Light Scattering By Optically-Trapped Vesicles Affords Unprecedented Temporal Resolution Of Lipid-Raft Dynamics

A spectroscopic technique is presented that is able to identify rapid changes in the bending modulus and fluidity of vesicle lipid bilayers on the micrometer scale, and distinguish between the presence and absence of heterogeneities in lipid-packing order. Individual unilamellar vesicles have been isolated using laser tweezers and, by measuring the intensity modulation of elastic back-scattered light, changes in the biophysical properties of lipid bilayers were revealed. Our approach offers unprecedented temporal resolution and, uniquely, physical transformations of lipid bilayers can be monitored on a length scale of micrometers. As an example, the deformation of a membrane bilayer following the gel-to-fluid phase transition in a pure phospholipid vesicle was observed to take place across an interval of 54 ± 5 ms corresponding to an estimated full-width of only ~1 m°C. Dynamic heterogeneities in packing order were detected in mixed-lipid bilayers. Using a ternary mixture of lipids, the modulated-intensity profile of elastic back-scattered light from an optically-trapped vesicle revealed an abrupt change in the bending modulus of the bilayer which could be associated with the dissolution of ordered microdomains (i.e., lipid rafts). This occurred across an interval of 30 ± 5 ms (equivalent to ~1 m°C).


Heating rate during the light-scattering measurements
The data shown in Figure S2 is the temperature-time data from the same experiment from which the light-scattering intensity data was used in Figure 1(a) of the original manuscript. For each measurement of the modulated light scattering intensity during the heating of an optically-trapped liposome, a new droplet of the liposome suspension was dispensed onto a cover glass. The heating rate was uncontrolled (i.e. there was feedback control from the thermocouple), however, the heating rate was still smooth, with an approximately linear rate of 0.029 C s -1 in the region of the gel-to-fluid transition, which occurred at ~180s (44 C) in Figure 1(a), and is the region highlighted in the plot shown in Figure S2(b).

Figure S2:
The profile of the temperature ramp for the experiment reported in Figure 1(a). The gel to fluid phase transition was observed between 43.9 -44.0 C, approximately 180 s after the start of the experiment. The heating rate at this time was 0.029 C s -1 .

Further detail on the accuracy of the temperature measurement
The design of the instrument enables the width of a phase transition to be measured on both time and temperature scales. A current limitation is that an accurate value for the actual temperature of the phase transition cannot be obtained. For example, the main transition temperature for DPPC liposomal bilayers was recorded at 44 C (see Figure 1), wherea temperature of 41.6 C has been determined by dilatometry.
In our experiments, the temperature is reported based on a measurement with a contact K-type thermocouple positioned at the edge of the heated aluminium plate (see Figure S3). There was a positive temperature gradient between the edge of the aluminium plate (1) and the surface area of the aluminium plate that was actually in contact with the coverglass (2); this area of the aluminium plate is directly above the bonded-Kapton flexible element. Furthermore, there was a negative temperature gradient between the area of the coverglass in contact with the heating plate (2) and the centre of the coverglass (3); the liposome was optically-trapped at a height of ~50 m above the centre of the coverglass.
An estimate of the magnitude of the temperature difference, T(1)  (3), when the bonded thermocouple at the edge of the aluminium plate reports a temperature, T3, of 44 C can be made; i.e. the temperature at which the main phase transition of the DPPC liposomal bilayer was observed (see  The precision of a temperature measurement by the infrared thermometer at (2) was  0.5 C. The error reported in Fig. S2 for T3 takes into account the uncertainty in the emissivity value and the precision of T2.
To achieve a more accurate temperature calibration in the future, we intend to make a simultaneous measurement of a temperature-sensitive spectroscopic band. For instance, the Raman band for O-H stretching in water shows a temperature dependence that could be monitored alongside the elastic-light scattering intensity. 1 However, the modulation of the O-H stretching band with temperature is subtle. A more realistic approach would be the addition of a temperature-sensitive fluorescent probe to the aqueous phase.

IV. Lipid structure-related interpretation of Raman spectra
Our work has focussed on the C-H band, which is by far the strongest feature in lipid-Raman spectra.
However, we will start by giving some background on other spectral regions. The main features in Raman spectra of lipid molecules are analogous to those of long chain n-alkanes. There are distinct Raman bands within the C-C stretching region corresponding to tight and loose packing of the hydrocarbon chains in different phases, namely: the asymmetric and symmetric C-C stretching bands at c.1060 and 1130 cm -1 , which originate from all-trans C-C bonds; and the skeletal C-C stretching band at c.1080 cm -1 from gauche segments of the alkyl chain. The relative intensity of the bands at c.1130 and 1080 cm -1 has been used to estimate the number of trans-conformational segments per alkyl chain, which would be higher in the Lo phase relative to the Ld phase. 2 The intensities of all the bands in the C-C stretching region are markedly lower than the bands in the C-H region, which is the reason why the latter was recorded in our experiments.
In a further spectral region, a blue-shift in the CH2 twisting mode at c.1300 cm -1 is also consistent with the disordered phase of lipid bilayers and structures that contain a broader distribution of gauche rotamers. The packing of the hydrocarbon chains also strongly affects the relative intensity of bands for the CH2 twisting mode and the CH2 scissor mode at 1440 cm -1 but, unlike bands in the C-C stretching region, these methylene bands are not unique for trans and gauche segments, i.e. ordered and disordered lipid structures. The frequencies of the bending and stretching methylene modes are sensitive to the number of gauche segments. This is due to increasing steric repulsion between neighboring chains. However, the blue shift can be difficult to resolve in the C-H stretching region due to the overlapping band structure. All the methlyene bands in this region are much weaker than the C-H stretching bands. The CH2 wagging mode at c.1370 cm -1 is an extremely weak band.
Our work has focussed on the C-H band, which is by far the strongest feature in the lipid-Raman spectrum. There are a similar number of peaks in the C-H band for different lipid and hydrocarbon molecules. 3 The C-H region comprises the symmetric (d + ; c. 2840 cm -1 ) and antisymmetric (d -; c. 2870 cm -1 ) methylene stretch, the Fermi resonance of the symmetric methyl stretch (rFR + ; c. 2920 cm -1 ) and the antisymmetric methyl stretch (r -; c. 2960 cm -1 ). 4 A weak band is also observed at 3030 cm -1 for the antisymmetric CH3 stretch of the choline head group. The intensity ratio between the dand d + bands, or, alternatively, the dand rFR + bands, has been reported in the literature to provide a qualitative measure of short-range packing order of the hydrocarbon chains, where a larger ratio is indicative of greater order (alternatively the ratio has been used to measure intermolecular interactions or vibrational/torsional motion of the lipid chain). 5,6 It should be noted that the integrated intensity of the dand d + band are not believed to be affected by the relative number of trans and gauche segments, and only the relative intensity of the peak maxima. With a larger number of gauche segments, the d + band is broadened with substantial intensity gained at c.1855cm -1 . 5 The dfrequency has also previously been associated with the extent of chain decoupling and rotational diffusion of lipids. As the number of gauche segments increases, the dband broadens in the spectrum and merges with the Fermi resonance band. Thus, the quantitative significance of the d -/d + ratio, which some researchers refer to as the order parameter, is uncertain.
Although the d + , d -, rFR + and rpeaks are observed in the C-H band for different lipid molecules, the intensity ratios are different, which further impacts on any ability to deduce the degree of trans- V. Raman spectra of 1,2-dipalmitoyl-sn-glycero-3-phosphocholine, DPPC, at various temperatures. Figure S11 shows the spectrum of inelastic (Raman) scattered-light from an optically-trapped DPPC liposome measured in the gel (L), ripple (P) and fluid (L) phases. The C-H region comprises the symmetric (d + ) and antisymmetric (d -) methylene stretch, the Fermi resonance of the symmetric methyl stretch (rFR + ) and the antisymmetric methyl stretch (r -). The assignment and notation is in accord with Ref. (7). A weak band is also observed at 3030 cm -1 for the antisymmetric CH3 stretch of the choline head group. The intensity ratio between the dand d + bands, or, alternatively, the dand rFR + bands, has been reported in the literature to provide a qualitative measure of short-range packing order of the hydrocarbon chains, where a larger ratio is indicative of greater order. 9 The decrease in both of these ratios above the main transition for a single DPPC liposome is consistent with previous measurements on planar supported lipid bilayers 11 (1:1 mixture of 1palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, POPC, and cholesterol) as a function of temperature: The transition from a liquid-ordered phase to co-existing liquid-ordered and disordered phases. In Figure S12(a), the Raman spectrum of an optically-trapped POPC/cholesterol liposome is shown as a function of temperature. All the spectra were recorded from the same trapped vesicle at regular intervals of time during a temperature ramp from 20 to 60 C (80 spectra, 30 s intervals; approx.linear ramp of +0.025 C s -1 ). The d + , d -, rFR + and rbands can be distinguished in the broad C-H region, but the spatial resolution of the individual bands is not as clear as that for a pure DPPC liposome. The change in the structure of the lipid bilayer between room temperature, where a pure Lo phase should exist, and high temperature, where Lo and Ld phases should co-exist, appears to result in a small increase in the ratio d -/ d + . This is not the direction of change that was expected for the ratio of the CH2 stretching bands. An increase in the ratio is usually observed with greater short-range order of the hydrocarbon chain packing; 5 thus, a decrease in the ratio d -/ d + was expected following the Lo to just below 1.0). We have observed, in Figure S4 (a), an increase in d -/d + from approximately 0.95 to 1.10, between 293 and 333 K. The significance of the increase, apparent in Figure S12(a), indicates that it is not strictly reliable to associate the relative intensities of Raman peaks with the physical properties of lipid membranes.

VI. Raman spectra of an optically-trapped liposome
Although the change in ratio d -/ d + with temperature for a POPC/cholesterol bilayer does not conform to changes observed for other bilayer compositions, a shift in the dband to higher frequency can be seen in the raw spectral profiles in Figure S12(a) which is consistent with interpretation in the older literature. The dfrequency has previously been associated with the extent of chain decoupling and rotational diffusion of lipids in bilayers. 5 Furthermore, there is a substantial increase in the intensity of the rFR + band relative to the dband in Figure S12(a), which has also been used as an indicator of packing disorder. 5 The results of multivariate curve resolution on the sequence of Raman spectra from Figure S12(a) is shown in S12(b) and (c). A fitting of two components captured 99.0% of the variance in the experimental data. In S12(d), the residual is shown for a representative example from the sequence of experimental spectra shown in S12(a). The residual in S12 ( There is debate in the literature as to whether the properties of binary mixtures of lipids should be described by phase-separated Lo and Ld regimes, or by gradual changes in a largely homogeneous lipid bilayer. 20 By measuring the light scattering signal from an optically-trapped liposome, we have been able to probe an area of lipid bilayer corresponding to a few m 2 and detect heterogeneities due to differences in light scattering from the Lo and Ld phases, which have been difficult to characterise using conventional thermal analysis techniques or Raman spectroscopy.

VII. Control measurement of an optically-trapped liposome (1:1 mixture of 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphocholine, POPC, and cholesterol) at fixed temperature.
A sequence of Raman spectra was recorded from an optically-trapped POPC/cholesterol vesicle maintained at ambient temperature. The raw spectral data are shown in Figure S13(a). The spectra were recorded from the same trapped vesicle at regular intervals of time (36 spectra, 30 s intervals).
No change in the frequencies or intensities of peaks in the C-H stretching region was observed at constant temperature. The results of multivariate curve resolution on the sequence of Raman spectra from Figure S13(a) is shown in (b) and (c). The spectral data set was fitted with two components, using the same approach made to analyse the temperature-dependent Raman spectra (see Figures 3 and S12). A fitting of two components captured 99.7% of the variance in the experimental data, however, the fitted components, A and B, show nearly no change in the relative intensities and positions of the d + , d -, rFR + and rbands. Due to a change in the overall intensity of the recorded Raman spectra, which is the result of focus drift during the course of the experiment, the components A and B reflect the small difference in the recorded spectra following background subtraction, and the component weighting still shows a gradual change between A and B as a function of time. In Figure   S13(d), the residual is shown for a representative example from the sequence of experimental spectra in S13(a).
Raw spectral data are superimposed in Figure S14 illustrating the initial and final measurements for the control experiment (fixed temperature in (a)) and the heating experiment (in (b); reproduced from