Detecting membrane lipid microdomains by two-photon fluorescence microscopy

This article discusses our ability to obtain, using two-photon excitation, microscopy images of the generalized polarization (GP) of the lipid probe 2-dimethylamino-6-lauroylnaphthalene (LAURDAN) in phospholipid vesicles and in natural membranes. The images show different distribution of the GP value depending on membrane composition. The use of linearly polarized excitation allowed the attribution of the GP heterogeneity to coexisting membrane domains of different dynamic properties. Based on the photoselection operated by the excitation polarization, we propose a model to explain our results and to estimate the domains' dimension.

excitation, such as the reduced sample photodamage and probe photobleaching, and a better background rejection compared to one-photon excitation, allow the prolonged observation and the study of samples that are difficult to measure using one-photon fluorescence microscopy [2,3], The spectroscopic properties of the emitting fluorophores can be characterized using two-photon microscopy [4].
This article discusses our ability to obtain, using two-photon excitation, microscopy images of the generalized polarization (GP) of the lipid probe 2-dimethylamino-6-lauroylnaphthalene (LAURDAN) in phospholipid vesicles and in natural membranes. The images show different distribution of the GP value depending on membrane composition. The use of linearly polarized excitation allowed the attribution of the GP heterogeneity to coexisting membrane domains of different dynamic properties. Based on the photoselection operated by the excitation polarization, we propose a model to explain our results and to estimate the domains' dimension.

Overview
Our interest in the study of the dynamic properties of lipids in membranes has been focused on establishing a methodology for the detection of lipid microdomains. In the bilayer aggregation form of lipids, microdomains are intended as locally separated areas with distinct dynamic properties. In biological systems, microdomains with specific properties may serve to separate com-partments with specific functions and to modulate the activity and the delivery of information during the physiological life of thecell [5,6]. Alteration of the topography and of the properties of such domains may also have a role in the development of cell malfunction and pathologies.
We have previously characterized the fluorescence response of the LAURDAN polarity lipid probe using "cuvette" spectroscopy, in both model and natural membranes [7,81. LAURDAN possesses a steady-state sensitivity to polarity changes in its lipid environment, showing a red spectral shift directly related to the increase in the polarity. The GP function has been used to quantify the spectral shift of LAURDAN: (1) where Z,and ZR are the steady-state intensities in the blue (440 nm) and the red (490 nm) part of the emission spectrum, respectively [7]. These blue and red spectral regions correspond to the maximum emission of LAURDAN in phospholipids in the gel and in the liquid-crystalline phase, respectively.
Calculation of the GP in the gel and in the liquid-clystalline phase following Eq.
(1) gives high and low values, respectively. The polarity ohanges in the different phospholipid phases originate from variations of the water content around the probe fluorescent moiety, at the level of the gjycerol backbone. In pure phospholipid bilayers of unknown composition, the dependence of the GP value on the excitation wavelength gives information on the coexistence of phase microdomains.
Similar to the fluorescence polarization, the GP possesses an additive property, so that the measured GP value represents an average of the GP value of each emitting LAURDAN molecule. The general limits of the GP values for the gel and for the liquid crystalline phase have been determined, so that for samples of unknown composition, once the coexistence of domains has been ascertained, the fraction of the membrane domains can be quantitatively resolved. More difficult is the detection of coexisting domains of different dynamic properties when the bilayer is composed of unknown lipids, such as in the case of mammalian cell membranes. Moreover, the presence of cholesterol induces the formation of phase states with properties different from those of the pure phospholipid bilayer. Dynamic properties between those of the gel and of the liquid-crystalline phase have been measured when cholesterol is present. Phase diagrams have been constructed and phases such as the solid-ordered, liquid-ordered, and liquid-disordered have been defined [9]. LAURDAN GP "cuvette" studies on mammalian cell membranes did not detect coexisting domains [lo]. The next step for the investigation on coexisting domains in the membranes and, in particular, in cell membranes, was to use fluorescence microscopy. LAURDAN absorption occurs in the W region, between 320 nm and 420 nm. Using one-photon excitation, the very fast fading of LAURDAIi fluorescence due to photobleaching hindered any detailed spectroscopic study of its emission. However, by using two-photon excitation, prolonged observations on LAURDAN-labeled cell membranes were possible [ 111.
Images of LAURDAN GP values were obtained, revealing different GP values in the various cell membrane compartments

Culture, Labeling, and Mounting of OK Cells
Opossum kidney renal tubular epithelial cells (OK cells) [13] were grown in Dulbecco's modified Eagle's high-glucose medium (DMEM) containing 10% fetal calf serum, 100 IU/ml penicillin G, and 0.1 mg/ml streptomycin. For microscopy measurements, cells were seeded on dishes, containing microscope coverslips, for 2-4 hours. During this time, cells adhered to the coverslips, maintaining their round shape, which we preferred for the experiments with polarized excitation. LAURDAN labeling was performed by directly adding 1 pl of a 2 mM probe solution in DMSO per milliliter of the growth medium in the culture dishes. Cells were incubated for 30 min in the dark, then gently washed with fresh medium. The coverslips were mounted on the hanging drop microscope slides in fresh medium.

Brush Border and Basolateral
Membrane Preparation, Labeling, and Mounting Apical brush border (BBM) and basolateral (BLM) membranes from the rat renal cortex were simultaneously isolated by differential centrifugation, magnesium precipitation, and discontinuous sucrose gradient methods [14, 151. Purified membranes were diluted to a concentration of 0.1 mg protein/ml. LAURDAN labeling was performed by adding 1 pl of the 2 mM probe solution in DMSO per milliliter of the membrane sample. The sample was vortexed for 30 s, then a drop was evaporated on a coverslip with a nitrogen stream. The coverslip was mounted on the flat microscope slide with a drop of distilled water.

Two-Photon Microscopy Measurements
A titanium (Ti):sapphire laser (Mira 900, Coherent, Palo Alto, CA) pumped by an argon ion laser (Innova 3 10, Coherent) was used as the excitation light source. The laser wavelength was tuned at 770 nm. The laser light was guided by a galvanometer-driven x-y scanner (Cambridge Technology, Watertown, MA). The scanning rate was 26 s per frame (256 x 256 pixels). The incident laser power on the sample was about 2 mW. A quarter-wave plate (CVI Laser Corporation, Albuquerque, NM) was placed after the polarizer to change the polarization of the laser light from linear to circular for polarization-independent excitation. To change the laser polarization, apolarizer was placed after the quarter-wave plate. Two optical bandpass filters (Ealing Electro Optics, New Englander Industrial Park, Holliston, MA) were used to collect the fluorescence in the blue (440 nm) and in the red (490 nm) regions of the LAURDAN emission spectrum. The two filters were exchanged each time a full frame was scanned. To compensate for photobleaching, three successive pictures in a sequence of red-blue-red were collected, and the two red frames were averaged. Then the GP was calculated following Eq. 1 [ll]. A miniature photomultiplier (R5600-P, Hamamatsn, Bridgewater, NJ) amplified through a AD6 discriminator (Pacific, Concord, CA) was used for light detection in the photon-counting mode. The fluorescence intensity was collected using a custom-made data acquisition card located in a personal computer. This heterogeneity can be well appreciated in the plots of the histograms of the GP value (Fig. 2).

Natural Membranes
The GP images obtained from OK cells show different GP values in the different cellular compartments (Fig. 4). In particular, the plasma membrane shows high GP values. For the round shape of the plasma membrane, an excitation photoselection can beperformed, and the, high GP values are displayed along the direction parallel to the excitation polarization. By selectively plotting high and low GP values (Fig. 5), the low GPpixels mainly appeared in the direction perpendicular to the excitation polarization [ Figs. 4 (a-d)].
The GP images obtained from the BBM and the BLM samples show relatively high values, with no photoselection effect (Fig. 4). Compared to the BLM, the BBM samples show higher GP values. A selective plotting of low GP values of both the BBM and the BLM images shows the localization of these low values at the border of the membranes (Fig. 5).

Discussion
We have obtained LAURDAN GP images in the phospholipid vesicles and in the natural membranes using two-photon excitation. These measurements have been made possible for the reduced overall sample photobleaching when using two-photon excitation. Im-Images of selected pixels with low GP values from Fig. 1. The scale bar in all images is 5 pm. The plotted GP values are less than -0.25 for DOPC (a); less than 0.10 for DLPC (h); less than 0.40 for DPPC (c); and less than 0.10 for the DOPC-DPPC mixture (d).
ages of LAURDAN GP previously reported in mouse fibroblasts [ 111, as well as the GP images of the OK cell presented in this work, showed a heterogeneous distribution of the GP values. Both in fibroblasts and in OK cells, higher GP values were observed in the plasma membranes and in inner membranes corresponding to the Golgi apparatus. Lower GP values were observed in the complex membrane cytoplasmic network, especially around the nucleus. These different GP values are consistent with the lipid composition of cell membrane compartments [17]. By the use of polarized excitation, we could ascertain that the GP heterogeneity is due to real membrane heterogeneity, thus excluding possible artifacts due to measurements noise. We also attempted an evaluation of the size of the domains in the membrane, based on the spatial separation of the different GP values.
Our rationale for using polarized excitation in imaging LAURDAN GP in membranes is based on our previous characterization of LAURDAN spectroscopy. We determined that the orientation of the transition dipole of the fluorescent moiety of LAURDAN was parallel to the phospholipid acyl chains [8]. Moreover, high values of both LAURDAN polarization and GP are observed in the gel phase, where the phospholipid molecular motion is hin-

Seplember/Ottober 1999 ME ENGINEERING IN MEDICINE AND BIOLOGY
dered and the water content in the bilayer is low. In the liquid-crystalline phase, the molecular dynamics of the phospholipid is faster (i.e., more water penetrates the bilayer), and consequently, LAURDAN GP is also lower [7,8]. By considering the round shape of the vesicles and the axial resolution of the two-photon excitation that corresponds to about 600 nm with our microscope, we associated the photoselection of the polarized excitation with the value of LAURDAN GP. The actual observation of this association allows the attribution of the GP heterogeneity to the heterogeneity of the membrane.

Excitation Photoselection is Associated With High GP Values
The possible orientations of the phospholipids in the round vesicle, with respect to the excitation polarization axis, are schematically represented in Fig. 6. The microscopic images correspond to a cross-section through the structure of the vesicles, with the phospholipids in aradial orientation. Using polarized excitation, we photoselect those LAURDAN molecules with their dipoles parallel to the excitation polarization [Panel (a) of Fig. 61. LAURDAN molecules with a perpendicular orientation with respect to the excitation polarization will be weakly excite

IEEE ENGINEERING IN MEDICINE AND BIOLOGY
only in the liquid-crystalline phase of the bilayer, where more rotational mobility is allowed (Panel (b) of Fig. 6 ) and a projection of LAURDAN dipole moment along the excitation polarization exists.
If the phospholipid environment is heterogeneous (i.e., when LAURDAN GP values are heterogeneous), the excitation polarization photoselects higher GP values in the regions where thephospholipids, and the probe, are oriented parallel to the excitation polarization. Examples of this case are represented by the images obtained with vesicles composed of DOPC, DLPC, equimolar mixture of DOPC and DPPC, and the plasma membrane of the OK cells.
In regions of the vesicles where the bilayer is oriented perpendicularly to the excitation polarization, only the more fluid environments with lower GP values will be photoselected, since in these fluid environments the LAURDAN transition dipole possesses a higher rotational freedom. When the two-photon excitation occurs close to the bottom (or to the top) of the vesicles, the excitation photoselection results in GP images having a pattern as that sketched in Panel (c) of Fig. 6, with areas of higher GP values parallel to the excitation polarization, and areas of lower GP in the center and/or perpendicularly to the excitation polarization.
A clear example of this case is given by the image of the DLPC vesicle in Fig.  I(b). When the vesicles are very small or when we image a flat bilayer surface, we obtain medium and low GP values, respectively, and there is no apparent photoselection [Panels (d) and (e) of Fig.  61. Examples of a flat bilayer surface with relatively homogeneous GP values are represented by the BBM and BLM samples and images. In these membranes, the cholesterol concentration is relatively high, and consequently, the average GP values are also high [18].

Evaluation of the Size of Membrane Domains
For this purpose, we considered the photoselection caused by the excitation polarization in relation with the spatial resolution of our images, which is about 300 nm in the radial direction. In the simplest case, the membrane domains are larger than the pixel size, so that pixels of similar GP values are present in large areas. This is the case of various cellular compartments in the OK cells, where we observed differences between the high GP values of the plasma and the Golgi mem-SeptembedOttober 1999 branes and lower GP values in the complex network of cytoplasmic membranes, including the membrane of the rough and smooth endoplasmic reticulum. Nevertheless, also in the OK cells, a more complex heterogeneity is present within each of these cellular compartments. As discussed above, the GP heterogeneity of the membranes is not due to measurement errors, but it represents a real heterogeneity due to the presence of spatially distinct domains. We now propose a model to evaluate the dimension of these membrane microdomains. We can separate the case of domains about the size of a single pixel and smaller. These two circumstances are schematically represented in Fig. 7. By using the excitation photoselection method, we can distinguish among them.

Membrane Domains Comparable to the Pixel Size
In this case, each pixel contains only one of the various domains. Due to the rotational freedom of the LAURDAN molecule in the fluid liquid-crystalline phase, with polarized excitation the probe can be excited all around the vesicle circumference. Instead, in the rigid gel phase, the probe will only be excited when aligned along the excitation polarization axis. Consequently, a selective plotting of the high GP values will show pixels only in the direction parallel to the excitation polarization, while a selective plotting of low GP values will show pixels all around the circumference of the vesicle. This case was only observed in red blood cells [12]. We observed a homogeneous distribution of relatively low GP values also in the vesicles composed of DPPC. However, due to the overall homogeneity of the GP values in this sample, as judged by the narrow GP distribution

Membrane Domains Smaller than the Pixel Size
In this case, each pixel contains both fluid and rigid domains. In a pixel with the phospholipids oriented along the excitation polarization, the rigid microdomains with higher GP values will be preferentially excited. In those pixels with the phospholipids oriented perpendicularly to the excitation polarization, fluid domains September/October 1999 will be weakly excited, while the rigid domains will not be excited. Consequently, a selective plotting of lower GP values will show pixels only in the perpendicular di-rection with respect to the laser polarization. A complete separation of regions with higher and lower GP values inthe direction parallel and perpendicular to the excitation GP e 0.3 polarization, respectively, occurs only in this case of membrane microdomains smaller than the microscope image resolution. We observed this case in the vesicle samples (except for DPPC) and in the plasma membrane of the OK cells.

Images of selected GP values from the images in
Using our model, we predict that the dimensions of the membrane domains in phospholipid vesicles and in OK cells are smaller than the pixel size. This result is in agreement with previous spectroscopic studies, from which the estimated domain dimensions were in the range between 2 nm and 5 nm [16]. In the case of BBM and BLM membranes, due to the random orientation of the probe molecules within these samples, the heterogeneity in terms of domain size is difficult to interpret using our probably representing more fragile areas where membranes have been broken during their purification procedures.

Conclusions
Most fluorescent membrane probes are susceptible to photobleaching. The use of two-photon excitation opens new possibilities for microscopy studies of membranes. LAURDAN is a membrane probe with a steady-state sensitivity to the polarity of its environment, thus valuable for microscopy studies of membranes. When imaging LAURDAN GP with samples of natural membranes, we observed a large heterogeneity. By using the photoselection operated by a linearly polarized excitation, we found that the high GP values are associated with membrane regions where the fluorescent probe is aligned with the excitation polarization. Thus, we believe that the GP heterogeneity observed in natural membranes is due to coexisting domains. We have also discussed the possibilities of using polarized excitation for the evaluation of the dimension of the membrane domains.