Direct visualization of raft-like lo (liquid-ordered) domains in model systems and cells using microscopic techniques requires fluorescence probes with known partitioning preference for one of the phases present. However, fluorescent probes may display dissimilar partitioning preferences in different lipid sys-tems and can also affect the phase behaviour of the host lipid bilayer. Therefore a detailed understanding of the behaviour of fluorescent probes in defined lipid bilayer systems with known phase behaviour is essential before they can be used for identifying domain phase states. Using giant unilamellar vesicles composed of the ternary lipid mixture DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine)/DPPC (1,2-dipalmitoyl-sn-glycero-3-phosphocholine)/cholesterol, for which the phase behaviour is known, we examined nine commonly used fluorescent probes using confocal fluorescence microscopy. The partitioning preference of each probe was assigned either on the basis of quantification of the domain area fractions or by using a well-characterized ld (liquid-disordered)-phase marker. Fluorescent probes were examined both individually and using dual or triple labelling approaches. Most of the probes partitioned individually into the ld phase, whereas only NAP (naphtho[2,3-a]pyrene) and NBD-DPPE [1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl] preferred the lo phase. We found that Rh-DPPE (Lissamine™ rhodamine B–1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine) increased the miscibility transition temperature, Tmix. Interestingly, the partitioning of DiIC18 (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate) was influenced by Bodipy®-PC [2-(4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-pentanoyl)-1-hexa-decanoyl-sn-glycero-3-phosphocholine]. The specific use of each of the fluorescent probes is determined by its photostability, partitioning preference, ability to detect lipid phase separations and induced change in Tmix. We demonstrate the importance of testing a specific fluorescent probe in a given model membrane system, rather than assuming that it labels a particular lipid phase.
More than 10 years after the first proposal of their existence , lipid rafts are now regarded as a key organizing principle of cell membranes [2,3]. These nanoscale assemblies, whose structure, size and dynamics are still being explored, are thought to give rise to membrane lateral heterogeneity. However, their role in cellular biomembranes remains controversial [4,5]. They are believed to play important roles as platforms in processes as varied as signal transduction , disease pathogenesis  and intracellular sorting . The properties of lipid rafts, in particular their immiscibility with bulk membrane lipids , were noted to closely resemble those of the lo (liquid-ordered) phase , which had been described previously in cholesterol-containing lipid bilayers [11–13]. These observations provided the impetus to study model membrane systems with coexisting lo–ld (liquid-disordered) phases as models for raft formation in biological membranes (for example, see ). Ternary lipid systems composed of a high-melting and a low-melting phospholipid, together with cholesterol, have been popular choices for such studies. The application of biophysical approaches, such as NMR spectroscopy, has resulted in a detailed physical understanding of the phase behaviour of these ‘raft-forming’ lipid mixtures [15,16], allowing more insight into the processes that might occur in the membranes of intact cells. NMR approaches yield information about the molecular organization within the different phases, but do not give much insight into the different physical forms of the phases or domains. However, with the advent of GUVs (giant unilamellar vesicles), raft-like domains can be visualized directly using fluorescence microscopy by incorporating probes that report on domain shape and size.
The number of fluorescence probes used for imaging membrane domains is steadily increasing [17–19]. However, studies using fluorescently labelled lipid analogues in different binary and ternary mixtures of phospholipids and cholesterol have shown that the same fluorescent probe can have very different partitioning preference in different lipid systems . For example, DiIC18 (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate) is an ld-phase marker in mixtures of SM (sphingomyelin)/DOPC (1,2-dioleoyl-sn-glycero-3-phosphocholine)/cholesterol , but an lo-phase marker in SM/DLPC (1,2-dilauroyl-sn-glycero-3-phospho-choline)/cholesterol mixtures . Rh-DPPE (Lissamine™ rhodamine B–1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine), showed a preference for the fluid ld phase in GUVs of DMPC (1,2-dimyristoyl-sn-glycero-3-phosphocholine)/DSPC (1,2-distearoyl-sn-glycero-3-phosphocholine), whereas in GUVs of DLPC/DPPC (1,2-dipalmitoyl-sn-glycero-3-phosphocholine) it preferentially partitioned into the gel-phase . The substitution of one type of lipid for another in a binary or ternary mixture can therefore perturb the preferential partitioning of the fluorescent probe. It is evident that fluorescent probe partitioning into particular membrane domains depends on the local chemical environment of the lipid domains, not the membrane phase state itself , and it is therefore crucially important to characterize the partitioning behaviour of a particular probe in a given lipid system before assigning a phase preference to it. In addition, some fluorescent lipid analogues perturb the phase behaviour of ternary mixtures by raising the miscibility transition temperature, Tmix, by several degrees .
The phase diagram for lo-forming mixtures of DOPC/DPPC/cholesterol was published recently . We have conducted a systematic study of partitioning behaviour using a panel of commonly used fluorescent probes that takes advantage, for the first time, of the detailed knowledge of the phase behaviour of this ternary lipid system. Results of the fluorescent probe phase preference are presented from both single-labelling studies, and an approach where each probe was compared with an established reference ld-phase marker. Our study was then extended to include dual-labelling (pairs of probes) and triple-labelling (three probes simultaneously) approaches. Some probes were found to alter the partitioning behaviour of another probe in these experiments. We demonstrate the usefulness of dual/triple labelling in situations where two or three phases coexist, if the probes are chosen carefully. We also report on how different fluorescent probes, either individually or in combination, affect Tmix in these lipid mixtures, which can be determined quantitatively. Bringing to light the different characteristics of fluorescence probes and their effects on phase separation will help to avoid some of the many pitfalls when interpreting the behaviour of membrane rafts in more complex biological systems.
DOPC and DPPC were obtained from Avanti Polar Lipids and cholesterol was from Sigma–Aldrich. TR-DPPE (Texas Red–1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine), DiIC18, DiIC16 (1,1′-dihexadecyl-3,3,3′,3′-tetramethylindocarbocyanine perchlorate), DiIC12 (1,1′-didodecyl-3,3,3′,3′-tetramethylindo-carbocyanine perchlorate), DiD (1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine 4-chlorobenzenesulfonate salt), Rh-DPPE and Bodipy®-PC [2-(4,4-difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-pentanoyl)-1-hexadecanoyl-sn-glycero-3-phosphocholine] were all from Invitrogen. NDB-DPPE [1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl] and NBD-DOPE [1,2-dioleoyl-sn-glycero-3 phosphoethanolamine-N-(7-nitro-2–1,3-benzo-xadiazol-4-yl)] were supplied by Avanti Polar Lipids and NAP (naphtho[2,3-a]pyrene) was purchased from Sigma–Aldrich. The molecular structures of the fluorescent probes are presented in Supplementary Figure S1 (at http://www.BiochemJ.org/bj/430/bj4300415add.htm).
GUVs were prepared by electroformation on a pair of platinum (Pt) wires by a method first developed by Angelova and Dimitrov [24,25], modified as described previously . Lipid stock solutions were prepared in 2:1 (v/v) chloroform/methanol at 0.27 mg/ml, and appropriate volumes of each were mixed. Labelling was carried out by pre-mixing the desired fluorescent probes with the lipids in organic solvent, except in one series of experiments using NAP. The concentration of individual fluorescent probes in each sample was 0.1±0.01 mol%. In the NAP experiments, 8.41 μl of a 1 μg/ml NAP stock solution in DMSO was added to the aqueous solution in the sample chamber following GUV formation, which corresponds to 0.1 mol% of the lipid present. Drops of ~2 μl of lipids and fluorescent labels in organic solvent were deposited on Pt wires under a stream of N2 gas. The Pt wires were placed in a vacuum for ~1 h to completely remove the organic solvent. One side of the chamber was then sealed with a coverslip using a small amount of silicone grease. Heated (~60 °C) high-purity water (Millipore SuperQ) was added to the chamber until it covered the Pt wires, and then the other side of the chamber was sealed with another coverslip. The chamber was placed in an oven (Isotemp 500 series), and the Pt wires were connected to a function generator (HP 3310A or BK Precision 4011A). A sinusoidal wave function of amplitude 3 V and frequency 10 Hz was applied for ~90 min, after which the electric field was turned off and the chamber was placed on the microscope stage. Typically at least 10 min equilibration time was allowed before proceeding with the temperature scans, which usually started at 42 °C.
CFM (confocal fluorescence microscopy)
A custom-made imaging chamber described previously  was employed for imaging the GUVs by CFM. The sample temperature was controlled (±0.1 °C) above room temperature (22 °C) by applying an electric current to the imaging chamber (TC-324B; Warner Instruments) and below room temperature by circulating water from a water bath (Isotemp 3016; Fisher Scientific) underneath the sample compartment. Two thermistors were used to monitor the temperature, one connected to the imaging chamber and the other placed in the sample compartment (cable assembly for series 20 chambers, CC-28; Warner Instruments). The temperature values given are the readings from the thermistor situated in the sample compartment between the two Pt wires, which were separated by 3 mm. The cooling rate from a set-point temperature was 1–2 °C per min; this rate slowed as the temperature approached the new set-point temperature. Typically, different sections of the Pt wires were checked for consistency of Tmix and domain area fractions, which were measured as described previously . Tmix values were determined from observations of 70–100 separate GUVs generated from at least two different lipid droplets on the Pt wires, by raising and lowering the temperature through Tmix. TR-DPPE was used as a reference probe for each lipid mixture.
Images were acquired on commercial Leica confocal microscopes, either on a CLSM-U, an upright Leica DM RE microscope connected to a Leica TCS SP2 system, or on a CLSM-MP, an upright Leica DM 6000B microscope connected to a Leica TCS SP5. The following objectives were used to image the samples: a 20× air objective (not touching the sample) with NA (numerical aperture) of 0.5, a 40× water-immersion objective with NA 0.8 and a 60× water-immersion objective with NA 0.9. The excitation wavelength used for NBD-DPPE, NBD-DOPE and Bodipy®-PC was 488 nm, whereas 543 nm was used for TR-DPPE, DiIC12, DiIC16, DiIC18 and Rh-DPPE. DiD and NAP were excited at 633 nm and 305 nm respectively. The confocal pinhole settings were in the range 1–2 Airy, 1 Airy being the default value used for a given objective. The intensity of the lasers was kept at ~60%. All images were routinely checked for under- and over-exposure using the QLut function in the Leica confocal software. We typically used 1024 pixel×1024 pixel resolution with a 400 Hz or 800 Hz scan speed, and a line average of typically two or four, but occasionally eight. The three-dimensional representation of a GUV from two-dimensional images was carried out with the Leica confocal software using a step size of 1 μm and 1.6 μm (see Figures 1D and 1E) in the vertical z direction. This step size was slightly larger than that suggested by the Leica confocal software, which optimizes the step size and number of images to be sequentially acquired according to the Nyquist criterion. When using combinations of fluorescent probes in CFM, care was taken to rule out artefacts relating to ‘bleed-through’ of one probe's fluorescence emission signal in the detection channel of another probe. This was especially true when using NBD-DPPE together with DiIC18, because of the large overlap of their emission signals. Typically, the excitation and/or detection in one channel were turned off, and the acquired image in the other channel was compared with the image acquired with both channels active.
Phase preference of individual fluorescent probes in DOPC/DPPC/cholesterol GUVs
RESULTS AND DISCUSSION
In the investigation described below, we chose to use two compositions of the lipid mixture DOPC/DPPC/cholesterol, the phase behaviour of which was previously established by both CFM  and NMR spectroscopy . To demonstrate the general applicability of the approach, we felt it was important to show that the observed partitioning behaviour of the fluorescence probes applied to more than one lipid mixture, with different Tmix values and different proportions of lo and ld phases in the temperature range under study.
Visualization of phase separation in GUVs by CFM
The most commonly used fluorescence probes in model membrane studies of lo–ld two-phase coexistence are dialkylcarbocyanine dyes with varying chain lengths of 12–20 carbons, headgroup-labelled DPPE species, such as NBD-DPPE, TR-DPPE and Rh-DPPE, and the acyl-chain-labelled fluorescent lipid, Bodipy®-PC. We prepared GUVs with compositions for which we had previously determined domain area fractions, using CFM with TR-DPPE labelling . Using TR-DPPE as a reference ld-phase marker was helpful in assigning the phases where the fluorescent probes were located. We examined the partitioning behaviour of a total of nine fluorescent probes (not including the reference probe TR-DPPE) in GUVs composed of ternary mixtures of DOPC/DPPC/cholesterol, which display coexisting lo and ld phases. The partial phase diagram was recently determined for this lipid system . All fluorescent lipid analogues, except Bodipy®-PC, are headgroup-labelled, whereas NAP is a polycyclic aromatic hydrocarbon (Supplementary Figure S1). As described previously , when GUVs are observed at a temperature well above the region of two-phase coexistence, the fluorescence intensity of each probe is homogeneously distributed over the surface. As the temperature is slowly decreased, phase separation begins and leads to fluctuations in surface fluorescence intensity. The highest temperature at which this is detected defines Tmix. Following cooling to a temperature below Tmix, many small lo domains form in each GUV and typically coalesce to a single large circular domain after reaching equilibrium, which takes approx. 5 min, after which the domain distribution remains stable for at least 20–25 min . The area fractions for the lo and ld domains can be quantified, as described in our previous studies . Assignment of phase partitioning for a particular fluorescent probe was carried out by inspecting the domain area fractions on cooling and/or comparing its domain distribution with that of TR-DPPE . The partitioning preference of the fluorescent probes for the ld phase compared with the lo phase was assessed qualitatively. Quantification of partition coefficients from CFM images is challenging , due to concentration-dependent self-quenching, varying fluorescent intensities in the different lipid phases and selective excitation of the fluorophores due to their differing orientations in different phases.
Probe partitioning in ternary mixtures of DOPC/DPPC/cholesterol: individual labelling with a single fluorescent probe
The two lipid mixtures used in the individual labelling experiments (DOPC/DPPC/cholesterol at 35:35:30 mol% and 32:48:20 mol%) have Tmix values (Table 1) determined previously by 2H-NMR  and by CFM using TR-DPPE . Both are in the ld–lo two-phase coexistence region below Tmix. CFM images resulting from labelling of GUVs with individual fluorescent probes are presented in Figure 1, and the observed values of Tmix for each label are shown in Table 1.
|Sample||Fluorescent probes||Tmix (°C)|
|DOPC/DPPC/cholesterol (35:35:30 mol%; Tmix=30.5±1 °C)||TR-DPPE||30.5±1|
|DOPC/DPPC/cholesterol (32:48:20 mol%; Tmix=35.5±0.5 °C)||TR-DPPE||35.5±0.5|
|Sample||Fluorescent probes||Tmix (°C)|
|DOPC/DPPC/cholesterol (35:35:30 mol%; Tmix=30.5±1 °C)||TR-DPPE||30.5±1|
|DOPC/DPPC/cholesterol (32:48:20 mol%; Tmix=35.5±0.5 °C)||TR-DPPE||35.5±0.5|
These Tmix values were judged to be elevated compared with the reference probe TR-DPPE.
The dialkylcarbocyanine dyes DiIC12 and DiIC16 preferentially partitioned into the ld phase (Figures 1A and 1C), as on cooling the bright regions on the GUV surface (ld phase) were observed to shrink, and new dark domains (lo phase) formed on the bright background. This behaviour is expected for a sample of this composition, since on cooling the lipid mixture contains a decreasing amount of the ld phase [16,26]. No significant change in Tmix was recorded with these two probes, compared with TR-DPPE (Table 1). By analysing the fluorescence intensity in the two phases, we concluded that the differential partitioning of DiIC16 into the ld phase is not as strong as that of DiIC12. When DiIC12 was used together with TR-DPPE, we found that the two molecules co-partitioned into the ld phase (results not shown). For these two probes, ld phase partitioning has been reported previously in other ternary membrane systems [18,27]. The longer chain fluorescent probe, DiIC18, showed no preference for either the ld or lo phase in the temperature range 42.6–14.7 °C, since no change in fluorescence intensity over the GUV surface was observed as the sample was cooled (Figure 1D). Interestingly, several studies have reported that DiIC18 changes its preference for the ld or lo phase in different lipid mixtures (Supplementary Table S1 at http://www.BiochemJ.org/bj/430/bj4300415add.htm). Thus the shorter probes, DiIC12 and DiIC16, prefer the ld phase in the ternary mixture studied, whereas the phase preference of DiIC18 clearly depends on the nature of the lipid bilayer. A recent study could assign no phase preference for a similar fluorescent probe, DiOC18 (3,3-dioctadecyloxacarbocyanine perchlorate), in DOPC/DPPC-d62 (1,2-diperdeuteropalmitoyl-sn-glycero-3-phosphocholine)/cholesterol (35:35:30 mol%) . The related probe DiD partitioned into the ld phase. Tmix was not altered, but the temperature difference between the upper and lower boundaries was larger than that observed for the other probes used singly (Table 1). This effect appeared to arise from slightly more inhomogeneous GUV formation.
Rh-DPPE was reported to partition preferentially into ld domains in mixtures of egg SM/DOPC/cholesterol . This partitioning was so complete that almost no fluorescence intensity was detected in the lo domains. We observed similar partitioning behaviour for this probe (Figure 1B shows a representative image). Among the probes examined, Rh-DPPE displayed the highest fluorescence intensity in the ld domains. Its only drawback was that we detected the onset of the phase separation at Tmix ~2.5 °C higher than expected (Table 1).
NBD-DPPE is one of the rare fluorescent probes that has been reported to preferentially partition into lo domains in several model membrane systems [27,29,30]. In agreement with these earlier investigations, we concluded that NBD-DPPE also favoured lo domains in DOPC/DPPC/cholesterol (32:48:20 mol%). According to the phase diagram  this lipid mixture is in the ld–lo–gel three-phase region at the temperature presented in Figure 1(E). The onset of the three-phase separation for this composition is at ~23 °C when using DPPC-d62 as one component , and in a sample with the same composition prepared using DPPC, the onset of the three-phase region should occur at most ~3.5 °C higher, at ~26.5 °C. Since this mixture has an almost equal lo and ld domain area fraction just before it enters the three-phase region , a particular phase cannot be unambiguously assigned to the bright domain on the surface of the GUV in Figure 1(E) by visual inspection. However, additional proof of NBD-DPPE lo-phase preference will be presented below.
Evaluation of the phase partitioning behaviour of NBD-DOPE by simply comparing domain area fractions at 23.8 °C (Figure 1F) is also challenging for the reasons mentioned above. However, when examining the sample thermal history, we concluded that NBD-DOPE partitioned into the ld phase. Phase separation started at ~35 °C with the appearance of dark domains on the GUV surface and, as the temperature was lowered, they increased in size, whereas the bright domains shrank. According to the phase diagram , the area fraction of the lo domains increases as temperature decreases.
Bodipy®-PC typically prefers the less-ordered phase because of the presence of a covalently linked fluorophore on its acyl chain, and was found to partition into the ld phase (Figure 1G). Phase separation started with the formation of small dark domains on a bright background. As the temperature was lowered, the dark lo domains on the GUV surface increased in size and the bright fraction decreased, suggesting that Bodipy®-PC prefers the ld phase. No significant shift in Tmix was observed (Table 1). Bodipy®-PC was reported to partition into the ld phase in similar membrane systems [31,32].
Incorporating NAP, an aromatic fluorescent probe, into GUVs proved to be challenging (see the section below for details). When NAP was used as a single probe, incorporated by addition to pre-formed GUVs, it did not alter Tmix (Table 1).
Overall, these results indicate that if the phase behaviour of a lipid mixture is well defined, then the single label approach can be useful for characterizing the phase partitioning behaviour of a number of fluorescent probes. In some cases, further experiments (described below) were necessary to confirm these assignments.
Probe partitioning in ternary mixtures of DOPC/DPPC/cholesterol: comparison with the reference probe TR-DPPE
Phase assignments for the various probes were confirmed by co-labelling with the reference probe TR-DPPE, which we previously identified as a reliable ld-phase marker in the ternary lipid mixture used . In many cases, it was possible to detect fluorescence from each probe simultaneously in separate emission channels.
TR-DPPE/Rh-DPPE and TR-DPPE/DiIC12
Simultaneous detection of the fluorescence signals in separate channels was not possible due to overlap in the excitation/emission wavelengths of the two probes. As expected from the results of individual labelling, all probes partitioned into the ld phase in DOPC/DPPC/cholesterol (32:48:20 mol%), and assignment of their phase preference was straightforward. By lowering the temperature through Tmix, small dark domains formed on the surface of GUVs and then increased in size. After equilibration at room temperature, two large domains (one lo phase, the other ld phase) were observed, similar to those in Figure 1(C). GUVs containing both TR-DPPE and Rh-DPPE were observed to start phase-separating at 38 °C, above the expected Tmix for this sample. The upward shift in Tmix was also observed when Rh-DPPE was used alone (Table 1).
DiD, a dialkylcarbocyanine dye with 18-carbon chains, was previously observed to partition strongly into the ld phase in ternary mixtures [33–35]. We examined the partitioning of this probe in DOPC/DPPC/cholesterol (35:35:30 mol%) when used together with TR-DPPE, which also labels the ld phase. The images in Figures 2(A) and 2(B) were acquired simultaneously in two separate detection channels. As the two probes have very different excitation/emission spectra, emission ‘cross-talk’ can be easily ruled out in the acquired images. At 32.4 °C, shortly after the sample had passed through Tmix and before domain equilibration, both probes are strongly excluded from lo domains (Figures 2A–2C). Fading of DiD is noticeable in the overlay image (Figure 2C), where the absence of green fluorescence leaves an increased red intensity. We observed that the Tmix with the combination of these two labels was ~2 °C higher than with either probe alone (Table 1).
Fluorescent probe partitioning compared with the reference probe TR-DPPE
Figures 2(D)–2(F) shows the different partitioning properties of TR-DPPE and NBD-DPPE; the images in Figures 2(D) and 2(E) were acquired simultaneously in two separate detection channels. Both probes displayed the same partitioning behaviour as when observed individually. Thus the fluorescence of TR-DPPE was observed to arise almost entirely from the ld phase (Figure 2D), and weak partitioning into the lo phase was observed for NBD-DPPE (Figure 2E). According to the observed distribution of TR-DPPE, the ld domain on the GUV surface occupies a smaller area fraction than the lo domain labelled by NBD-DPPE, in agreement with earlier results [16,26]. No shift in Tmix was recorded with this pair of labels (Table 1).
Figures 2(G)–2(I) present the partitioning behaviour of TR-DPPE and Bodipy®-PC at 25.7 °C. On lowering the temperature, large fluorescent intensity fluctuations were detected starting at 30.7 °C (Table 1) and complete phase separation was recorded at 29.9 °C, when all the GUVs in the field of view had phase-separated. We found that both labels preferred the ld phase, although Bodipy®-PC had a weaker partitioning preference when compared with TR-DPPE. This effect is noticeable in the overlay image in Figure 2(I) where some green fluorescence from Bodipy®-PC is seen in the lo domain. As the excitation and emission wavelengths of these two fluorescent probes are well separated, we can unambiguously assign an ld-phase preference for both fluorescent probes.
Initially we employed the same sample preparation method as for the other fluorescent probes, using DOPC/DPPC/cholesterol (32:48:20 mol%) with NAP dissolved in DMSO. However, solvent evaporation took considerably longer (overnight compared with 1 h) and GUV growth did not result in a homogeneous sample. According to the phase diagram , no gel phase should exist above 22 °C for this sample composition, yet we found regions on the surface of several GUVs where solid-like phases (non-circular domains) were detectable at temperatures as high as 34.8 °C (Figure 3A). Thick fibril domains were labelled by NAP and depleted of TR-DPPE at 27.9 °C (Figure 3C), and the NAP fluorescence intensity was increased around dark non-circular domains depleted of both fluorescent labels, such as the ones shown in the overlay image in Figure 3(A) (the result is more obvious in images acquired in the NAP channel alone; results not shown). Despite the sample inhomogeneity, individual vesicles that did not show solid-like phases could be selected (Figures 3B and 3D–3F). These images are similar to those acquired of samples labelled with TR-DPPE/NBD-DPPE (Figures 2D–2F), except that NAP showed stronger preferential partitioning into the lo domain than NBD-DPPE, and had a much higher quantum yield and photostability.
NAP incorporation efficiency
We next incorporated NAP into GUVs by adding the probe to the aqueous phase after their formation. Following co-incubation for ~1 h at 26.5 °C, the fluorescent probe started to incorporate into the vesicles (Figures 3G–3I), and it was obvious from these images that NAP did not promote the formation of solid-like phases, such as those observed when it was premixed with lipids in solvent. Labelling of the lo phase by NAP could be greatly improved by increasing the temperature to 42 °C and optimizing the probe concentration. NAP and several other fluorescent polycyclic hydrocarbon probes were reported previously to preferentially partition into lo phases in egg SM/DOPC/cholesterol mixtures . Perylene was also found to partition into lo phases in egg SM/DOPC/cholesterol , yet showed no preferential partitioning in brain SM/DOPC/cholesterol . The major advantage of these polycyclic aromatic hydrocarbon fluorophores is their high photostability, satisfactory quantum yield and short excitation wavelength, so that they can be easily used in dual- and triple-labelling experiments with other fluorescent probes.
In summary, co-labelling experiments with the reference probe TR-DPPE are an excellent means to make (or confirm) assignments of phase partitioning preference for any fluorescent probe.
Combinations of fluorescence probes: a dual-labelling approach
The dual-labelling approach may be very useful in visualizing lipid domains and assigning their phases. However, a full understanding of whether the two probes alter each other's behaviour, and how they affect the lipid system under study, is essential for correct interpretation of the results.
Samples of DOPC/DPPC/cholesterol (35:35:30 mol%) co-labelled with Bodipy®-PC and DiIC18 showed interesting behaviour. When close to Tmix, as detected by Bodipy®-PC (~30 °C, see Table 1), fluorescence intensity fluctuation was observed, and as the temperature was slowly decreased and domain equilibration was achieved, DiIC18 co-partitioned with Bodipy®-PC into the same domains (Figures 4A–4C). We propose that both molecules partition into the ld phase in this mixture, since the dark domains on a bright background increased in size, and additional small dark domains formed on the GUV surface, as the temperature was decreased from ~29 °C to 26.7 °C. We ruled out bleed-through between the two detection channels by shutting off excitation/detection of Bodipy®-PC while detecting DiIC18 fluorescence (Figure 4A), and vice versa (Figure 4B). Figure 4(C) shows an overlay image resulting from simultaneous excitation of both fluorophores, where it is clear that the two labels co-partition into the same ld domains; however, DiIC18 had weaker differential partitioning than Bodipy®-PC. Clearly, Bodipy®-PC influences the partitioning behaviour of DiIC18.
Dual-labelling in DOPC/DPPC/cholesterol GUVs: DiIC18/Bodipy®-PC and DiIC18/DiD
NBD-DPPE/DiIC18, NBD-DPPE/DiD and DiIC18/DiD
Owing to the low quantum yield and weak partitioning preference of NBD-DPPE, it was challenging to assign its phase preference when used together with DiIC18 (results not shown); however, it appeared that DiIC18 did not have a phase partitioning preference when used with NBD-DPPE. When NBD-DPPE was used together with DiD, the well-separated emission spectra of the two probes allowed unambiguous phase assignment. Thus NBD-DPPE preferred the lo phase, from which DiD was almost completely excluded (results not shown). To test the partitioning preference of DiIC18 together with another ld-phase-preferring probe, we prepared GUVs labelled with both DiIC18 and DiD, which differ only in the length of the headgroup region linker (Supplementary Figure S1). At 24 °C, a temperature at which the lipid mixture is well below Tmix [16,26], no noticeable partitioning preference of DiIC18 was observed (Figure 4D), whereas DiD was found to segregate into ld domains (Figure 4E). The fact that DiIC18 has no preference for either of the two phases can be seen in the overlay image (Figure 4F), where the lo domains are red (DiIC18) and the ld domains are yellow due to the combination of red and green fluorescence from both probes.
The dual-labelling approach clearly showed that one fluorescent probe may alter the partitioning behaviour of the other, so it cannot be assumed that the phase preference displayed when using single fluorophores always applies in a two-probe situation.
Combinations of fluorescence probes: a triple-labelling approach
We extended our investigation to examine lipid mixtures that were labelled with three probes simultaneously, as this novel approach might be useful in domain identification where two or three lipid phases coexist.
We tested DiIC18 partitioning in DOPC/DPPC/cholesterol (20:60:20 mol%) in the presence of both an lo-preferring probe NBD-DPPE and an ld-preferring probe DiD. This lipid mixture enters the three-phase region (with coexisting ld, lo and gel phases) below 30 °C . According to the phase diagram  the onset of the two-phase region for this lipid composition is ~35 °C. The highest temperature examined for this sample was 38.2 °C, where homogeneous fluorescence intensity was recorded in the emission channel for each probe, and in overlay images. We observed the onset of phase separation at ~35.1 °C, as expected, in agreement with the observation that the presence of DiD results in a consistent increase in Tmix. We detected the formation of bright domains on a dark background in the DiD detection channel, which decreased in size as the temperature was lowered, suggesting that DiD partitioned into ld domains. In the NBD-DPPE detection channel we observed exactly complementary behaviour, since this probe has a preference for the lo phase. The temperature was then further decreased to ~16 °C, where the sample is well within the three-phase region, and contains almost equal amounts of lo and gel phases, together with smaller amounts of ld phase . As can be seen in Figures 5(A) and 5(C), both NBD-DPPE and DiD maintained their partitioning preference for the lo and ld phases respectively, whereas DiIC18 was observed to have no preference for any particular phase (Figure 5B). None of the three fluorescence probes could distinguish the presence of gel-phase domains.
Triple-labelling in DOPC/DPPC/cholesterol GUVs: NBD-DPPE/DiIC18/DiD and NBD-DPPE/TR-DPPE/DiD
DOPC/DPPC/cholesterol (35:35:30 mol%) was labelled simultaneously with three probes, two preferring the ld phase (TR-DPPE and DiD), and one preferring the lo phase (NBD-DPPE). As expected from the incorporation of DiD, a ~3 °C upshift in Tmix was recorded (Table 1). Phase separation started with intensity fluctuations, and dark domains formed on a bright background in both the TR-DPPE and DiD emission images. These dark domains coalesced relatively quickly and formed one or two large domains. These details could not be detected in NBD-DPPE images due to the dye's poor fluorescence properties, but when the domains reached equilibrium, the differential partitioning of NBD-DPPE was obvious, showing that this probe partitioned into domains from which TR-DPPE and DiD were excluded (Figures 5E–5H). The fluorescence intensity of NBD-DPPE (Figure 5E) is higher in regions where intensities for TR-DPPE (Figure 5F) and DiD (Figure 5G) are considerably lower. By comparing Figures 5(F) and 5(G), we can conclude that DiD has a stronger differential partitioning into ld domains than TR-DPPE. Figure 5(H) shows the overlay of images acquired in the three separate channels; the contrast between the lo and ld phases is obvious.
We conclude that labelling experiments with three different fluorophores can clearly demonstrate the contrast between the lo and ld phases, and the complementary behaviour of probes with different phase-partitioning preferences is readily observed.
The use of fluorescent probes to explore the existence of lo and ld phase lipid domains has increased greatly over the past decade, and obtaining quantitative information is now possible using CFM [26,37]. However, it is very important to characterize the partitioning behaviour of a fluorescence probe before assigning it a phase preference and employing it to visualize or quantify domains by CFM. Such information on the phase preferences of several commonly used probes, often thought to be domain-specific, has been provided by the present systematic investigation in a ternary lipid bilayer membrane system. Knowledge of the detailed phase diagrams for the lipid systems in the present study provides certainty as to the type and relative amount of the various phases present. The simultaneous use of two and three different probes in the present study was found to have advantages, but unappreciated complexities of this approach were also revealed, including the ability of one probe to alter the partitioning behaviour of another.
Most of the fluorescent probes we examined preferred the ld phase over the more ordered lo phase in the ternary mixtures of DOPC/DPPC/cholesterol. According to our phase assignments for this ternary lipid system (Table 2), we found two reliable lo-phase labels, NBD-DPPE and NAP. However, care has to be taken when visualizing GUVs with NAP due to sample inhomogeneity and a post-formation incorporation technique is advisable. The relatively small NBD fluorophore linked to the headgroup of saturated-chain DPPE makes it possible for this lipid to also pack into lo domains induced by cholesterol. The drawbacks of this fluorescent probe are its low fluorescence quantum yield, sensitivity to photobleaching and weak preferential partitioning. Acquisition of sharp fluorescence images exhibiting contrast between the two phases can thus sometimes be challenging for lipid probes labelled with NBD. When NBD-DPPE is used as a single fluorescent probe at a concentration of 0.1 mol% it also cannot provide fine details of the phase separation, such as Tmix or domain evolution. Although NBD-DPPE has a low excitation maximum of ~466 nm, its emission spectrum covers a wide range (485–700 nm) and care must be taken in dual-labelling experiments to avoid overlap with the emission spectra of other probes with emission maxima in the 550–580 nm range. Nevertheless, when used in dual-labelling experiments with an ld-preferring probe, such as TR-DPPE (emission maximum ~608 nm), NBD-DPPE offers good contrast in visualizing ld and lo domains in DOPC/DPPC/cholesterol mixtures.
|Phase partitioning preference|
|DiIC18||No preference||ld/no preference||No preference|
|Phase partitioning preference|
|DiIC18||No preference||ld/no preference||No preference|
We conclude that TR-DPPE is by far the most reliable ld-phase label in GUVs of DOPC/DPPC/cholesterol. One reason for this is that the recorded Tmix and the previously measured domain area fractions are in full agreement with the 2H-NMR results . Furthermore, TR-DPPE showed consistent preferential ld phase partitioning when used in any combination with other fluorescent labels, and it also has the advantage of high photostability. Owing to its higher excitation wavelength (633 nm), DiD is also a good choice for an ld-phase label, although photobleaching of the dye was observed and an increase of ~2.5–3 °C in Tmix was also noted. Among the other probes we investigated, Rh-DPPE has the advantage of high fluorescence yield and high photostability and the disadvantage of raising Tmix by ~2.5 °C. NBD-DOPE is the poorest ld-phase label among all the probes included in the present study, because of its low quantum yield and weak preferential partitioning. Bodipy®-PC is a general ld-phase marker with relatively strong preferential partitioning. Surprisingly, it affected the partitioning behaviour of DiIC18, a probe with no phase preference in DOPC/DPPC/cholesterol mixtures. Overall, our results demonstrate the importance of thorough testing of various fluorescent probes in a given model membrane system, rather than making the assumption that a specific probe labels a particular lipid phase.
It seems highly unlikely that the differences in phase partitioning preferences between the probes could arise from the formation of lipid peroxides, which have been reported to form during fluorescence microscopic illumination in lipid mixtures containing NBD-DPPE and Rho-DOPE (rhodamine dioleoylphosphatidylethanolamine), and promote the formation of large raft domains . First, the lipid mixtures we used contained no SM, which was found to be the culprit in previous studies. Secondly, we used very low concentrations of fluorescent probes (0.1% compared with 1–3% in ). Finally, the phase preferences summarized in Table 2 are supported by other studies of individual fluorescence probes in various lipid mixtures (see the Results and discussion section).
The GUV model membrane system can be made more complex and biologically relevant by the incorporation of membrane proteins (e.g. [38,39]), allowing the distribution of these proteins between domains to be examined in raft-forming mixtures. Giant plasma membrane vesicles (e.g. [40–42]) also represent native-like systems for exploring the presumed link between domain formation and protein behaviour. Expanding the application of CFM using domain-preferring probes to these systems will probably be an important tool to understand their behaviour.
confocal fluorescence microscopy
1,1′-dioctadecyl-3,3,3′,3′-tetramethylindodicarbocyanine 4-chlorobenzenesulfonate salt
giant unilamellar vesicle
Lissamine™ rhodamine B–1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine
miscibility transition temperature
Janos Juhasz performed all of the experiments and wrote the first draft of the manuscript. Frances Sharom and James Davis assisted in design of the experiments and edited the manuscript.
We are grateful for the technical assistance of Michaela Struder-Kypke and Joseph Chu.
This work was supported by grants to F. J. S. and J. H. D. from the Natural Sciences and Engineering Research Council of Canada, the Canada Foundation for Innovation and the Ontario Research Fund.