Phase Separation of Lipid Membranes Analyzed with High-Resolution Secondary Ion Mass Spectrometry

See allHide authors and affiliations

Science  29 Sep 2006:
Vol. 313, Issue 5795, pp. 1948-1951
DOI: 10.1126/science.1130279


Lateral variations in membrane composition are postulated to play a central role in many cellular events, but it has been difficult to probe membrane composition and organization on length scales of tens to hundreds of nanometers. We present a high-resolution imaging secondary ion mass spectrometry technique to reveal the lipid distribution within a phase-separated membrane with a lateral resolution of ∼100 nanometers. Quantitative information about the chemical composition within small lipid domains was obtained with the use of isotopic labels to identify each molecular species. Composition variations were detected within some domains.

Imaging and quantifying the static and dynamic variations in lateral composition that result from interactions among membrane components is a major challenge in structural biology. Although biological membranes are fluid structures and fluidity is essential for function, it is widely believed that some degree of lateral organization is present and that this organization is also essential for function (13). The relevant distance scale is larger than that of individual membrane proteins or protein assemblies (>10 nm), whose structures can be determined by x-ray crystallography or inferred from atomic force microscopy (AFM), but is substantially below the diffraction limit of light microscopy. Fluorescence microscopy is widely used and is extremely sensitive and specific to the labeled component (48), but only the labeled component is observed, and, at least for relatively small components such as lipids, the fluorophore may greatly alter the delicate interactions that are present in the membrane (9). Infrared (10) and coherent anti–Stokes Raman (11) imaging offer greater chemical specificity, but thus far the lateral resolution and sensitivity are limited. AFM provides much better resolution of topographical features but does not yield information on chemical composition (9, 1214). Imaging mass spectrometry offers distinct advantages over these methods (1521), and we applied this approach to imaging and analyzing the chemical composition of small lipid domains with lateral resolution of ∼100 nm.

Secondary ion mass spectrometry (SIMS) was performed with a NanoSIMS 50 (Cameca Instruments, Courbevoie, France). During NanoSIMS analysis, a focused 133Cs+ primary ion beam is rastered across the sample; secondary ions generated by sputtering are extracted and analyzed according to their respective charge-to-mass ratios at high mass resolving power (Fig. 1). By selectively incorporating a distinctive stable isotope into each membrane component [e.g., 13C or 15N], NanoSIMS secondary ion images characteristic of each species [e.g., 13C1H or 12C15N, respectively] can be used to create a component-specific compositional map of the sample. We used this approach to demonstrate the ability to image and analyze quantitatively the composition of very small lipid domains within a phase-separated lipid membrane.

Fig. 1.

Schematic showing NanoSIMS analysis of phase-separated lipid bilayer (not to scale). At room temperature, gel and fluid phases, mostly composed of 13C18-DSPC (red) and 15N-DLPC (green), respectively, are present in the bilayer. The gel phase is ∼2 nm higher than the neighboring fluid phase and can be imaged by AFM (compare with Fig. 2D). The sample is freeze-dried to preserve the lateral organization within the bilayer (fig. S1) and analyzed with the NanoSIMS. During NanoSIMS analysis, a focused 133Cs+ ion beam generates secondary ions; the negative ions are collected and analyzed in a high-resolution mass spectrometer. The secondary ions that are characteristic of 13C18-DSPC and 15N-DLPC (13C1H and 12C15N, respectively) are used to identify each component in the NanoSIMS image. The 133Cs+ primary ion beam is focused to a spot ∼100 nm in diameter (fig. S2) and is rastered across the sample to generate an image.

Supported lipid bilayers were prepared from vesicles containing equal mole fractions of 15N-labeled 1,2-dilauroylphosphatidylcholine [15N-DLPC, melting temperature (Tm) = –1°C] and 13C-labeled 1,2-distearoylphosphatidylcholine (13C18-DSPC, Tm = 55°C), with 0.5 mol % of a fluorescent lipid added to allow the bilayer quality to be evaluated by fluorescence microscopy during sample preparation (22). The sample was maintained at 70°C (above the Tm of both lipid components) to ensure complete mixing, both during vesicle and supported bilayer formation on prewarmed (70°C) silicon wafers. The silicon wafers were prepared with a thin (17 nm) SiO2 layer that facilitated the formation of stable bilayers while permitting charge dissipation during the SIMS analysis, and the wafers were patterned with chrome grids to corral the lipid bilayers and provide landmarks on the surface for characterization of the same regions by fluorescence, AFM, and NanoSIMS imaging (22, 23). The homogeneous, supported bilayer samples were slowly cooled to room temperature to induce phase separation (Fig. 1), rapidly frozen, and freeze-dried to remove water without disrupting the lateral organization within the membrane (fig. S1). Note that the lipid bilayers are fully hydrated before being frozen. Before NanoSIMS analysis, the geometries of the gel-phase domains, which are thicker than the fluid-phase regions, were characterized by AFM for subsequent comparison to the NanoSIMS data. As shown in Fig. 2D, the AFM image of the freeze-dried supported lipid bilayer contained domains that extended ∼2.0 nm above the neighboring bilayer, in good agreement with the reported height difference (1.8 nm) between gel-phase DSPC and fluid-phase DLPC in a hydrated supported lipid bilayer on mica (24, 25).

Fig. 2.

A phase-separated supported lipid bilayer that was freeze-dried and imaged by NanoSIMS and AFM. (A to C) NanoSIMS images of the normalized 12C15N signal that localizes 15N-DLPC (A), the 13C1H signal that localizes 13C18-DSPC (B), and the overlaid 12C15N and 13C1H signals (C). (D) An AFM image of the same region on the sample taken before NanoSIMS analysis. The contrast levels within the NanoSIMS images reflect the normalized signal intensity, corresponding to 100 and 0 mol % of the appropriate isotopically labeled lipid, as determined from calibration curves [see text and (22)]. Arrows indicate objects in the AFM images that are unlabeled debris, not domains, and their corresponding locations in the NanoSIMS images. Domains with diameters as small as ∼100 nm, as measured by AFM, were visible in the SIMS images (e.g., those highlighted with circles). NanoSIMS images were acquired with a pixel size of ∼100 nm by 100 nm.

For chemical imaging, the 13C1H and 12C15N NanoSIMS secondary ion signals were used to evaluate the distributions of 13C18-DSPC and 15N-DLPC, respectively, within the supported lipid bilayer. Although other secondary ions with nominal masses of 14 amu (12C1H 2) and 27 amu (13C14N) were generated during analysis, the mass resolving power was sufficient to resolve the 13C1H and 12C15N ions from these interfering isobars while maintaining high lateral resolution, which permitted un-ambiguous identification of the species of interest. The component-specific NanoSIMS secondary ion images in Fig. 2, A to C, show that the bilayer was not homogeneous. Distinct microdomains enriched in 13C18-DSPC, as evidenced by an increased 13C1H signal and decreased 12C15N signal, were dispersed within a 15N-DLPC–rich matrix. The area occupied by 13C18-DSPC within the bilayer was lower than that based on the molar ratio of the lipids in the vesicle solution as prepared. This difference in the lipid composition between the vesicle solution and the phase-separated supported lipid bilayer is likely due to selective adsorption of these very different lipid species (26).

Close examination of the sizes and shapes of the 13C18-DSPC–enriched domains observed in the NanoSIMS secondary ion images revealed that they were nearly identical to the domain geometry imaged by AFM at the same sample locations (Fig. 2D). Phase-separated domains with complex edge structures and domains as small as ∼100 nm in diameter, as measured by AFM, are visible in the NanoSIMS secondary ion images (Fig. 2, circles), confirming the high lateral resolution and sensitivity of the NanoSIMS technique. A few of the features in the AFM image did not produce lipid-specific secondary ion signals (Fig. 2, arrows); the height difference between these features and the bilayer (>5 nm, measured by AFM) confirmed that these objects were unlabeled debris and not lipid domains.

Quantitative information on the lipid composition within specified regions of the bilayer was obtained by calibrating the secondary ion yields against standard samples (22). Briefly, NanoSIMS measurements were made on sets of homogeneous supported lipid bilayers that systematically varied in the 13C18-DSPC or 15N-DLPC content, and calibration curves were constructed that correlated the normalized 13C1H or 12C15N signal intensities (13C1H/12C or 12C15N/12C) to the mol % of 13C18-DSPC or 15N-DLPC, respectively, within each sample (fig. S3). With this approach, the gel-phase lipid composition and uniformity were investigated by converting the component-specific secondary ion intensities collected at numerous locations within a single micrometer-sized domain into mol % concentrations (Fig. 3).

Fig. 3.

Details of correlated NanoSIMS and AFM images showing domain composition and topography. The 13C1H/12C and 12C15N/12C Nano-SIMS isotope ratio images show the abundance of 13C18-DSPC and 15 N-DLPC, respectively, within the bilayer, as determined from calibration curves (fig. S3). AFM images acquired at the same sample locations reveal topography. Lower concentrations of both lipids were detected in the locations where debris was identified (arrows). The lipid composition within the gel phase was usually consistent with the phase diagram predictions (domain A), but elevations in the amount of 15N-DLPC within the gel phase were occasionally detected at localized areas within the domains (domains B and C). AFM imaging indicated a small (< 200 nm) depression that could be a fluid-phase subdomain (circle) trapped within the gel phase (domain C); this is confirmed by the NanoSIMS image, which shows an elevated amount of 15N-DLPC across this region (see also Fig. 4). NanoSIMS images were acquired with a pixel size of 100 nm by 100 nm and are smoothed over three pixels. Scale bar, 1 μm.

We could often detect compositional heterogeneity within the gel phase. Although the majority of the domain consisted of a ∼9:1 mol ratio of 13C18-DSPC to 15N-DLPC, as predicted by the phase diagrams for DSPC and DLPC mixtures (2729), higher concentrations of 15N-DLPC were occasionally detected within the gel phase. To determine whether the 15N-DLPC distribution within the gel-phase domains varied in a statistically significant manner, we divided each domain into regions of 3 pixels by 3 pixels that did not include the domain edges or debris (Fig. 4) and used the calibration curves to determine the amount of 15N-DLPC within each region. The variations in the 15N-DLPC content within domains B and C (Fig. 4) were greater than the uncertainty in the measurements, which indicates that these domains contained statistically significant differences in lipid composition. AFM imaging revealed that the elevated 15N-DLPC concentration localized within one 13C18-DSPC–enriched domain (Figs. 3 and 4, domain C) corresponded to a small (diameter <200 nm) fluid-phase subdomain within the gel phase (Fig. 3, circle). We hypothesize that small gel-phase domains (tens of nanometers in diameter) that form early in the phase separation process coalesced around a small amount of 15N-DLPC, thereby trapping the fluid-phase subdomain within the growing gel-phase domain [see (30) for a theoretical model that may be relevant to this process]. A similar process may have produced the elevated concentrations of 15N-DLPC that were detected at localized regions within other gel-phase domains; however, the absence of topographical features that are characteristic of gel-fluid interfaces at these regions implies that either the fluid-phase subdomains were smaller than the lateral resolution of these AFM images, or the 15N-DLPC was well dispersed within these small regions of the gel phase. [Note that the lateral resolution of these AFM images, ∼70 nm, is significantly lower than the highest resolution attainable with AFM because relatively large areas (35 μm by 35 μm) were imaged at 512 pixels by 512 pixels to locate regions for NanoSIMS analysis.] Lower concentrations of both lipids were measured in regions where debris is visible in the AFM images, which indicated that the nonlipid particles were embedded in the bilayer and could have served as nucleation sites. In the fluid phase, the lipid composition was again well approximated by phase diagrams (27, 29); the ratio of 15N-DLPC to 13C18-DSPC was greater than 19:1, although tiny gel-phase domains scattered throughout the fluid phase may have been included in this value.

Fig. 4.

Quantitative analysis of the gel-phase domains shown in Fig. 3. Each gel-phase domain was divided into regions of 3 pixels by 3 pixels; specific regions of interest (ROIs) within each domain are shown on the NanoSIMS images. The graphs illustrate the amount of 13C18-DSPC (◯) and 15N-DLPC (♢) detected within each domain for the numbered ROI, where each data point represents a region of 3 pixels by 3 pixels within a domain, and the error bars represent the uncertainty calculated from counting statistics (fig. S4) (22). Statistically significant lateral variations in lipid composition were detected in domain C, where ROIs 5, 8, and 9 are in the vicinity of the fluid-phase subdomain that was detected by AFM (Fig. 3, domain C, circled region). NanoSIMS images were acquired with a pixel size of 100 nm by 100 nm and are smoothed over three pixels. Scale bar, 1 μm.

With the use of component-specific secondary ion imaging performed with the NanoSIMS, domains as small as ∼100 nm in diameter were successfully imaged within a phase-separated lipid membrane, the lipid composition within small regions of the bilayer were quantified, and heterogeneous lipid distributions within gel-phase domains were identified. This example of phase-separated membrane domains also demonstrates the advantage of combining quantitative lipid composition analysis performed by the NanoSIMS with multiple imaging modalities. Because supported lipid bilayers are amenable to isotopic substitution and freeze-drying, this approach can establish the distributions of multiple lipids and membrane-anchored proteins within more complex phase-separated supported membranes by incorporating a distinct stable isotope into each membrane component of interest and simultaneously imaging the secondary ions that distinguish each species. This approach can be extended to living cells by selectively incorporating stable isotopes into membrane components through the use of techniques to label lipid components in live cells (31, 32) as well as by isolating cell membranes with methods to detach intact membrane sheets from live cells (3336). In this way, quantitative information on multiple components within native cell membranes may be obtained with high lateral resolution.

Supporting Online Material

Materials and Methods

Figs. S1 to S4

Tables S1 to S4


References and Notes

View Abstract

Stay Connected to Science

Navigate This Article