the challenges of liposome LD spectroscopy, as discussed above, is that overlaid on any liposome LD spectrum one collects is a scattering or turbidity LD, as the unilamellar lipid vesicles typically used in LD experiments are of the order of the wavelength of light. Since a spectrometer detector simply counts the photons that are incident onto it, photons that do not arrive due to scattering are not differentiated from those missing due to absorbance. The spectrum in Figure 7C shows the original liposome flow LD spectrum of a helical peptide before and after correction for scattering [66].

Hybrid techniques are invariably more complex to implement than either of their components but are richer in information content. Fluorescence-detected LD (FDLD) is one such example where the sensitivity and selectivity of fluorescence can be used to probe the orientation of selected chromophores in a sample (Figure 8). By inserting high quality long-pass cut-off filters into a standard CD machine that has been adapted for LD one can measure FDLD [67, 68]. It should be noted that some features of more modern instruments designed to ensure users collect good CD data may need to be disabled.

Vibrational spectroscopies give information about bonds, and hence conformations of molecules as a function of their environments [39]. For liposome-protein systems vibrational spectroscopies are particularly attractive as means of accessing the lipid bonds which are invisible in electronic spectroscopy. Infra-red absorbance (IR) is most widely used in general applications. However, most protein–lipid work in the literature is for samples deposited on surfaces rather than in solution. In addition, deuterated samples are often preferred due to overlap of water and protein amide I signals [71–73]. Thus, although there is great potential for membrane/protein IR, here we focus on how Raman spectroscopy can be used to give information to understand and control cellular processes.

Raman has the advantage over IR, whereby H₂O scattering is comparatively weak so that the signals of analytes can be extracted from higher concentration solutions. However, Raman scattering by its nature — it is a second-order process [74] — is generally weak with poor signal to noise ratios. Signals can be enhanced by Raman tags or ‘reporter’ molecules which are bound to targeted compartments of the membrane to generate enhanced Raman signals. Alternatively, so-called label-free ‘surface-enhanced’ Raman spectroscopy (SERS), which is based on the use of plasmonic nanomaterials to obtain enhanced Raman signals, may be used [9, 75]. There is a consensus in modern scientific society on the mechanisms underlying the enhancement of the Raman signal (enhancement factor or EF). The EF builds on the combination of electromagnetic enhancement that occurs due to plasmon resonance on the metal particles, and chemical enhancement caused by the (partial) transfer of electrons between metal particles and molecules near their surface. As illustrated in Figure 9, label-free SERS can enhance Raman signals by at least two orders of magnitude, enabling single molecule SERS in some cases [76].

The incubation of liposomes with metal nanoparticles [77–80] is simple and does not require much preparation, with quality of data largely being dependent on the quality of the nanoparticles. The method has proved to be useful for lipid membrane analysis [81]. However, the descriptor, e.g. ‘gold nanoparticle’ is misleading as the metallic centre invariably has another ion, such as citrate, or, any other purpose-specific functionalisation [82] on the surface of the metal core. The surface-functionalisation [83] as well as particle size [84] and shape [85] influence the membrane-particle interaction and should be considered. If a membrane protein is being studied by SERS with metal nanoparticles it is advisable to check whether the particle–protein interaction is changing the secondary structure of the protein. CD is usually the simplest option.

Examples of how SERS can be used with liposomes are provided by Faried et al. [77] who performed location-specific analysis on the DMPG membranes (Figure 10) using 1-octanethiol capped gold nanoparticles of sizes 100 nm and 5 nm. The larger 100 nm particles preferentially enhanced signals from the surface region of the monolayer (e.g. C = O), while 5 nm particles became embedded inside the DMPG bilayer, enhancing the hydrophobic region (e.g. trans C–C signals and others in the fingerprint region (500–2000 cm^–1) and C−H stretching regions (2700−3100 cm⁻¹)). Small particles influenced the lipid fluidity and polarity of the membranes by weakening hydrophobic interactions between lipids’ molecules. However, overall the effect was not significant and it was concluded that embedding of the particles into the lipid bilayer does not disturb the membrane properties, which is also supported by their previous observation for DOPC/CH liposomes [82].

Novel approaches to membrane (phosphatidylcholine, sphingomyelin, and cholesterol) characterisation with citrate-stabilised gold nanoparticles for SERS in tandem with cryogenic electron microscopy (cryo-EM) have also been reported [80]. The EM images indicate nanoparticles of 30–50 nm in size and lead to the conclusion that the nanoparticles do not affect the membrane structure whereas higher concentrations of citrate do. Often, a principal component analysis is required to extract differences between preparations. For example, Delgado-Coello et al. [78] used 30 nm citrate-capped gold nanoparticles to characterise sphingomyelin/cholesterol-rich liposomes derived from cell membranes in order to describe the catalytic activity of a Ca²⁺-ATPase calcium transporter as a function of the local lipid environment.

SERS gold nanoparticles also show potential to characterise not only artificial membrane systems but also the membranes of whole cells. Živanović et al. [86] studied (with 30 nm citrate capped gold particles) the distribution and interaction of the lipid molecules in parasitophorous vacuoles and the parasite of Leishmania-infected macrophage cells. Ramya et al. [79] were able to clearly distinguish lipids from other macromolecules including proteins, carbohydrates, and photosynthetic pigments via SERS (with 40 nm citrate capped gold particles) in Scenedesmus quadricauda CASA-CC202 under nitrogen starved condition and different pH.

Overall, SERS is a powerful technique for biomolecule characterisation. The different approaches and methodologies under this tool allow selective, dynamic, and site-specific study of the membrane compartments and processes. SERS can significantly enhance signals thus enabling very small amount of sample to be studied. Label-free SERS with gold nanoparticles is a relatively simple approach to obtain Raman signal of the membrane lipids and proteins. By tuning particle properties such as size, capping agent and shape, different sites can be targeted. It can be used in various applications including the monitoring of self-assembled lipids monolayers and characterisation of the whole cell membrane.

• Roughly 30% of all protein types in biological systems are membrane proteins. However, this is not representative with membrane protein being 1% of available types in our protein data bank. To bridge this inconsistency, methods that probe their structures in environments that mimic a biological system are required.

• Spectroscopic techniques, including absorbance, fluorescence, circular dichroism, linear dichroism, and surface enhanced Raman spectroscopy, can be applied to peptides and proteins bound to unilamellar liposome lipid bilayers to give structural information that can be combined to form a picture of the membrane–protein interaction.

• Combinations of techniques and their advancements are being developed to further enhance selectivity and sensitivity. Examples include surface enhanced Raman spectroscopy and fluorescence-detected linear dichroism.

The authors declare that there are no competing interests associated with the manuscript.

A.T. acknowledges funding from Macquarie University Postgraduate Scholarship.

Open access for this article was enabled by the participation of Macquarie University in an all-inclusive Read & Publish pilot with Portland Press and the Biochemical Society under a transformative agreement with CAUL.

The authors each wrote the first draft of half the text and half the diagrams and they wrote the final document together.

apoA-I (A83)

Apolipoprotein A1

ATP

Adenosine Triphosphate

AuNP

Gold Nanoparticle

CD

Circular Dichroism

CH

Cholesterol

DiynePC

Diacetylene Lipid

DMC

4-Dimethylaminochalcone

DMPG

1,2-Dimyristoyl-Sn-Glycero-3-Phospho-Rac-(1-Glycerol)

DNA

Deoxyribonucleic Acid

DOPC

Dipalmitoyl Phosphatidylcholine

DtpA

diethylenetriaminepentaacetic acid

EF

Enhancement Factor

EM

Electron Microscopy

FDLD

Fluorescence Detected Linear Dichroism

FRAP

Fluorescence Recovery After Photobleaching

FRET

Fluorescence Resonance Energy Transfer

FTIR

Fourier-transform infrared spectroscopy

IR

Infra-Red

LD

Linear Dichroism

MPP1

Membrane Palmitoylated Protein-1

NBD-PE

Nitrobenzoxadiazole Labelled Lipid

NP

Nanoparticles

NP₁₀₀

100 nm nanoparticle

NP₅

5 nm nanoparticle

PE

Phosphatidylethanolamine

PS

Phosphatidylserine

Rhod-PE

Rhodamine Labelled Lipid

SERS

Surface Enhanced Raman Spectroscopy

SM

Sphingomyelin

TFE

Trifluoroethanl

UV

Ultra-Violet