Evaluating membrane structure by Laurdan imaging: Disruption of lipid packing by oxidized lipids

Abstract

Impact of different lipids on membrane structure/lipid order is critical for multiple biological processes. Laurdan microscopy provides a unique tool to assess this property in heterogeneous biological membranes. This review describes the general principles of the approach and its application in model membranes and cells. It also provides an in-depth discussion of the insights obtained using Laurdan microscopy to evaluate the differential effects of cholesterol, oxysterols and oxidized phospholipids on lipid packing of ordered and disordered domains in vascular endothelial cells.

1. Introduction

Lipid packing or order is a fundamental property of biological membranes that has a major impact on the function of membrane proteins, formation of lipid-protein signaling platforms and membrane recycling (Cooper, 1977; Yeagle, 1991). In earlier studies, changes in membrane microstructure were detected primarily by measuring the degree of movement of different fluorescent probes, which in most cases inversely correlates with an increase in lipid order. Currently, increasing number of studies turn to a fluorescent probe, Laurdan, that has spectral sensitivity to the membrane phase, its excitation/emission spectra depending on the gel-like/liquid crystalline-like phases of the phospholipids (Gaus, Zech, & Harder, 2006; Jay & Hamilton, 2017; Gunther, Malacrida, Jameson, Gratton, & Sanchez, 2021). This property allows imaging the local lipid order in model membranes (Dietrich et al., 2001) and cells (Gaus et al., 2003). Laurdan imaging can be used in both live and fixed cells (Gaus et al., 2003; Gaus, Le Lay, Balasubramanian, & Schwartz, 2006), as well as in intact tissues (Owen, Rentero, Magenau, Abu-Siniyeh, & Gaus, 2011). Most importantly, it allows to assess local membrane properties and obtain space-resolved maps of lipid order heterogeneity in live cells and tissues.

2. Basic principles of Laurdan fluorescence

LAURDAN (6-dodecanoyl-2-(dimethylamino) naphthalene), was synthesized and first characterized as an environmentally sensitive fluorescent probe by Gregorio Weber and Fay Farris in 1979 (Weber & Farris, 1979; Gunther et al., 2021). It is composed of naphthalene hydrophilic head linked to a hydrophobic tail composed of lauric fatty acid and it partitions to the hydrophilic-hydrophobic interface of the phospholipid bilayer with its lauric acid tail aligning with the phospholipid acyl chains (Fig. 1).

Fig. 1.

Chemical structure and incorporation of Laurdan in membrane. (A) Laurdan (6-dodecanoyl-2-dimethylaminonaphthalene, C24H35NO) has tswo aromatic rings and lauric fatty acid tail. When it is fluorescently excited, a charge separation between the amino and the carbonyl groups create a dipole moment sensitive to the presence of water dipoles in its environment. (B) Incorporation of Laurdan into the membrane bilayer (a single leaflet is shown). Purple indicates hydrophilic head, black line indicates hydrophobic acyl chain, and red line indicates Laurdan.

When it is fluorescently excited, a charge separation creates a strong dipole between the amino and the carbonyl groups. This strong dipole, in turn, aligns the surrounding dipole molecules, such as water, to restrict their free rotational motions. The energy dissipation for aligning water molecules or restricting the free rotational motions of water molecules results in a longer emission wavelength. Therefore, the spectra of the probe is sensitive to the water contents in the membrane. A more disordered membrane (liquid crystalline-like) should has more defects (or voids), which allow more water molecules to partition into the bilayer. Consistently, studies showed that Laurdan undergoes a red spectral shift during the membrane phase transition from gel-to liquid crystalline-like, which was attributed to dipole relaxation and the sensitivity to the polarity of the environment (Parasassi, De Stasio, Ravagnan, Rusch, & Gratton, 1991). Furthermore, the sensitivity of Laurdan to the lipid order of the membrane is a result of its sensitivity to the presence and mobility of water dipoles within the lipid bilayer and as more mobile water dipoles surround the Laurdan ring, it causes a shift in the emission spectrum. Thus, as water penetration into the bilayer decreases with an increase in the lipid order of the membrane, Laurdan emission spectrum shifts accordingly (Parasassi et al., 1991; Parasassi & Gratton, 1995). It is also important to note that Laurdan partitions equally to solid and liquid lipid phases, does not bind to specific fatty acids and is not soluble in water (Bagatolli, Sanchez, Hazlett, & Gratton, 2003). Another parameter that might be important to take into the account is the sub-cellular distribution of the Laurdan dye between the apical and basal membranes, which might affect the polarization properties.

Since Laurdan undergoes a 50-nm red shift in the emission maximum in polar vs no-polar environments, measuring fluorescence intensity at the two specific wavelengths provides a ratiometric method to assess the local membrane properties. This method defines the general polarization (GP) function to quantify the emission shift (Parasassi, De Stasio, d’Ubaldo, & Gratton, 1990; Parasassi et al., 1991). Specifically, the emission peaks of Laurdan immersed into lipids in the gel phase is 440nm and in liquid crystalline phase is 490nm. Thus, the GP ratio is defined as

$$\text{GP}=\left({\text{I}}_{440}-{\text{I}}_{490}\right)/\left({\text{I}}_{440}+{\text{I}}_{490}\right),$$

where I represents the emission intensities at the given wavelengths. The excitation is done at 340–360nm. At very low concentration, variations in Laurdan concentration do not affect the GP measurement. The GP values can vary from −1.0 to +1.0 with the more positive values indicating increase in lipid order. Experimentally though, the values typically range between −0.6 and +0.6 with the liquid phase yielding values between −0.3 and +0.3 and gel phase typically yielding values between 0.5 and 0.6.

In early studies, Laurdan GP values of lipid vesicles with different lipid compositions were measured using conventional fluorometer in a “cuvette” to determine the homogeneity and the presence of domains in lipid vesicles (Parasassi, Di Stefano, Loiero, Ravagnan, & Gratton, 1994; Parasassi & Gratton, 1995). Interestingly, Laurdan measurements led to a surprising observation about the effect of cholesterol on membrane fluidity: while it was considered to be well known that increase in membrane cholesterol decreases membrane fluidity, GP measurements suggested that the effect is more complex. Surprisingly, it was found that cholesterol has a “homogenizing” effect on the membrane, decreasing the fluidity of the liquid-crystalline phase, whereas also increasing the fluidity of the gel phase (Parasassi et al., 1994). Specifically, they found that the addition of 30mol%, cholesterol, a physiological level of cholesterol in biological membranes, decreases the fluctuation rate between the liquid and gel phases of the bilayer, while at the same time decreasing water concentration on hydrophobic-hydrophilic interface of due to higher lipid packing. These observations provided significant insights into the impact of cholesterol on the physical properties of lipid bilayers.

Clearly, the use of Laurdan is not limited to the studies of cholesterol effects on the lipid bilayers. Multiple studies applied this approach to evaluate effects of various lipids or other hydrophobic molecules that can be incorporated into the bilayers on lipid packing of the membrane. For example, using Laurdan, Amyloid-β peptide was found to alter membrane phase properties in astrocytes through its direct insertion and indirectly through the phospholipase A2 pathway (Hicks et al., 2008). More recently, Morandi et al. (2021), investigated the impact of plastic pollutants found in sea water that can enter marine organisms and consequently, the human food chain. Morandi et al found that incorporation of polystyrene, one of the widely used plastics, into liposomes resulted in a strong shift in Laurdan emission indicating a shift in the gel-to-liquid transition and lipid packing of the membrane suggesting that plastic nanopollutants may disrupt the function of biological membranes. Another very interesting recent study by Salvador-Castell, Brooks, Winter, Peters, and Oger (2021) used Laurdan fluorescence to evaluate the effects of unique phospholipids of archaeal membranes on the stability of the membrane under different temperatures to get insights into the ability of these membranes to keep their integrity under extreme temperatures. Thus, Laurdan GP provides a versatile approach to evaluate the physical properties of lipid bilayers.

Measuring Laurdan fluorescence in a cuvette, however, provides only the average behavior of a vesicle population without the information about spatial distribution of the domains.

The main constraint of using Laurdan probe in microscopy was limited by its rapid photobleaching, which made it non-reliable for conventional microscopy (Bagatolli et al., 2003).

3. Laurdan two photon imaging: Visualizing domains in membrane vesicles

Laurdan imaging was developed with the advance of the two-photon fluorescence microscopy, an approach in which a fluorescent probe absorbs two photons simultaneously with each photon contributing half of the required energy and thus shifting excitation to the wavelengths longer than the emission wavelengths. Typically, a probe is excited by two photons in near-infrared range, substantially reducing photobleaching and scattering. Most of the current Laurdan studies use two-photon microscopy.

One of the earliest studies to shift from “cuvette” to two-photon microscopy by Parasassi, Gratton, Yu, Wilson, and Levi (1997), suggested that while different domains co-exist in membrane vesicles, their size is smaller than the microscope resolution of 200nm, an important observation in light of subsequent debate about the dimensions of lipid ordered domains in biological membranes. They also found that cholesterol induces heterogeneity in the gel phase: in contrast to previous observations that cholesterol increases the fluidity of the gel phase, Parasassi et al. (1997), did not see a disordering effect of cholesterol and instead observed a general shift of the GP values towards the more ordered state. A more recent study by Aguilar et al. (2012) showed a clear spatial separation of Laurdan GP values in ordered vs. disordered domains in giant unilamellar vesicles (GUVs) containing cholesterol in temperature below the transition temperature. This separation was not observed above the transition temperatures of 37–38°C and the membrane becomes more homogenous. The authors concluded that an increase in the cholesterol content in the vesicles results in the stabilization of a liquid-ordered (Lo) phase, an intermediate between gel (So) and liquid-crystalline (Ld) phases.

Sanchez, Tricerri, and Gratton (2007) applied Laurdan GP imaging of GUVs to study the kinetics of cholesterol removal by high density lipoproteins (HDL) from different membrane domains (pools). As expected, the GP values decreased as a function of methyl-β-cyclodextrin (MβCD), a well-known cholesterol-depleting agent. Imaging of the GUVs show that at temperatures below the phase transition (30°C), small islands/domains of membrane with high GP values floating in a larger membrane pool with lower GP. These domains vary significantly in size among the vesicles. Above 30°C, the two phases merged, and separate domains were not visible. Most interestingly, exposing the GUVs containing cholesterol to HDL particles resulted in a decrease in the GP values in the liquid but not from the ordered state suggesting that HDL particles selectively removes cholesterol from the liquid phase of the membrane. A similar approach was used by Sanchez, Gunther, Tricerri, and Gratton (2011) to study the kinetics and the pool/domain specificity of cholesterol removal by MβCD. Consistent with the observations described above for HDL particles, MβCD was also found to preferentially remove cholesterol from the liquid/disordered state in spite of the fact that it is the ordered phase that is more rich in cholesterol. Thus, Laurdan imaging allows distinguishing between cholesterol kinetics of two separate membrane pools.

These observations have major implications both for understanding the nature of cholesterol interactions with the surrounding phospholipids and for the use of MβCD to assess the physiological roles of cholesterol rich membrane domains. A molecular mechanism proposed for the preferential removal of cholesterol from the disordered phase was that the kinks in the structure of unsaturated phospholipids present in the disordered state prevent cholesterol from penetrating deeper into the bilayer and thus making it more accessible to the acceptor molecules. In contrast, increased hydrophobic thickness and packing of the ordered phase impedes its removal by MβCD or other acceptors. It is most important to note that based on these observations, MβCD should not be used as a selective tool for the disruption of the ordered domains. In fact, not only MβCD removes cholesterol from the disordered domains, it appears to preferentially target this phase.

4. Laurdan two photon imaging: Visualizing membrane domains in living cells

Laurdan imaging of living cells was pioneered by Gaus et al. (2003) and Gaus, Le Lay, et al. (2006). First, Gaus et al. (2003) showed that Laurdan-stained macrophages have highly heterogenous distributions of the GP values with a punctate distribution of areas of low and high GPs corresponding to ordered “gel” and disordered “fluid phase” domains throughout the entire surface of the cells. The phase separation is not observed but rather the cell surfaces appeared to have a continuum of GP values varying from low to high. Multiple studies including from our group showed a similar pattern of highly heterogenous punctate staining through a cell with the ordered domains typically concentrated at cell edges (Ayee, LeMaster et al. 2017; Levitan and Shentu 2011; Shentu et al., 2010) (see a typical image of GP distribution in Fig. 2A).

Fig. 2.

Typical Laurdan image and the distribution of GP values in a living cell. (Left) A typical GP image of an endothelial cells (human aortic endothelium) shown in pseudo-color, as indicated in the gradient bar: the disordered regions are shown in blue and green and the ordered regions in orange and yellow. (Right) GP histograms (navy dots) fitted by two-Gaussian distributions, with right-shifted curve (yellow) representing ordered domains and left-shifted curve (blue) representing fluid domains. The sum of the Gaussians is shown as a solid, mauve curve. GP distribution is obtained from the region −0.35 to + 0.8. Adapted from Ayee, M. A., LeMaster, E., Shentu, T. P., Singh, D. K., Barbera, N., Soni, D., et al. (2017). Molecular-scale biophysical modulation of an endothelial membrane by oxidized phospholipids. Biophysical Journal 112(2), 325–338.

The areas of the high GP puncta were also very heterogenous in size varying from hundreds of nanometers to micrometers. This is larger than the estimated size of lipid rafts, cholesterol-rich membrane domains that play major roles in cell signaling (e.g., Lingwood and Simons, 2010; Sonnino and Prinetti, 2013). Indeed, the size, the exact composition, the dynamics of lipid rafts have been investigated in numerous studies with different approaches but still remain controversial. A working definition of a lipid raft was adopted during the Keystone Symposium on Lipid Rafts and Cell Function in 2006. The question “What is a raft?” was answered with the following statement: “Membrane rafts are small (10–200nm), heterogeneous, highly dynamic, sterol- and sphingolipid-enriched do-mains that compartmentalize cellular processes. Small rafts can sometimes be stabilized to form larger platforms through protein-protein and protein-lipid interactions.” (Pike, 2006). This is below the resolution of the two-photon microscopy at 800nm (~180nm) and thus, the puncta are unlikely to represent individual rafts. Instead, they may represent areas where rafts coalesce into larger domains or areas with a higher fraction of ordered domains. Notably, a subsequent study by Gaus, Le Lay et al. (2006) established that in endothelial cells and in embryonic fibroblasts, the puncta with high GP values appear to co-localize with the markers of focal adhesions, phosphorylated FAK (pFAK) and phosphorylated caveolin-1 (pYCav1), whereas Tranferrin, a marker of disordered domains was shown to co-localize with low GP puncta.

Typically, GP histograms for single cells show double peak distributions with the two peaks overlapping to different degrees depending on the degree of the separation between the areas with low and high GP values (see a typical histogram in Fig. 2B). The distributions are fitted by a sum of two Gaussian curves, which are interpreted to represent ordered (peak with higher GP values) and fluid (lower GP values) membrane domains. It is important to note that the distribution of the GP values is continuous and the fit to a sum of two Gaussian is an approximation. The specific positions of the two peaks represent the degree of lipid order for each population of the domains and shift to higher or lower values depending on the experimental conditions.

As described in more detail below, increasing or decreasing membrane cholesterol also shifts the peaks appropriately. The areas under the two curves represent the coverage of the cells surface by ordered or disordered domains. Both values vary between different cell types or other parameters. For example, in macrophages, Gaus, Gratton et al. (2003) found the peaks corresponding to the two phases at 0.14GP and at 0.55GP for the fluid and the gel phase respectively. The areas covered by ordered and disordered domains were comparable with disordered area being somewhat larger. In endothelial cells (porcine aorta) (Gaus, Le Lay, et al., 2006), found that the separation between the two peaks was a little less pronounced (peaks at 0.17 and 0.51, respectively) and the % coverage was significantly different than what was found in the previous study with macrophages. In contrast to macrophages, where % coverage of the membrane by the ordered domains was only slightly smaller than % coverage by the disordered domains, in endothelial cells, the difference was clearly very significant with the ordered domains being a relatively minor component of the membrane. Our later studies (Ayee, LeMaster, et al., 2017; Shentu et al., 2010) also showed somewhat smaller separation between the peaks in human aortic endothelial cells but still two clear peaks with the disordered area being higher than the ordered one. Thus, this method constitutes a unique approach to assess the impact of different experimental conditions on distinct membrane domains in living cells.

5. Impact of cholesterol depletion on Laurdan GP values

A decrease in membrane cholesterol is well known to have profound effects on membrane fluidity and lipid packing. Numerous studies, using a variety of techniques, showed that cholesterol removal results in increased membrane fluidity and decreased lipid packing (Zidovetzki and Levitan, 2007). As such, testing the impact of cholesterol depletion on Laurdan signal serves a valuable control to validate this approach. Indeed, as expected cholesterol depletion using MβCD, the most commonly used tool to remove membrane cholesterol from biological membranes (Zidovetzki and Levitan, 2007), was shown to decrease the GP values in several preparations indicating that the probe is sensitive to experimental perturbations of membrane cholesterol in living cells (e.g., Gaus, Gratton, et al., 2003; Shentu et al., 2010).

A much more controversial question is what is the dynamics of cholesterol depletion from ordered vs. disordered domains. This question is critical for the interpretation of studies that use MβCD as a tool to disrupt lipid rafts, a widely used strategy that frequently assumes that MβCD specifically affects lipid rafts and disregards its possible effect on the disordered domains. This assumption was previously questioned in studies that used biochemical techniques to separate between low-density (cholesterol-rich) and high density (cholesterol-poor) membrane fractions and which showed that cholesterol can be depleted from both pools. The degree to which MβCD takes cholesterol from one pool vs. another varied between different studies, as described in detail in our earlier review (Zidovetzki and Levitan, 2007). Clearly, though, biochemical separation of membrane domains may introduce multiple artifacts for their lipid composition and is not the most reliable approach to address this question.

In this regard, Laurdan imaging that discriminates between ordered and disordered domains in living cells presents a much more powerful tool to establish whether MβCD has preferential effects on ordered vs disordered domains. The results, however, reported by different studies are still controversial. As mentioned above, studies in GUVs suggested that MβCD actually preferentially removes cholesterol from fluid rather than ordered domains (Sanchez, Gunther, Tricerri, & Gratton, 2011). In living cells, the results are more complex and more controversial. Gaus, Gratton, et al. (2003) showed that in macrophages, cholesterol depletion with MβCD resulted in a significant decrease in the cell coverage by the ordered domains but the shifts in the GP values were not analyzed in detail.

Our study, Shentu et al. (2010) provided a detailed quantitative analysis of the impact of MβCD on the distribution of the GP values in aortic endothelial cells and established that both ordered and disordered domains are affected. As shown previously, punctate distribution of ordered and disordered domains was observed throughout the cell surface, with ordered domains concentrating on the cells periphery (Fig. 3). Exposure to MβCD resulted in a shift of the GP values of both types of the domains to less ordered and more fluid membrane structure and the magnitudes of both shifts were comparable (Fig. 3, Table 1). In contrast, the relative coverage of the cell surface by the two types of the domains remained unchanged with approximately 30% of the cell surface covered by the ordered and 70% by the disordered domains. Clearly, these observations further challenge the general belief that MβCD-induced cholesterol depletion is a specific tool for the disruption of lipid rafts. There is no doubt that exposing cells to MβCD results in a significant decrease in membrane cholesterol but it appears that a decrease occurs both in rafts and in non-rafts fraction. The significance of this observation is in the notion that at least in this case, MβCD-induced cholesterol depletion does not eliminate lipid rafts but instead alters the lipid order within the domains making them less ordered and that a similar change develops in the disordered domains in parallel. We also found that a decrease in the GP values or disruption of lipid packing is associated with an increase in the elastic modulus of endothelial cells, an effect dependent on the activation of RhoA/ROCK signaling cascade (Oh et al., 2016). The mechanism underlying the connection between lipid packing and activation of RhoA is currently under the investigation.

Fig. 3.

MβCD-induced cholesterol depletion shifts Laurdan GP to more fluid values in both ordered and disordered domains. (A) Typical GP images for control cells and MβCD-treated cells. All images are shown in pseudocolor, as described above (green/blue—disordered, yellow/orange—ordered). (B) Zoomed images showing the puncta in more details. (C) GP histograms showing the experimental values fitted with a sum of two Gaussians, as described above. Adapted from Shentu, T. P., Titushkin, I., Singh, D. K., Gooch, K. J., Subbaiah, P. V., Cho, M., et al. (2010). oxLDL-induced decrease in lipid order of membrane domains is inversely correlated with endothelial stiffness and network formation. American Journal of Physiology Cell Physiology 299(2), C218–229.

Table 1.

Peak GP values and % coverage for control and MβCD-treated cells.

	Control	MβCD
GP value
Peak 1	0.09 ± 0.01	0.03 ± 0.01*
Peak 2	−0.06 ± 0.01	−0.12 ± 0.01*
General GP	−0.02 ± 0.01	−0.08 ± 0.01*
% Coverage
Peak 1	31 ± 4	33 ± 3
Peak 2	69 ± 4	67 ± 3

Adapted from Shentu, T. P., Titushkin, I., Singh, D. K., Gooch, K. J., Subbaiah, P. V., Cho, M., et al. (2010). oxLDL-induced decrease in lipid order of membrane domains is inversely correlated with endothelial stiffness and network formation. American Journal of Physiology Cell Physiology 299(2), C218–229.

Interestingly, the loss of caveolin-1, the structural component of caveolae, known as cholesterol-rich membrane invaginations, also led to significant shifts in the GP values in both ordered and disordered domains to more fluid values in embryonic fibroblasts (Gaus, Le Lay, et al., 2006). A lack of caveolin-1 also resulted in a significant decrease in membrane coverage by the ordered domains and both effects were reversible upon reconstitution of caveolin-1. Since caveolin-1 is known to regulate cholesterol trafficking to the membrane (Frank, Woodman, et al., 2003), the observed shift in the GP values could be the result of lower cholesterol content of the membrane in Cav-1 knock-out cells. It is also possible that Cav-1 itself may play a role in regulating lipid order of the membrane. It is also interesting to note that the GP values differ significantly between the different studies, which might be due to differences in the lipid composition in different cellular sub-types or culture conditions. It would be interesting to explore the heterogeneity of membrane fluidity on the single cell basis.

A new Laurdan imaging approach was developed more recently to allow discriminating between changes in membrane cholesterol and changes in membrane fluidity (Bonaventura, Barcellona, et al., 2014; Golfetto, Hinde, et al., 2013). This approach is based on the dual dependence of Laurdan fluorescence on the polarity of the environment and on dynamics of dipolar relaxation and quantifies the decay of the Laurdan fluorescence using phasor analysis. Done in living cells, the phasor approach allows monitoring dynamic changes in local membrane fluidity/lipid packing during physiological processes, such as cell migration, independently of changes in membrane cholesterol. Notably, using time-resolved Laurdan fluorescence analysis of live HeLa cells, Ma, Benda, et al. (2018) showed that the time-resolved GP data is consistent with 20–30mol% of cholesterol and suggested that most of the membrane heterogeneity comes from the local variations of membrane cholesterol.

6. Opposite effects of low-density lipoproteins (LDL) and oxidized low-density lipoproteins (oxLDL) on the lipid order of endothelial cells

The major cholesterol carrier in blood is low-density lipoprotein (LDL), a lipoprotein particle that emulsificates cholesterol and other lipids allowing their transport in blood stream (Baigent, Blackwell, et al. 2010; Castelli, Anderson, et al. 1992). It is well known that elevated level blood LDL correlates the development of fatty streaks in the vessel walls of the arteries, which progress to atherosclerotic lesions and clogging of the arteries. LDL particles are internalized by cells via LDL receptor-mediated endocytotic pathway, which is then degraded and free unesterified cholesterol incorporated into the plasma membrane. Indeed, numerous studies showed that plasma hypercholesterolemia leads to significant cholesterol loading of vascular tissues in vivo, particularly macrophages that become cholesterol-loaded foam cells. However, a controversy arose about whether LDL is directly responsible for cholesterol loading of cells or whether the loading requires LDL to be oxidized. The source of this controversy was the discovery that macrophages exposed to high levels of LDL in vitro, do not show significant cholesterol loading. It was then proposed that a lack of cholesterol loading in macrophages exposed to LDL is due to the downregulation of the LDL receptor and that the loading is the result of exposing the cells to oxidative modifications of LDL, oxLDL (Steinberg, 2002). Indeed, oxLDL, is also found in blood concurrently with LDL although at much lower levels was shows to be highly pro-inflammatory and cytotoxic further supporting the idea that it is a major proatherogenic form of LDL (Levitan, Volkov, et al., 2010).

We have studied the effects of LDL and oxLDL on membrane cholesterol and lipid packing of endothelial cells using Laurdan two-photon microscopy imaging and found that the two types of lipoproteins have opposite effects on lipid packing: while exposure to LDL resulted in increased lipid order of the membrane, exposure to oxLDL resulted in the fluidization of the membrane.

LDL:

Exposure of human aortic endothelial cells to clinically-relevant high levels of LDL resulted in a relatively mild but significant cholesterol loading (~20% increase in free cholesterol) and a pronounced shift in Laurdan GP values towards more ordered lipid structure (Bogachkov, Chen, et al., 2020). Fig. 4 shows typical Laurdan images of endothelial cells (HAECs) exposed to normal (50mg/dL) or hypercholesterolemic (250mg/dL) levels of LDL.

Fig. 4.

Exposure to high levels of LDL increases lipid packing of endothelial cells in both ordered and disordered domains. Left: typical GP images for endothelial cells exposed to 50 vs 250mg/dL (pseudocolor, as described above). Right: GP histograms showing the experimental values fitted with a sum of two Gaussians, as described above. Adapted from Bogachkov, Y. Y., Chen, L., Le Master, E., Fancher, I. S., Zhao, Y., Aguilar, V., et al. (2020). LDL induces cholesterol loading and inhibits endothelial proliferation and angiogenesis in Matrigels: Correlation with impaired angiogenesis during wound healing. American Journal of Physiology. Cell Physiology 318(4), C762–C776.

As described above, the images are pseudo-colored with higher GP values shown as red and lower as yellow. The shift to higher GP values in cells exposed to hypercholesterolemic LDL level is apparent both in the representative images and in the corresponding histograms of the GP values. Notably, the rightward shift is observed in both Gaussians, the one corresponding to the fluid domains and the one corresponding to the ordered domains indicating that both domains shift to the more ordered structure. The relative abundance of the two types of domains did not change, however, suggesting that there is no increase in the membrane coverage by the ordered domains. This is important because it allows discriminating between two conceptually different models for the impact of cholesterol on membrane domains in living cells: one model is to suggest that increase in membrane cholesterol results in increased abundance of lipid rafts/ordered domains, whereas the alternative is that the main changes occur within the domains. Our observations clearly support the second possibility.

OxLDL:

We found that in contrast to LDL, exposing endothelial cells to oxLDL does not load the cells with cholesterol but results in changes in the elastic modulus of the cells similar to those induced by cholesterol depletion, rather than cholesterol enrichment (Byfield, Tikku, et al., 2006). Furthermore, cholesterol depletion-like changes in the cellular elastic modulus in oxLDL treated cells were not accompanied with a decrease in cellular cholesterol. To understand better this phenomenon, we analyzed the impact of oxLDL on lipid packing using Laurdan and found that exposure to oxLDL shifts the GP values to lower values, an effect that is indeed similar to the effect of cholesterol depletion describe above (Shentu et al., 2010). Furthermore, providing surplus of cholesterol to oxLDL treated cells by exposing them to MβCD saturated with cholesterol reverses the effect of oxLDL on lipid packing. The reversal is particularly apparent in a strong shift of the ordered domains towards higher GP values in cells exposed to oxLDL and then to MβCD-cholesterol, as compared to oxLDL alone (Shentu et al., 2010).

These observations indicate that, in contrast to previous belief, exposure to oxLDL is not a more efficient method to load cell with cholesterol but has a completely different and opposite effect on the biophysical properties of the lipid bilayer. Our next question, therefore, was to investigate the mechanism responsible for this dichotomy. We found that this phenomenon can be attributed to the incorporation of oxidized lipids (Fig. 5).

Fig. 5.

OxLDL results in the fluidization of membrane domains. (A) Typical GP images for control cells and MβCD-treated cells. All images are shown in pseudocolor, as described above (green/blue—disordered, yellow/orange—ordered). (B) Zoomed images showing the puncta in more details. (C) GP histograms showing the experimental values fitted with a sum of two Gaussians, as described above. Adapted from Shentu, T. P., Titushkin, I., Singh, D. K., Gooch, K. J., Subbaiah, P. V., Cho, M., et al. (2010). oxLDL-induced decrease in lipid order of membrane domains is inversely correlated with endothelial stiffness and network formation. American Journal of Physiology Cell Physiology 299(2), C218–229.

7. Disruption of lipid packing of endothelial membrane by oxidized lipids: Oxysterols and oxidized phospholipids

7.1. Oxysterols

Earlier studies showed that in contrast to cholesterol, incorporation of oxysterols into liposomes disrupt the formation of the ordered domains (Wang, Megha, & London, 2004). We hypothesized, therefore, that a decrease in lipid packing we observed in endothelial cells exposed to oxLDL could be due to the incorporation of oxidized lipids. To test this idea, we exposed cells to 7ketocholesterol, one of the major oxysterol components of oxLDL (Garcia-Cruset, Carpenter, Guardiola, Stein, & Mitchinson, 2001; Oh et al., 2016). Our observations showed that exposure to 7ketocholesterol resulted in a significant shift in Laurdan GP to more fluid values, an effect similar to that of oxLDL and cholesterol depletion (Fig. 6, (Shentu et al., 2010)). Interestingly, this effect was observed mostly in the ordered domains. Furthermore, a decrease in lipid packing was also observed in cells exposed to anderstenol, another oxysterol that was shown previously to disrupt membrane the formation of lipid ordered domains in liposomes (Xu and London, 2000).

Fig. 6.

Fluidization of membrane domains by 7ketocholesterol. Top panels: typical GP images for a control and 7ketocholesterol-treated cells. Botttom panels: GP histograms. Adapted from Shentu, T. P., Titushkin, I., Singh, D. K., Gooch, K. J., Subbaiah, P. V., Cho, M., et al. (2010). oxLDL-induced decrease in lipid order of membrane domains is inversely correlated with endothelial stiffness and network formation. American Journal of Physiology Cell Physiology 299(2), C218–229.

More recently, we provided a theoretical explanation for the differential effects of cholesterol and 7ketocholesterol on lipid order of the membrane bilayer (Ayee & Levitan, 2021). Briefly, we used Molecular Dynamics simulation to analyze structural and dynamic changes in phospholipid bilayers that result from the incorporation of physiological levels of cholesterol or 7ketocholesterol, the amounts of both sterols were based on mass spectrometry analysis of cells exposed to LDL or oxLDL respectively. Our analysis showed that due to the different tilt in the orientation of cholesterol and 7ketocholesterol within the phospholipid bilayer, the presence of cholesterol increases the order parameter of all bonds in the lipid tails of the three major phospholipids, POPPC (1-palmitoyl-2-oleoyl-glycero-3-phosphocholine), DPPC (1,2-dipalmitoyl-sn-glycero-3-phosphocholine) and SM (sphingomyelin). In contrast, the presence of 7ketocholesterol decreased the order parameter of the lipid tails of POPC, DPPC and SM lipid species.

7.2. Oxidized phospholipids

We also found that similarly to oxysterols, incorporation of bio-active oxidized phospholipids, specifically, oxPAPC (1-palmitoyl-2-arachidonoyl-sn-glycero-phosphocholine), a component of the minimally oxidized LDL, a highly reactive oxLDL species, also disrupts lipid packing in endothelial membranes, as detected by Laurdan microscopy (Ayee et al., 2017). The effect was observed for two oxPAPC species, POPC (5-oxovaleroyl)-sn-glycero-3-phosphocholine and PGPC (1-palmitoyl-2-glutaroyl-snglycero-3-phosphocholine), both with truncated tails generating a tilt in their orientations within the bilayer.

Thus, Laurdan microscopy provides a unique tool to clearly discriminate between the effects of different lipid species on lipid order/packing of biological membranes.

8. Challenges and new directions

8.1. Sensitivity to fading and development of new probes

The first limitation that discussed above is the rapid photo-bleaching of Laurdan dye, which makes it unsuitable for standard confocal and total internal reflection fluorescence (TIRF) systems, limiting its current use to multi-photon microscopy. To resolve this limitation, several new membrane probes that work similarly to Laurdan but are more stable have been developed. Specifically, Di-4-ANEPPDHQ, a voltage-sensitive dye that was first developed to analyze neural networks (Obaid, Loew, Wuskell, & Salzberg, 2004), was demonstrated to be sensitive to lipid packing and capable to identifying cholesterol-rich domains in model membranes (Jin, Millard, Wuskell, Clark, & Loew, 2005). Similarly to Laurdan, Di-4-ANEPPDHQ, has a spectral shift when incorporated in ordered vs. disordered domains (Jin et al., 2006). Di-4-ANEPPDHQ is more photostable than Laurdan and was successfully used in confocal and TIRF microscopy (Owen, Rentero, Magenau, Abu-Siniyeh, & Gaus, 2012). Interestingly, it was also shown that the GP values generated using Laurdan are more influenced by the temperature that those generated using Di-4-ANEPPDHQ, which were shown to be more sensitive to the cholesterol content of the membrane (Amaro, Reina, Hof, Eggeling, & Sezgin, 2017). Another lipid-packing sensitive dye that was shown greater photostability than original Laurdan dye is M-Laurdan a close derivative of Laurdan but with less sensitivity to photobleaching (Mazeres, Joly, Lopez, & Tardin, 2014). Moreover, an interesting modification to Laurdan dye was the addition of a membrane anchor that stabilizes it in the outer leaflet of the plasma membrane and prevents its internalization and labeling of the internal membrane (Danylchuk, Sezgin, Chabert, & Klymchenko, 2020). Development of these new dyes also opens the door to develop super-resolution applications to overcome the spatial resolution limitations. Indeed, the resolution of multi-photon or confocal microscopy is not sufficient for imaging of individual rafts unless the rafts coalesce. To this end, the attention is turned to super-resolution fluorescence microcopy that increases the lateral resolution from 200nm diffraction limit of conventional microscopy to 5–30nm range, as reviewed by (Owen and Gaus 2013). It still needs to be established whether Laurdan or its derivatives can be suitable for the super-resolution imaging. It is also important to note, however, that these environmentally-sensitive dyes were also shown to have significant effects on membrane properties, particularly at higher concentrations (Suhaj, Gowland, Bonini, Owen, & Lorenz, 2020).

8.2. Laurdan imaging in vivo

Another important advance is applications of Laurdan or its derivatives not only to individual cells but also to tissues, organs or even organisms to study changes in lipid packing in intact tissues. Laurdan imaging was successfully implemented to visual lipid packing in model organisms, such as zebra fish and in living vertebrate embryos (Owen, Magenau, Majumdar, & Gaus, 2010; Owen, Rentero, Magenau, Abu-Siniyeh, & Gaus, 2012). Development of another Laurdan derivative, C-Laurdan advances this possibility by increased sensitivity to membrane polarity and yielding brighter images in multi-photon microscopy (Kim et al., 2007). This latter approach was used to image ordered domains in brain slices at the depth of 100–250μm in live tissue (Kim et al., 2008). Developing further applications for imaging lipid order of membrane in intact tissues will provide major insights into the impact of multiple pathological conditions on cell function.