Fluorescence lifetime imaging microscopy of flexible and rigid dyes probes the biophysical properties of synthetic and biological membranes

Abstract

Sensing of the biophysical properties of membranes using molecular reporters has recently regained widespread attention. This was elicited by the development of new probes of exquisite optical properties and increased performance, combined with developments in fluorescence detection. Here, we report on fluorescence lifetime imaging of various rigid and flexible fluorescent dyes to probe the biophysical properties of synthetic and biological membranes at steady state as well as upon the action of external membrane-modifying agents. We tested the solvatochromic dyes Nile red and 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl) (ammonium salt) (NBD), the viscosity sensor Bodipy C₁₂, the flipper dye FliptR, as well as the dyes 3,3′-dioctadecyloxacarbocyanine perchlorate (DiO), Bodipy C₁₆, lissamine-rhodamine, and Atto647, which are dyes with no previous reported environmental sensitivity. The performance of the fluorescent probes, many of which are commercially available, was benchmarked with well-known environmental reporters, with Nile red and Bodipy C₁₂ being specific reporters of medium hydration and viscosity, respectively. We show that some widely used ordinary dyes with no previous report of sensing capabilities can exhibit competing performance compared to highly sensitive commercially available or custom-based solvatochromic dyes, molecular rotors, or flipper in a wide range of biophysics experiments. Compared to other methods, fluorescence lifetime imaging is a minimally invasive and nondestructive method with optical resolution. It enables biophysical mapping at steady state or assessment of the changes induced by membrane-active molecules at subcellular level in both synthetic and biological membranes when intensity measurements fail to do so. The results have important consequences for the specific choice of the sensor and take into consideration factors such as probe sensitivity, response to environmental changes, ease and speed of data analysis, and the probe’s intracellular distribution, as well as potential side effects induced by labeling and imaging.

Significance

Fluorescence lifetime imaging microscopy (FLIM) is a powerful technique able to provide environmental maps at (sub-)cellular resolution. To maximize FLIM capabilities, a combination of advanced instrumentation, optimized data analysis, and suitable fluorescent reporters is paramount. In this study, we address the latter. We critically compare the performance of several well-established environmental sensors, as well as probes with no previous report of environmental sensing capability to detect the environment, in both synthetic and biological membranes. We provide a guide on which probe may be the best FLIM reporter depending on the specific application and model system. We show that an optimal probe sensitively detects dynamic environmental changes at optical resolution in conditions where many other methods fail to do so.

Introduction

Fluorescence microscopy has long been recognized as a powerful tool to unravel the structure and function of cells (1,2). It enables detailed observation of cellular constituents at and beyond the diffraction limit thanks to advances in imaging capabilities and design of fluorescence probes (3,4). More than resolving structure, fluorescence imaging can be used to track and quantify the activity of cells and single molecules, reporting on molecular concentration, stoichiometry, and mobility, to name a few (5,6). Sensitive detection requires bright and photostable dyes that label the target structure with high specificity and sensitivity, which is ideally combined with high spatial and temporal resolution. Beyond mere passive probes, fluorescent dyes have long been used as reporters of local physical-chemical properties (7,8), with major examples including ion (e.g., calcium) sensors (9,10).

Fluorescent reporters that sense the local environment of membranes have recently regained attraction due to the development of specific and sensitive sensors, advances in spectroscopy and microscopy, and the ever-growing popularity of membrane biophysics (11,12). Environmental-sensitive probes are molecules that change one or more of their fluorescent properties (i.e., intensity, spectra, anisotropy, and lifetime) in response to the surrounding environment. They can be classified as 1) solvatochromic probes, molecules that undergo excited-state charge transfer and change their color in response to polarity (12); 2) molecular rotors, dyes that can twist and respond to viscosity by changes in quantum yield (QY) and fluorescence lifetime (13,14,15); and 3) flippers, flexible molecules that change both their spectrum and lifetime in response to changes in membrane tension (16) and order (17). Typical solvatochromic probes include the well-characterized and commercially available Nile red (NR) (18), 1,2-dimyristoyl-sn-glycero-3-phosphoethanolamine-N-(7-nitro-2-1,3-benzoxadiazol-4-yl) (ammonium salt) (NBD) (19), and Laurdan (20). Environmental sensing is either measured ratiometrically (e.g., generalized polarization (GP)) (21) or upon scanning the whole emission spectrum (22). Regardless of the molecular mechanisms, all these dyes have in common their sensitivity to factors such as hydration, packing, and viscosity, and, not surprisingly, environmental sensors have found a wide range of applications in cell biology and membrane biophysics (13,16,17,23).

Membranes composed of short and/or unsaturated lipids are known to be fluid and in the liquid disordered (Ld) state (24), a phase characterized by low packing, low order, high hydration, and mobility (25), whereas membranes composed of long and saturated lipids are in the solid (So) state, characterized by highly ordered, packed, and dehydrated bilayers with virtually frozen molecules (25). Both phases are soluble in nonionic detergents (26). The addition of cholesterol to either membrane leads to the formation of an ordered, yet fluid, liquid ordered (Lo) phase with intermediate physical properties (25), rendering them insoluble in detergents (26,27). These phases are classically identified based on the dye’s chemical affinity to a certain phase. In general, Ld membranes give rise to low GP (22) and short lifetimes (28), whereas Lo membranes exhibit high brightness, high GP, and longer lifetimes. Membranes formed by mixtures of saturated and unsaturated lipids with cholesterol separate into multiple phases depending on the ratio of their constituents (24), and changes in fluorescence (mostly brightness) can be used to identify the domain identity. Changes in brightness are typically assumed to be a sole result of concentration, whereas the phase itself may induce changes in the probe’s QY, a factor that is commonly ignored.

Environmental sensing can be studied purely spectroscopically or in combination with microscopy. The latter has the additional advantage that structural information can be obtained at optical (or sub-optical) resolution. In either case, it can be assessed by changes in 1) fluorescence intensity, 2) spectral changes, 3) polarization changes, or 4) lifetime changes. The first approach is simpler but it is not suitable for cell experiments as the local probe concentration is typically unknown. Polarization can be used to study homo- Förster resonance energy transfer (FRET) or viscosity (29), the latter via a change in the rotational mobility of the probe. Spectral and lifetime changes are complementary approaches and can sense a wider range of environmental parameters than polarization (e.g., polarity, pH, ion concentration). The former is simpler but the obtained data depend on the settings used and the obtained parameters (e.g., GP) do not have a physical meaning. In lifetime-based experiments, such as fluorescence lifetime imaging microscopy (FLIM), the fluorescence decay is specific to the probe used and less dependent on the detection settings. FLIM contrast is given by the fluorescence lifetime τ, which is the average time the fluorophore remains in the excited state after excitation and is defined as the inverse sum of the radiative rate and nonradiative rate constants k_r and k_nr:

$$\tau =1/\left({\mathrm{k}}_{\mathrm{r}}+{\mathrm{k}}_{\text{nr}}\right)$$ (Equation 1)

The nonradiative rate constant k_nr includes the rate for intersystem crossing to the triplet state and the interaction with other molecules (e.g., oxygen or other quenchers). For the fluorescent molecular rotor Bodipy C₁₂, the nonradiative rate constant is a function of the free volume, which is proportional to the viscosity of the solvent (30). The recorded lifetime can be converted into meaningful physical properties, such as viscosity (13) and hydration (18). In FLIM, an ideal reporter 1) should sensitively report environmental changes by clear changes in lifetime; 2) its fluorescence decay should ideally be a monoexponential decay to facilitate analysis; and 3) longer-lifetime probes may be preferred to maximize the dynamic range. Furthermore, in cell applications, cellular distribution may be critical.

Fluorophores for protein characterization have been reviewed in the past (31). In this work, rather than developing new dyes, we use FLIM to characterize a series of commercially available and already developed custom-based rigid and flexible dyes to assess and compare environmental sensing and applicability in synthetic and biological membranes at steady state or upon the action of membrane-active molecules. More specifically, the study offers a compilation of a series of environmental probes used in FLIM to report on various membrane biophysical properties. Although this report is not comprehensive, it provides a robust comparison of the performance of different commonly used probes, especially when performing challenging biophysical experiments that require excellent probe performance. Sensing was tested by doping the individual dyes in the Ld or Lo giant unilamellar vesicles (GUVs) and assessing the changes in QY and fluorescence lifetime in well-defined membrane environments, as well as the dynamic changes in membrane biophysics upon the action of external membrane-active molecules. All classes of probes sensitively detect membrane’s biophysical properties. Several ordinary commercial dyes with no previous report of environmental sensing also detect membrane phases, although not as sensitively as well-established reporters. In cells, these probes were able to sensitively map spatial biophysical heterogeneities statically or dynamically. None of the fluorescent probes induced significant toxicity, although prolonged imaging with strong illumination may cause photoinduced-environmental responses. We provide a guide on the choice of the most appropriate reporter based on the specific application depending on its performance as well as experimental system. We anticipate the results to aid researchers on their choice of the appropriate environmental sensor and FLIM as a highly sensitive method for dynamically monitoring the biophysical properties of biomimetic or biological membrane at and beyond optical resolution, especially in conditions where intensity measurements are insufficient.

Materials and methods

Fluorescent dyes and chemicals

All materials and chemicals were used without further purification. The phospholipids 1,2-dioleoyl-sn-glycero-3-phosphocholine (DOPC), sphingomyelin (brain, porcine), cholesterol (Chol), and the fluorescent probes 1,2-dipalmitoyl-sn-glycero-3-phosphoethanolamine-N-(lissamine-rhodamine B sulfonyl) (ammonium salt) (PE-Rh) and NBD were purchased from Avanti Polar Lipids (Alabaster, AL). Lipid solutions were prepared in chloroform and stored at – 20°C until use. NR and 3,3′-dioctadecyloxacarbocyanine perchlorate (DiO) were purchased from Sigma-Aldrich (St. Louis, MO, USA). Flipper-TR plasma membrane (FliptR) was purchased from Spyrochrome (Stein am Rhein, Switzerland). DOPE-Atto647N was purchased from AttoTech (Siegen, Germany). 4,4-Difluoro-5,7-dimethyl-4-bora-3a,4a-diaza-s-indacene-3-hexadecanoic acid (Bodipy C₁₆) was purchased from Thermo Fisher (Waltham, MA). Glucose and sucrose were purchased from Sigma-Aldrich (St. Louis, MO, USA). Bodipy C₁₂ was synthesized according to (32). The Annexin V-APC and propidium iodide (PI) Apoptosis Detection Kit was purchased Thermo Fisher Scientific (Waltham, MA).

Sample preparation and handling

GUVs were produced by the PVA-lipid hybrid film method (33) with minor modifications (34). A 2% weight solution of PVA was prepared in Milli-Q water, stored at room temperature, and used within 2 weeks. About 50–100 μL of this solution was spread on an untreated glass coverslip and the coverslip was placed onto a heat plate (∼60°C) until complete water evaporation, typically within 5 min. During the evaporation process, the PVA solution was continuously spread on the glass with the help of a pipette tip to form a homogeneous polymer film. Next, a 10-μL lipid solution (from a 3 mM stock) of a given lipid mixture was homogeneously spread onto the PVA film and chloroform was evaporated under a stream of argon. A pair of coverslips was then sandwiched between a Teflon spacer forming a ∼1.8-mL chamber volume and then closed with the help of clippers. A 200 mM sucrose solution was then added to the chamber for 20–30 min to swell the hybrid film of polymer-lipid. The formed GUVs were subsequently harvested and were ready for use. For samples in the Ld phase, the swelling step was performed at room temperature (∼21°C), whereas, for Lo samples, swelling was performed with preheated sucrose and left in the oven at 50°C until harvesting. Since all samples were fluorescently labeled (0.1–0.5 mol % of the respective dye), growth was performed in the dark. For imaging, the GUVs were typically diluted in a 1:1 volume in isotonic glucose, enabling vesicle sedimentation at the bottom of the chamber and facilitating imaging. The samples were placed between two thin coverslips separated by a ring rubber spacer creating a ∼100-μL chamber.

Multicolor confocal experiments

Multicolor confocal experiments were performed on a Zeiss LSM 710 scanning confocal microscope. Samples were imaged using a 40× (1.2 numerical aperture [N.A.]) W Korr M27 water immersion objective. Images consisted of 512 × 512 pixels with four line averages (unidirectional scanning) and 1 Airy unit. The dyes NR, FliptR, Bodipy C₁₂, and DiO were excited using a 488-nm argon laser and their emission detected in the range of 500–565 nm. Annexin V-APC was excited with a HeNe 633-nm laser and its emission detected in the range of 640–800 nm. PI was excited with a HeNe 543-nm excitation laser and its emission detected in the range of 555–600 nm. Each green dye was imaged simultaneously with Annexin V-APC and PI in the sequential scanning mode to minimize bleed-through. Laser intensity and detector gain were adjusted for maximum contrast without image saturation.

Fluorescence lifetime imaging and analysis

The radiative rate constant k_r is a function of the refractive index of the fluorophore’s environment, extinction coefficient ε, and the average emission wavenumber ν according to the Strickler-Berg equation (35):

$$k_r=2.88\times 10^{-9}{n}^2\frac{\int \mathit{I}\left(\mathit{\nu }\right)\mathit{d}\mathit{\nu }}{\int \mathit{I}\left(\mathit{\nu }\right){\mathit{\nu }}^{-\mathbf{3}}\mathit{d}\mathit{\nu }}\int \frac{\mathit{\epsilon }\left(\mathit{\nu }\right)}{\mathit{\nu }}d\nu$$ (Equation 2)

where n is the refractive index of the fluorophore’s environment, I is the fluorescence emission, ε the extinction coefficient, and v the wavenumber (v = λ⁻¹, λ wavelength). The n² dependence is empirical (36). If k_nr = 0 in Eq. 1, then τ = k_r⁻¹, which is an upper limit to the fluorescence lifetime.

FLIM was performed on an inverted microscope (Olympus IX73) equipped with time-correlated single-photon counting (PicoQuant). Samples were illuminated using a 100× (1.4 N.A.) oil immersion objective lens (UPLSAPO, Olympus). Emission was collected via the same objective and filtered from the excitation light by band or long-pass filters depending on the dye used. The green dyes FliptR, DiO, NBD, Bodipy C₁₂, and Bodipy C₁₆ were excited with a 481-nm laser and their emission (except for FliptR) was collected using a 525/50-nm band-pass filter. FliptR’s emission was collected using a 582/75-nm band-pass filter. DPPE-Rh and NR were excited with a 532-nm laser and their emission collected with a 594-nm large-pass filter. DOPE-Atto647N was excited with a 638-nm laser and its emission detected with a band-pass filter 690/70 nm. The images were acquired using the SymPhoTime 64 software and all samples were excited with a pulsed 20-MHz repetition rate. Unless stated otherwise, the samples were imaged with 128 × 128 pixels, 1-ms dwell-time, and ∼300 μm/pixels with typical acquisition times of 30 s. For analysis, the GUV membrane signal at the equator was manually selected for individual GUVs and all pixels used for fitting without binning. The fluorescence decays were fitted using an n-exponential tail fit with a single (n = 1) or bi-exponential (n = 2) decay model

$$\mathrm{I}\left(\mathrm{t}\right)={\mathrm{I}}_0\left({\mathrm{A}}_1{\mathrm{e}}^{\left(-\mathrm{t}/\tau 1\right)}\right)+\left({\mathrm{A}}_2{\mathrm{e}}^{\left(-\mathrm{t}/\tau 2\right)}\right)+\mathrm{B}$$ (Equation 3)

where I(t) is the intensity at time t and I₀ is the intensity at t = 0. A₁ and A₂ are pre-exponential factors associated with lifetime components τ₁ and τ₂, respectively, and B is the background. For FliptR, we used the long lifetime for analysis as it corresponds to environmental sensing (16).

We calculated the mean error (Err) using error propagation analysis from mean (M) and Err propagated from the calculated averages for lifetime ratio, difference, or intensity ratio according to

$$Err=\sqrt[2]{{\left(\frac{{Err}_{Ld}}{M_{Ld}}\right)}^2+{\left(\frac{{Err}_{L0}}{M_{Lo}}\right)}^2}$$ (Equation 4)

The subscripts refer to membrane phase state.

To study the effects of Triton X-100 (TX-100) on FliptR’s lifetime, a concentrated 5-μL volume (from a 50 mM stock solution) of TX-100 was added to the chamber and the GUVs were immediately observed. The externally added lipids were dissolved in DMSO (100 mM lipids) and small aliquots were added to the chamber containing the cells. DOPC was added up to 12 mM. Due to induced changes in cell morphology, SM:Chol was added up to 1 mM. TX-100 and lipids reach the vesicles from a distance via diffusion. The typical operation time between addition and image acquisition is approximately 10 s. The time between administration of the external agents and imaging was 10–15 min for each concentration. In the photobleaching experiments, GUVs labeled with 0.5 mol % Bodipy C₁₂ were imaged for FLIM before and after strong illumination with an X-Cite 120Q lamp through the GFP filter (band-pass 460- to 480-nm excitation and 495- to 540-nm emission) for 30 s.

The spectral shift experiments were carried out on multilamellar vesicles/liposomes labeled with 0.5 mol % each probe with a Jasco FP-8300 spectrofluorometer at room temperature. Multilamellar vesicles (MLVs) were prepared in 200 mM sucrose at 2 mM total lipid concentration and the measurements were carried out without further dilution. For each dye, the excitation was selected as identical to those used in the FLIM experiments.

Cell handling and labeling

THP1 cells, derived from human monocytic leukemia, were cultured in RPMI 1640 (Thermo Fischer Scientific, Waltham, MA, USA), supplemented with 10% fetal calf serum (Sigma-Aldrich, Zwijndrecht, the Netherlands) and 1% penicillin/streptomycin (p/s) (GIBCO) in a humidified incubator at 37°C and 5% CO₂. Cells were kept at 400,000 cells/mL and split with fresh medium every other day. For imaging, 400 μL of cells were seeded on μ-Slide 8 Well with no. 1.5 polymer coverslip bottom (Ibidi, Gräfelfing, Germany). The selected dyes were dissolved in DMSO (Sigma-Aldrich, Zwijndrecht, the Netherlands) at a concentration of 500 mM (DiO was dissolved at 50 mM) and added to the cells before imaging in the following dilutions at a final concentration of 0.5 mM and incubated for 15 min in a humidified incubator at 37°C and 5% CO₂. The cells were imaged at room temperature after the labeling period for a maximum period of 1 h without further removing the dyes.

Viability experiments

Annexin-V-APC, PI staining, and trypan blue

A volume of 400 μL of cells per condition was harvested and washed with PBS (Sigma-Aldrich, St. Louis, MO), spun down at 450 relative centrifugal force (RCF), and resuspended in PBS and 10% AnnexinV-APC binding buffer (Thermo Fisher Scientific, Waltham, MA). The cells were then incubated with the respective probes along with both Annexin V-APC (4 μL) and PI (4 μL) and incubated for 15 min in a humidified incubator at 37°C and 5% CO₂. For the labeled cells, viability was further tested with trypan blue staining. Trypan blue (Thermo Fisher Scientific, Waltham, MA) was added to the wells at 1:10 volume and the cells were imaged in bright field. Tripan blue (TP)-positive cells were identified based on their darker appearance.

Flow cytometry

Viability of the dye was assessed using LSR-II cytometer (BD Biosciences, San Jose, CA, USA). Four-hundred microliters of labeled cells were incubated for 15 min in a humidified incubator at 37°C and 5% CO₂ with DMSO control. Samples were acquired using the DIVA 8.0 software. Analysis was performed on FlowJo Vx.0.7 (Tree Star, Oregon, USA). Gating round live cells, a distinct population in the scatter plot with forward scatter (FSC-A) on the x axis and side scatter (SSC-A) on the y axis was defined and kept for all conditions, and the percentage of counts in the gate was used as a measure of viability.

Data analysis

All images were analyzed using Fiji (NIH, USA). The plots were generated using OriginLab (Northampton, MA), versions 6 and 2023b. Unless stated otherwise, means and standard deviation were used to compare samples.

Results

We use FLIM to assess the biophysical properties of synthetic and biological membranes in combination with probes with known environmental responses, as well as widely used commercially available fluorescent dyes with no reported environmental response. The selected dyes and their most common use in membrane biology are shown in Fig. S1 and include the rigid solvatochromic dyes NR, NBD, as well as dyes with no reported environmental response such as Atto647 (hereafter Atto), lissamine-rhodamine, Rh (the last three conjugated to phospholipid headgroups), Bodipy C₁₆, and DiO. As flexible dyes, we used the flipper molecules FliptR and the fluorescent molecular rotor Bodipy C₁₂. Thus, the model probes used belong to the three classes of environmental sensors. By biophysical properties, we refer to any mesoscale intrinsic material property of membranes that could be sensitively sensed by changes in fluorescence lifetime and/or intensity for probes located in well-characterized environments. Alternatively, lifetime responses to more specific biophysical properties such as hydration and viscosity are sensed by well characterized and specific sensors such as NR and Bodipy C₁₂, respectively. In this work, we thus use these terms interchangeably. Fig. 1 shows the molecular mechanisms by which they change their optical properties in response to the local environment. In general, solvatochromic dyes are assayed by monitoring their spectral changes (e.g., color), including GP measurements, whereas rotors and flippers are assayed based on changes in their lifetime. Although such characterization is important as it widens the range of fluorescent probes that could be used in environmental sensing, including dyes with improved optical properties (e.g., higher QY, photostability), it is also important so as to characterize the magnitude of sensing for dyes that should be insensitive to environmental factors.

Figure 1.

Environmentally sensitive reporter dyes respond to environmental changes. Liquid disordered (Ld) membranes made of unsaturated lipids (e.g. DOPC) and liquid ordered (Lo) membranes made of mixtures of fully saturated lipids and cholesterol (Chol) exhibit very distinct material properties. Reporters that sense these properties can be reconstituted in these membranes and change in lifetime (τ) in response to various environmental cues. The changes in lifetime can be represented by a color-coded scheme in which short and long lifetimes are denoted by blue to red colors, respectively.

Sensitivity in homogeneous membranes

To assess environmental response sensitivity to membrane properties, we measured the fluorescence decay of a series of fluorescent dyes reconstituted in small fractions in GUV model membranes. The phase diagram of an unsaturated, a fully saturated, and a sterol lipid exhibits a rich behavior of membrane phases, with unsaturated lipids alone forming a homogeneous Ld phase, and binary mixtures of saturated lipids and cholesterol forming a homogeneous Lo phase (24). These membranes exhibit large differences in bending rigidity (37), packing (38), and hydration (20) and therefore they provide an excellent testing system for environmental sensing. We thus prepared homogeneous GUVs with very distinct mechanical properties made of either DOPC or mixtures of 1,2-dipalmitoyl-sn-glycero-3-phosphocoline (DPPC) and cholesterol (7:3 mol ratio), forming respectively very fluid Ld and very packed (i.e., viscous) Lo membranes. Fluorescence decays were detected on single pixels forming a color-coded lifetime map, and the data were analyzed with least-squares fitting with a single or multiexponential decays (depending on the probe and/or membrane phase). Fig. 2 A shows FLIM images and the respective fluorescence decays for Ld and Lo GUV membranes labeled with NR, and Fig. S2 shows results with several other probes studied. The fluorescence decays for NR were fitted with a monoexponential decay function (Eq. 3) and the randomly distributed flat residuals and chi square close to 1 indicate an excellent fit. FliptR exhibited a double-exponential decay for both membranes, as previously reported (16,17). Bodipy C₁₂ exhibited a single-exponential decay in Ld membranes and double-exponential decay in Lo membranes, as previously observed (28); we use its long decay for analysis and viscosity calculations (see below). All other dyes for both membrane compositions exhibited a single-exponential decay.

Figure 2.

Fluorescence lifetime sensing of various fluorescent probes. (A) Representative FLIM images of GUVs in the Ld (DOPC) and Lo (DPPC:Chol, 7:3 mol) phases labeled with 0.5 mol % NR along with their fluorescence decays, monoexponential fits, and residuals. Inset, fluorescence intensity images for the respective FLIM images obtained under identical conditions. Scale bar: 9 μm. In (B), measured fluorescence lifetimes for all dyes tested in both membrane phases. Each circle represents a measurement on an independent GUV. For each probe in Ld and Lo phases, a paired-sample t-test was performed and the level of difference in significance at the 0.05 confidence interval is shown. Inset: histogram distribution of measured membrane viscosity using Bodipy C₁₂ as a reporter. (C) Mean lifetime ratio, (D) dynamic range (lifetime difference between Ld and Lo phase), and (E) intensity ratio for the measurements in (B). Error bars are not visible due to their small values. Red circle and green square data represent higher measured values in Lo and Ld, respectively. Note that longer lifetime in one phase does not necessarily correspond to higher brightness for that phase.

As for many dyes, there is a large difference between Ld and Lo decays for most (but not all) dyes in the more packed Lo membranes as expected for NR, FliptR, and Bodipy C₁₂. Strikingly, DiO, a dye with no reported sensing capability, also very sensitively detects the different phases (Fig. S2). The lower fluorescence intensity at the GUV equator is due to the orientation of the transition dipole moment parallel to the membrane and the horizontal polarization of the laser excitation due to photoselection (39), discussed below.

We performed a similar analysis on a total of eight different fluorescent dyes, five of which are known environmental sensors (NR, FliptR, NBD, Rh, and Bodipy C₁₂) and three with no report of environmental sensing (DiO, Atto, and Bodipy C₁₆). Table 1) summarizes the lifetimes of a series of fluorescent probes in homogeneous membranes along with reported literature data using the same dyes in similar membranes. As can be seen in Fig. 2 B, most dyes, including dyes with unknown sensing capabilities, change their lifetimes in response to the different membrane environments, although with large differences in their intrinsic lifetimes and dynamic ranges.

Table 1.

Reported lifetimes of different probes in homogeneous membranes

Probe	Membrane Composition	Membrane Phase	Temperature	Lifetime (ns)	Reference	Dynamic Range (ns)
NR	DOPC	Ld	18°C	3.9 ± 0.1	this study	3.5
NR	DOPC	Ld	23°C	∼2.9	aMukherjee (40)
NR	DOPC:Chol (9:1)	Ld	23°C	3.34	Mukherjee
NR	DOPC:Chol (8:2)	Ld	23°C	3.51	Mukherjee
NR	DOPC:Chol (7:3)	Ld	23°C	3.77	Mukherjee
NR	DOPC:Chol (6:4)	Lo	23°C	3.62	Mukherjee
NR	DPPC:Chol (7:3)	Lo	21°C	7.4 ± 0.2	this study
NR	DMPC	solid	25°C	4.1 ± 0.4	Jana (41)
NR	DMPC	solid	25°C	3.5 ± 0.3	Jana
FliptR	DOPC	Ld	18°C	4.2 ± 0.5	this study	2.3
FliptR	DOPC	Ld	not specified	3.8 ± 0.1	Colom (16)
FliptR	DOPC:Chol (6:4)	Lo	not specified	5.3 ± 0.1	Colom
FliptR	DPPC:Chol (7:3)	Lo	18°C	6.5 ± 0.1	this study
FliptR	SM:Chol (7:3)	Lo	not specified	6.4 ± 0.1	Colom
FliptR	POPC:SM:Chol (57:14:29)	close to phase separation	not specified	∼5.4	Colom
FliptR	DOPC:SM:Chol (30:30:40)	close to phase separation	not specified	∼4.6	Colom
PM flipper	DOPC	Ld	RT	∼2.8	Goujom (42)
PM flipper	SM:Chol (1:1)	Lo	RT	∼5.9	Goujom
Lyso flipper	DOPC	Ld	RT	∼3.1	Goujom
Lyso flipper	SM:Chol (1:1)	Lo	RT	∼6.6	Goujom
ER flipper	DOPC	Ld	RT	∼3.0	Goujom
ER flipper	SM:Chol (1:1)	Lo	RT	∼5.8	Goujom
Mito flipper	DOPC	Ld	RT	∼2.9	Goujom
Mito flipper	SM:Chol (1:1)	Lo	RT	∼5.9	Goujom
DiO	DOPC	Ld	18°C	0.9 ± 0.1	this study	0.4
DiO	DPPC:Chol (7:3)	Lo	18°C	1.3 ± 0.1	this study
NBD	DOPC	Ld	18°C	7.7 ± 0.2	this study	2.4
NBD	DOPC	Ld	25°C	7.1 ± 0.2	bStöckl (43)
NBD	DOPC:DOPS (1:1)	Ld	25°C	∼6	Stöckl
NBD	DPPC:Chol (7:3)	Lo	18°C	10.1 ± 0.7	this study
NBD	DOPC:SM:Chol (8:1:1)	Ld	25°C	7.6 ± 0.1	Stöckl
NBD	DOPC:SM:Chol (4:4:2)	Lo	25°C	7.3 ± 0.1	Stöckl
Liss-Rhodamine	DOPC	Ld	18°C	2.9 ± 0.1	this study	0.5
Liss-Rhodamine	DOPC:Chol (7:3)	Lo	18°C	2.4 ± 0.1	this study
Liss-Rhodamine	DPPC	solid	24°C	∼2.6	cAlmeida (44)
Liss-Rhodamine	DOPC:Chol (4:1)	Lo	24°C	∼2.5	Almeida
Liss-Rhodamine	DPPC:Chol (7:3)	Lo	24°C	∼2.5	Almeida
Atto 647N	DOPC	Ld	18°C	3.2 ± 0.1	this study	0.1
Atto 647N	DPPC:Chol (7:3)	Lo	18°C	3.0 ± 0.1	this study
Bodipy C16	DOPC	Ld	18°C	5.8 ± 0.6	this study	0.1
Bodipy C16	DPPC:Chol (7:3)	Lo	18°C	5.7 ± 0.1	this study
Bodipy C12	DOPC	Ld	18°C	2.0 ± 0.1	this study
Bodipy C10/C12	DOPC	Ld	RT	∼1.8	dWu (28)/Dent (45)
Bodipy C12	DPPC:Chol (7:3)	Lo	18°C	4.3 ± 0.3	this study	0.8
Bodipy C10/C12	DOPC:Chol (6:4)	Lo	RT	∼1.9	Wu
Bodipy C10/C12	SM	solid	RT	∼5.1	Wu
Bodipy C10/C12	SM:Chol (6:4)	Lo	RT	∼4.8	Wu

The dynamic range is defined as the difference between the long and short lifetimes. RT, room temperature.

^aThe short lifetime is more sensitive to Chol content and used as an environmental reporter.

^bLong lifetime.

^cAmplitude-weighted lifetime.

^dSingle decay in DOPC and double-decay in DOPC:Chol (7:3) GUVs.

The fluorescence lifetime depends on the radiative rate constant k_r and nonradiative rate constant k_nr (Eq. 1), in which k_r is a function of the refractive index of the fluorophore’s environment (n), extinction coefficient ε, and the average emission wavenumber ν (see Eq. 2). If changes to the refractive index of the membrane as it goes from Ld to Lo are negligible and if ε is constant, then only a spectral shift between the emission in the Ld and Lo phase can affect k_r. However, the magnitude of this effect is relatively small (46), as for all probes tested (Fig. S3). Thus, a significant change in fluorescence lifetime can be safely attributed to the nonradiative rate constant k_nr.

To assess environmental sensitivity, we calculated both the lifetime ratio and lifetime difference (dynamic range) for each probe in the Ld and Lo membranes; larger ratio or difference indicates higher environmental sensing (Fig. 2 C and D). The color codes represent higher values in a given phase—red circles for higher lifetime in Lo and green squares in the Ld, respectively. When looking at the lifetime ratio data, the molecular rotor Bodipy C₁₂ has the strongest sensitivity from all probes tested. Strikingly, the ratio of DiO is comparable to that of very sensitive dyes and even higher than that of the well-known solvatochromic probe NBD. Rh exhibits only moderate sensing, with the Lo phase showing a slightly shorter lifetime than the Ld phase, as previously reported (44), in contrast to the other dyes. This may be due to hydration, i.e., water molecules penetrating the membrane in the Ld phase, which are excluded in the Lo phase. Water has a lower refractive index than DPPC or DOPC membranes, and a lower refractive index in the immediate environment of the dye in the Ld phase may lead to a longer fluorescence lifetime (47), because the radiative rate constant k_r is lower, if there is no spectral shift and k_nr is not affected. This depends on the position of the dye in the membrane (47,48). Atto and Bodipy C₁₆ are weakly sensitive to environmental differences. Using the molecular rotor Bodipy C₁₂ as a viscosity sensor and previously reported calibration curves (32), the viscosities of the Ld and Lo phases are 260 ± 10 cP and 1330 ± 440 cP, respectively. Thus, ordered membranes composed of DPPC:Chol (7:3 mol) are approximately five times more viscous than their fluid DOPC counterparts. Given the very diverse dynamic ranges of the probes tested, we also analyzed the lifetime differences (dynamic range) between the Ld and Lo phases for all probes. Compared to lifetime ratios, this provides a more sensitive assessment of the probe’s response as it takes into account the real changes in lifetime. As shown in Fig. 2 D, NR exhibits the highest dynamic range of all probes tested. Similarly, other long-lifetime dyes are also highly sensitive in discriminating different environments. In contrast, DiO, which exhibits an intrinsically short lifetime, has a poorer environmental discrimination, thus DiO is less likely to resolve mild environmental differences (see section “real-time changes in membrane biophysical properties”).

Environmental sensors are known to change both their lifetimes and QY in response to the environment, with a general increase in both lifetime and QY in more ordered membranes (13). This is due to the relationship between fluorescence lifetime τ and QY ϕ:

$$\varphi =\frac{k_r}{k_r+k_{nr}}=k_r\tau$$ (Equation 5)

We also assessed changes in QY by measuring the intensity of the probes on the membrane of Ld and Lo GUVs. As can be seen in Fig. 2 A (inset) and Fig. 2 E, there are also clear changes in intensity at otherwise identical probe concentration and imaging settings for most dyes, but the intensity responses do not always mirror lifetimes. Moreover, there is no clear relationship between exhibiting a longer lifetime and intensity in a given phase—some probes with long lifetime in the Lo phase also exhibit an increase in brightness in this phase (e.g., FliptR), whereas others are brighter in the Ld phase (e.g., NR). It was not possible to measure the ratio using Atto due to its inhomogeneous reconstitution in Lo membranes. These results bring about two important aspects: first, the changes in lifetime are not always mirrored by changes in intensity. Second, due to their environmental sensing, the probes may exhibit dramatically different brightness at identical concentrations. This is important, for instance, in the determination of partition coefficient based solely on intensity and should be accounted for. For this reason, we did not analyze the data in terms of intensity differences as environmental responses and changes in QY are not separated. Lastly, in conditions of FRET, changes in donor intensity (in the presence of a FRET acceptor) may also arise from changes in the environment. Although these issues are hard to disentangle in intensity analysis, they can be circumvented by time-resolved experiments (49,50).

We have observed that there is preferential excitation of some of the probes and this depends on membrane phase state, an effect that is manifested both as intensity as well as lifetime responses. Laser excitation light is typically polarized, and the fluorophore is only excited if the electrical vector of the excitation light is not perpendicular to the transition dipole moment of the fluorophore. The excitation probability scales with cos²θ, where θ is the angle between the electric vector of the excitation light and the transition dipole moment of the fluorophore. Thus for θ = 0°, the fluorophore is excited with maximum probability and the fluorescence is brightest, whereas, for θ = 90°, no excitation takes place (51). The fluorescence decay is also affected by photoselection, an aspect that is utilized in time-resolved fluorescence anisotropy measurements (52). If the sample is excited with collimated polarized light, the fluorescence decays parallel and perpendicular to that of the excitation are given by

$$I_{\Vert }\left(t\right)=\frac{1}{3}{I}_0\mathit{exp}\left(-\frac{t}{\tau }\right)\left[1+2r_0\mathit{exp}\left(-\frac{t}{\theta }\right)\right]$$ (Equation 6)

for the parallel decay, and the perpendicular decay is given by

$$I_{\perp }\left(t\right)=\frac{1}{3}{I}_0\mathit{exp}\left(-\frac{t}{\tau }\right)\left[1-r_0\mathit{exp}\left(-\frac{t}{\theta }\right)\right]$$ (Equation 7)

where I is the fluorescence intensity, t is the time, r₀ is the initial anisotropy, and θ is the rotational correlation time (53). For an isotropic distribution of fluorophores, the polarization effect can be eliminated by measuring the fluorescence decays at the “magic angle” of 55° to the polarization of the excitation (54). For fluorophores rigidly oriented in the membranes (i.e., in the Lo phase), this can lead to an apparent longer fluorescence lifetime at a perpendicular orientation of polarization of the excitation and the transition dipole moment of the fluorophore; see Eq. 7. Fig. 3 shows such effects for FliptR and NBD. In addition to being highly sensitive to membrane phase (longer lifetime in Lo membranes), both probes exhibit clear photoselection, especially in the more viscous Lo membranes. Furthermore, NBD exhibits two clearly different lifetimes depending on the membrane segment (orientation), whereas FliptR does not. Once more, such effects are much more pronounced in the more viscous Lo phase where the rotational mobility of the probe is reduced. In general, most dyes have some degree of photoselection (Figs. S4–S6) and, in practice, such lifetime differences may give rise to multiexponential decays if the signal from the entire object is binned.

Figure 3.

Photoselection effect on Ld and Lo membranes. (A) The measured fluorescence lifetime at the horizontal (H) and vertical (V) planes of Ld and Lo GUVs labeled with FliptR and PE-NBD. The laser polarization is in the direction of the scan (horizontal). Scale bar: 9 μm. (B) The fluorescence decays and their respective fitting and residuals for the horizontal (gray) and vertical (green) planes on the GUVs in the Lo phase in (A). (C) individual lifetime measurements in the horizontal and vertical planes of the GUVs for both Ld (green) and Lo (red) membranes. (D) The ratio of the horizontal and vertical (R_H/V) fluorescence lifetime measured on several GUVs in the Ld and Lo phases. The horizontal dashed line represents ratio = 1, which corresponds to no photoselection. Each point represents a measurement on a single GUV. Standard deviation and error of the mean are shown in blue.

Sensitivity in heterogeneous membranes

Although most of the experiments above could have been performed purely spectroscopically (i.e., using cuvette experiments with small liposomes), imaging methods have the potential to unravel local environmental heterogeneities that are not accessible with spectroscopic methods alone. An exception was the direct observation of polarization effects, which would have been averaged out in bulk measurements. We then set out to study whether it is possible to assess the physical properties of domains in phase-separated GUVs. Domains are typically identified based on differential intensities of bona fide markers thanks to their differential partition in a given phase. The disadvantage is that some probes may have similar partitions, making identification difficult if the domains have comparable brightness, requiring other alternatives for identification; alternatively, they may change their QY depending on the environment. We labeled phase-separated GUVs with probes that respond to changes in membrane compositions, as well as Bodipy C₁₆ as a nonresponsive probe. From intensity measurements, it is difficult to resolve the different domains with DiO as a marker (Fig. 4 A) as both phases exhibit similar brightness. Unlike intensity, FLIM shows a clear contrast between the phases, with a longer lifetime in the more ordered Lo phase (Fig. 4 B). As in homogeneous membranes, most probes exhibit a longer lifetime in the Lo phase (Fig. 4 C).

Figure 4.

FLIM can resolve spatial changes in membrane properties in phase-separated GUVs. (A) A DiO-labeled GUV exhibiting a large Ld and Lo domain as identified based on DiO’s lifetimes. Scale bar: 9 μm. (B) fluorescence decays, fit, and residuals from the fit for the vesicle shown in (A). (C) Measurements on both domains for several independent GUVs for six different probes. (D) Ratio between the long and short lifetime for both domains measured on the same individual vesicles. (E) Histogram of the calculated membrane viscosities for Ld (green) and Lo (red) domains using Bodipy C₁₂ as a reporter. FliptR and Bodipy C₁₂ fluorescence decays in both domains were fitted with two exponentials. The viscosity shown in (E) was calculated using the long decay time. For all other probes, data were fitted with a single-exponential decay.

As expected, probes that sensitively detect different environments in homogeneous membranes also sensitively detect Ld and Lo domains in phase-separated GUVs (Fig. 4 C and D), with DiO having an outstanding performance compared to well-known responsive dyes. Due to the environmental response of most probes, we did not further characterize the changes in QY and limited our analysis to lifetime measurements only. Table 2 summarizes the lifetimes of a series of fluorescent probes in phase-separated membranes. The lifetimes in the Ld and Lo domains of all probes tested have intermediate values compared to their lifetimes in homogeneous membranes—the different phases have somewhat intermediate properties, as expected (26,38). Using Bodipy C₁₂ as the viscosity sensor, the viscosities of the Ld and Lo phases in phase-separated GUVs are ∼560 ± 170 cP and ∼1190 ± 390 cP, and hence the differences in their mechanical properties are also more subtle compared to homogeneous membranes.

Table 2.

Reported lifetimes of different probes in phase-separated membranes

Probe	Membrane Composition	Membrane Phase	Temperature	Lifetime (ns)	Reference
NR	DOPC:Chol:DPPC (1:1:1)	disordered	18°C	3.5 ± 0.1	this study
NR	DOPC:Chol:DPPC (1:1:1)	ordered	18°C	5.8 ± 0.4	this study
FliptR	DOPC:Chol:DPPC (1:1:1)	disordered	18°C	4.4 ± 0.3	this study
FliptR	DOPC:Chol:DPPC (1:1:1)	ordered	18°C	5.9 ± 0.1	this study
DiO	DOPC:Chol:DPPC (1:1:1)	disordered	18°C	0.9 ± 0.1	this study
DiO	DOPC:Chol:DPPC (1:1:1)	ordered	18°C	1.3 ± 0.1	this study
NBD	DOPC:Chol:DPPC (1:1:1)	disordered	18°C	8.6 ± 0.2	this study
NBD	DOPC:Chol:DPPC (1:1:1)	ordered	18°C	9.9 ± 0.7	this study
NBD	DOPC:Chol:DPPC (3:3:4)	disordered	25°C	7.1 ± 0.4	Stöckl
NBD	DOPC:Chol:DPPC (3:3:4)	ordered	25°C	11.6 ± 0.1	Stöckl (43)
Bodipy C16	DOPC:SM:Chol (1:1:1)	disordered	18°C	6.1 ± 0.1	this study
Bodipy C16	DOPC:SM:Chol (1:1:1)	ordered	18°C	5.8 ± 0.1	this study
Bodipy C12	DOPC:Chol:DPPC (1:1:1)	disordered	18°C	3.0 ± 0.3	this study
Bodipy C12	DOPC:Chol:DPPC (1:1:1)	ordered	18°C	4.5 ± 0.6	this study

Real-time changes in membrane biophysical properties

Due to the required high number of photons necessary for sufficient statistics, FLIM is usually significantly slower than steady-state confocal imaging, a requirement that also poses consequences to sensing subtle lifetime changes. We bridge these two aspects by studying the biophysics of Ld and Lo GUV domains in real time upon the action of Triton X-100 (TX-100), a model detergent whose effects on membranes are well characterized (55). In homogeneous membranes, TX-100 efficiently inserts into the Ld phase, decreasing packing, increasing membrane area and fluctuations, and eventually causing pore formation and solubilization due to a reduction in edge tension (56,57). Conversely, TX-100 exhibits low insertion in homogeneous Lo membranes, making them resistant to solubilization (27). In phase-separated GUVs, TX-100 induces domain coalescence, budding, and solubilization of the Ld domains, as well as a decrease in the area of Lo domains by removal of low-melting-temperature lipids therein (58), leaving an insoluble and presumably more ordered phase behind. Although its morphological effects are well characterized, changes in membrane properties, especially in the course of membrane solubilization, are unknown.

The typical morphology of a phase-separated GUV in the absence of Triton X-100 is shown in Fig. 5 A. Upon injecting a concentrated solution of the TX-100 (5 mM final concentration upon equilibration), we followed its effects on phase fluidity from the beginning by measuring the changes in lifetime in both Ld and Lo phases at a temporal resolution of 24 s/frame. Fig. 5 and Video S1 show the lifetime response using FliptR as the environmental reporter (Video S2 shows a GUV labeled with NR). At time 0 s, both domains are clearly resolvable due to differences in FliptR lifetime. As the local concentration of TX-100 increases, Ld area increases, followed by an increase in membrane fluidity (decrease in lifetime). The reduction in intensity is a result of lipid removal, as previously observed with several other probes (56,57,59). On the other hand, the order of the Lo phase increases, most likely as a result of removal of low-melting-temperature DOPC from the ordered Lo phase (although we cannot rule out the solubilization of submicroscopic Ld domains from within the Lo phase). Domain coalescence, bulging, splitting, pore formation, and eventual solubilization of smaller Ld domains originating from interactions of the detergent with Lo phases are also observed (Fig. S7).

Figure 5.

Real-time changes in the packing of Ld and Lo domains induced by TX-100. (A) FLIM time-lapse images of a phase-separated GUV labeled with 0.5 mol % FliptR. The numbers refer to the time from the beginning of the video. Scale bar: 10 μm. Images are zoomed-in snapshots of Video S1. (B) fluorescence decays and their respective fits and residuals for the different domains measured at the time points shown in (A). (C) Temporal evolution of FliptR’s lifetime for both Ld and Lo domains for the GUV shown in (A). The asterisks correspond to the times in (A). (D) Mean changes in lifetime, and associated errors (s.d.) for both domains for eight individual GUVs.

Video S1. Addition of TX-100 (final concentration 5 mM) to phase-separated GUVs (DOPC:DPPC:Chol, 1:1: mol) labeled with 0.5 mol % FliptR

Note an increase in size of the Ld domains along with Ld domain fluidification (lifetime color change from green to blue) until full domain solubilization. The decrease in Ld brightness is due to solubilization of the FliptR probe from these membranes rather than to photobleaching. These morphological changes are accompanied by an increase in the order of the Ld domains (lifetime color change from light orange to red). The Lo domains are not solubilized by TX-100 and thus remain as the detergent-insoluble fractions of the membranes. Zoomed-in snapshots and GUV dimensions can be seen in Fig. 5 in the main text.

Video S2. Addition of TX-100 (final concentration 5 mM) to phase-separated GUVs (DOPC:DPPC:Chol, 1:1: mol) labeled with 0.5 mol % NR

At the beginning of the video, the Ld (green, short lifetime) and Lo (red, long lifetime) domains had already been separated. As the local TX-100 increases, the Ld is fully solubilized, whereas the Lo phase is resistant to solubilization. Note that a small Ld is expelled from the larger Lo phase (∼313.6 s). Time stamp in seconds.

The temporal changes in membrane fluidity can be well resolved from the fluorescence decays (Fig. 5 B and dynamics in 5C). The response is very reproducible (Fig. 5 D). The lifetime in the Lo domain increases and stabilizes to a maximum (∼5.8 ns) and does not change further, presumably because all DOPC has been removed from this phase. The lifetime in the Ld decreases to a minimum (∼3.3 ns) until the domain is fully solubilized. Of note, such changes are reproduced using other probes (Figs. S7–S9), showing the universality of these responses. These experiments, wherein environment changes are mild, require highly sensitive probes and ideally probes with comparable partition in both phases. In that sense, both FliptR and NR (Fig. S7) are good reporters. Due to its poor sensitivity, DiO does not sensitively respond to such changes (Fig. S8). Although Bodipy C₁₂ has good sensitivity, its strong partition in the Ld phases makes it less suitable due to the low signal in the Lo phase (Fig. S9). Hence, the experiments show that milder changes in membrane packing can be dynamically and sensitively detected provided highly sensitive probes with more homogeneous partitions are used.

Extended imaging in time-lapse experiments may result in probe photobleaching, potentially producing reactive species that can oxidize lipids, especially unsaturated ones (60). Thus, although FLIM should not be sensitive to photobleaching alone, production of reactive species may effectively change the local environment (61), which can be monitored by changes in lifetime. We next tested for a potential deleterious effect of strong illumination for extended periods on the Ld or Lo phases in GUVs doped with Bodipy C₁₂ with strong light power. Strong illumination causes a significant reduction in fluorescence intensity (not shown), along with an increase in lifetime for both phases (Fig. S10 A). In GUVs made of unsaturated DOPC lipids, bleaching leads to an increase of viscosity from 411 ± 38 cP to 891 ± 75 cP, comparable to observed values of membranes containing purely oxidized lipids (61). In GUVs made of a mixture of the long and fully saturated lipid DPPC and Chol (DPPC:Chol, 7:3 ratio) membranes, bleaching increased viscosity from 916 ± 38 cP to 1374 ± 93 cP. When the ratio of viscosity after/before bleaching is compared (R_Bef/Aft; Fig. S10 C–E), Ld DOPC membranes containing unsaturated lipids are more prone to such effects than their Lo DPPC:Chol counterparts. Altogether, the results show that 1) strong illumination induces environmental modifications in the membrane but to a different extent, 2) the changes in viscosity can be sensed by molecular sensors, and 3) the very sensor, when illuminated with strong light and for extended periods, has the potential to modify the local physical-chemical properties of the environment, similarly to previously reported results with oxidation-prone molecules (62).

Sensitivity in living cells

We next tested the ability of a selected set of probes to assess the biophysical properties in the membranes of living cells. Previous studies have shown that the plasma membrane (PM) is the most ordered (and most viscous) membrane in the cell due to its high fraction of long and saturated phospholipids as well as cholesterol (63). In fluorescent labeling, many plasma membrane dyes are very quickly internalized and the PM signal is quickly lost (64). Hence, in addition to checking the sensors’ performance as a reporter of mechanics and spatial heterogeneities, we also compared their intracellular distribution.

As shown previously by others (64), NR is readily internalized by cells (intensity image in Fig. 6 A). However, despite significant internalization, FLIM clearly shows variations in NR lifetime, longer at the PM (4.8 ± 0.2 ns) and shorter in intracellular membranes (4.0 ± 0.1 ns), as can be seen from the FLIM image in Fig. 6 A. The values are slightly higher than whole-cell measurements reported previously (64). FliptR stains the PM for extended periods (16) (Fig. 6 B), albeit with a lower but detectable intracellular staining. Some cells have bright intracellular puncta (Fig. S11), or PM protrusions of lower lifetimes, indicative of higher fluidity. DiO distributes homogeneously in cells with no obvious lifetime differences between PM and intracellular regions (Fig. S12). Bodipy C₁₂ exhibits intermediate internalization as well as a clear distinction in PM (4.7 ± 0.2 ns) and intracellular (3.9 ± 0.4 ns) lifetimes (Fig. 6 C). Fig. 6 D shows the lifetime for these probes measured at the PM and at the intracellular membranes, with lifetime ratios from these two environments shown in Fig. S13. The FliptR lifetime measured at the PM is 5.6 ± 0.2 ns, similar to previous results (16). The lifetime distribution of NR and FliptR in Ld GUVs (Fig. 2) and in intracellular membranes (Fig. 6 D) is similar, but the measured lifetime is longer at the PM. For Bodipy C₁₂, we find a higher lifetime in both intracellular membranes and the PM in cells (Fig. 6 D) compared to Ld GUVs, showing that the membrane composition and perhaps cellular milieu affect the effective viscosity that the Bodipy C₁₂ fluorophore senses. Using Bodipy C₁₂ as the reporter, the measured viscosities of the PM and intracellular membranes are ∼1379 ± 136 cP and 1000 ± 45 cP, respectively. When compared to model membranes, the PM has a similar viscosity to the Lo phase, whereas the viscosity of intracellular organelles is almost twice as high as the highly fluid Ld phase.

Figure 6.

Fluorescence lifetime and intracellular distribution of several probes in live THP1 cells. (A–C) Representative intensity (top left) and lifetime (top right) images of cells labeled with (A) NR, (B) FliptR, and (C) Bodipy C₁₂. The fluorescence decay, fit, and residuals are shown for selected regions. Scale bars: 7 μm. (D) PM and intracellular membrane lifetimes for several cells labeled with the three probes. Mean and standard deviation are also shown. Inset: calculated viscosities of intracellular organelles (green) and PM (red) using Bodipy C₁₂ as the viscosity reporter. Measurements were performed are room temperature (18°C ± 1°C).

Changes in membrane packing in living cells

We next studied whether modulating the composition of cellular membranes by externally added molecules could result in changes in membrane properties to demonstrate FLIM’s capability to resolve mild changes in membrane environment at subcellular level using a probe with high lifetime discrimination. We chose FliptR as a reporter based on its high sensitivity and stable PM labeling, whereas its modest internalization still allows robust measurements in intracellular membranes, with constant lifetime in control cells without treatment with the external additives for extended time periods (not shown), in agreement with prior studies (16,65). Hence, changes in packing upon the addition of the additives can potentially be visualized in different regions of the cell. To study the effects of externally added membrane-active molecules, we used TX-100 as a model membrane pore-forming agent that can easily penetrate and permeabilize cells (66); DOPC, a lipid that forms very fluid membranes; and binary mixtures of SM:Chol that form very viscous membranes. The maximal working concentrations were chosen based on the ability of cells to preserve their typical quasi-round shape (see bright-field images in Fig. 7 A–D). Concentrations that lead to clear morphological changes (Fig. S14) were not considered for analysis. TX-100 leads to a small decrease in the packing of both PM and intracellular membranes (Fig. 7 A, B, and E). Nevertheless, the effects are only observed at concentrations significantly higher than those used in GUVs, indicating that cells are more resistant to TX-100 and cells only exhibit signs of solubilization (i.e., large PM pores) at TX-100 concentrations much beyond the concentrations that fully solubilize the GUVs (not shown). The addition of 3 mM DOPC fluidizes both the PM and intracellular membranes (Fig. 7 C and F). At higher concentrations, the PM is further fluidized, whereas the effects in intracellular membranes level off. DOPC effects are much more pronounced than those for TX-100. We interpret this as a more efficient integration in cellular membranes given its lower critical micelle concentration. The addition of SM:Chol led to a decrease in membrane packing, but to a much lesser degree (Fig. 7 C and G). These results are somewhat surprising, and, although beyond the scope of this study, this response could potentially contain a contribution from SM:Chol on membrane refractive index, increasing the radiative rate constant k_r. Higher concentrations led to visible morphological changes and were not investigated further. In summary, FLIM and environmental sensors can map the local biophysical changes of multiple membranes in living cells upon the addition of membrane-perturbing agents.

Figure 7.

Effect of membrane-incorporating additives on PM and intracellular membrane packing in live THP1 cells labeled with FliptR. Representative cells in the presence of (A) no external additive, (B) DOPC, (C) TX-100, and (D) SM:Chol. Left and right columns are FLIM and corresponding bright-field images. Scale bar: 9 μm. (E–G) Mean and standard deviation of PM (red) and intracellular membrane (green) lifetime for increasing concentrations of (E) TX-100, (F) SM:Chol (7:3, mol), and (G) DOPC. Each measurement contains at least 10 individual cells. The experiments were performed at room temperature (18°C ± 1°C).

Labeling with environmental-sensitive probes does not induce cell cytotoxicity

For an effective use as an environmental sensor, the probe must be inert to cells. Cells labeled with each of the probes separately were co-stained with Annexin V-APC and PI as reporters for early and late apoptosis, respectively (67). As a control, we loaded the cells with the equivalent amount of DMSO. The assay is not suitable for cells labeled with NR given its broad fluorescence emission. Therefore, as an additional assay, we labeled cells with trypan blue (TB), a nonfluorescent probe that penetrates dead cells, creating a dark appearance under bright-field illumination (Fig. S15 A–E). The fraction of positively labeled cells (X_positive) for cells labeled with all tested dyes is similar to the control cells (Fig. S15 E). Additionally, the back and forward scattering of cells in a flow cytometer shows no indication of changes in morphology compared to the control for any of the reporters (Fig. S16). Using this range of assays, we thus conclude that labeling with probes that can report on the biophysical properties of membranes does not induce detectable toxicity.

Discussion

In this work, we show that FLIM combined with molecular sensors can be used to sensitively and dynamically map the biophysical properties of synthetic and biomembranes with optical resolution, including in conditions where intensity-based measurements fail to do so. This could be achieved with well-characterized rigid and flexible probes that belong to the main classes of sensors, as well as with probes with sensing capabilities that have not previously been reported, although well-characterized environmental sensors exhibit the best sensing capabilities. This has the positive consequence that dyes with improved optical properties (i.e., high QY, low price, high photostability, suitable color) could be used in combination with established sensors in multiparametric sensing. However, the results also call for special attention in applications based on fluorescence-intensity measurements with probes in different environments. In such cases, environmental sensing must be tested for each probe, especially if environmental response is an unwanted effect (i.e., phase-separated membranes).

Considering several technical aspects for fluorescence imaging being pre-optimized (e.g., low enough concentration to prevent self-quenching effects and toxicity but preserving a good signal and high photon count), we point out five main features that should be considered when choosing a specific environmental sensor, especially (but not exclusively) for FLIM analysis. First, the probe’s response to the environment. Determination of the phase-partition coefficient in synthetic or biomembranes based solely on intensity measurements may give erroneous results as the probe’s brightness may depend on its partition (e.g., concentration) but it may also exhibit significant environmental responses. Second, probes with intrinsic long fluorescence decays may more sensitively detect environmental changes due to their higher dynamic range. Third, in FLIM, a single rather than a multiexponential decay is preferred, minimizing ambiguities such as which of the decays best responds to environmental changes. Fourth, membrane partition and intracellular distribution. Chemical derivatives have been designed to aid organelle targeting (14,23) and strong partition in the PM (22,68), whereas, for other applications, nonspecific intracellular distribution may be preferred so that several compartments can be accessed. Fifth, potential side effects (i.e., bleaching and membrane oxidation or changes in cell viability) are important factors that need to be considered, especially for probes that bind membranes weakly and thus require higher concentrations for optimal signal (68).

Other factors are important to consider as well: 1) the physical meaning of the measured parameter—what the probe actually responds to (e.g., polarity, viscosity, etc.) and whether the parameter is quantifiable or only gives a qualitative indication (e.g., GP); 2) speed of acquisition, typically slow with FLIM (although this can be circumvented with the use of sensitive detectors (69)) and spectral imaging and faster with GP; 3) spectral changes, wherein probes with large changes in color occupy a large band of the spectra, reducing the opportunity for multiparametric analysis, and that is also the case for GP analysis (70); 4) whether the probe’s sensing potential is disturbed by external agents (71); and 5) probe localization, if the membrane dye (usually added in the cell medium) equally redistributes in both membrane leaflets or if it is mainly located in the outer leaflet. For the latter, probes that strongly insert in the outer leaflet of the PM but do not undergo considerable flipflop, the readout is biased toward assessing only one of the leaflets rather than the whole membrane. Likewise, upon internalization, the probe mainly gives information about the luminal leaflet of intracellular membranes.

From the library of probes studied, one may wonder what the best sensor for FLIM application really is. Not surprisingly, the answer depends on the model system under study and the question asked as well as the probe’s capability and ease of analysis. For model membranes, the most sensitive probes are NR, FliptR, NBD, and Bodipy C₁₂ given their high dynamic range and lifetime difference discrimination (and thus sensitivity)—they could be used almost interchangeably, in which the particular choice may depend on other factors, such as spectral window for multiplexing with other dyes. From these, FliptR is the only dye that always exhibits a double-exponential decay, although this is well characterized (16,65). Furthermore, NR has the advantage that its polarity sensitivity has also been previously characterized (18), whereas Bodipy C₁₂ is a specific viscosity sensor whose lifetime values can be directly retrieved from published calibrations (72,73). NR and NBD also exhibit a single-exponential decay (Bodipy C₁₂ occasionally too), which facilitates analysis. On the other hand, although DiO can clearly resolve large differences in the environment, it is less suitable for assessing milder environmental differences due to is low dynamic range. In cells, FliptR combines the high dynamic range with strong PM accumulation (or accumulation in other specific organelles with other FliptR derivatives (42)), much higher than the other reporters used here, and it also mildly portions in intracellular organelles, enabling clear lifetime resolution from different intracellular structures. NR is quickly internalized (within a few minutes), precluding assessment of the PM, although NR derivatives with strong PM have recently been reported (74). Bodipy C₁₂ exhibits intermediate PM accumulation with the advantage that its lifetime can be used to assess environmental viscosities, but its poor partition in very ordered domains may limit some applications. Lastly, at the imaging conditions required for fast and (relatively) straightforward environmental sensing, none of the probes studied exhibit significant toxic effects.

We used FLIM as a minimally invasive, contactless, nondestructive, and nontoxic method capable of providing environmental information dynamically and with optical resolution to map biophysical changes induced by membrane-active molecules at the subcellular level. Although the use of FLIM in cellular imaging is not new, it is nonetheless far from its full capabilities when compared to other more established methods. This is, in part, due to the suboptimal performance of probes used in lifetime-based environmental sensing, poor probe characterization, and the lack of useful commercially available probes or with probes that require customized synthesis but that often exhibit suboptimal properties. Instead, here we characterize the sensing response of a series of flexible and rigid fluorescent probes that belong to the three major classes of environmental sensors and check their performance in challenging measurements, including sensitive, subtle, and real-time detection of changes in membrane packing at subcellular level. The analysis is simple, quick, robust, and of much higher throughput compared to electrodeformation (75), which is limited to homogeneous and electrically neutral membranes and requires more complex analysis. In addition, the approach is contactless, in contrast to slow and contact-based methods that are prone to contamination, such as conventional atomic force microscopy (76), and it does not require rupturing the membrane to assess the intracellular milieu (77). Physical parameters (i.e., viscosity, polarity) can be obtained in approximately 1 min, the time between data acquisition to fitting. Although these experiments can be done spectroscopically, we advocate for microscopy analysis to obtain the additional structural dimension, especially if combined with other imaging modalities such as multiphoton excitation (78). The experiments shown here are not exhaustive, but they clearly represent the advantages of microscopy in environmental sensing when using FLIM to assess dynamical maps of the biophysical properties in synthetic and biological membranes at and beyond optical resolution. We hope to aid researchers in their choice of the appropriate environmental sensor for specific and sensitive applications.

Author contributions

R.B.L. devised the study, performed all experiments, analyzed the data, and wrote the first manuscript draft. Study design was performed by R.B.L. in consultation with W.H.R. and K.S. L.S.D. and J.-J.S. provided the cells and supported these experiments. G.Y. and K.S. provided the molecular rotor Bodipy C₁₂. K.S. and W.H.R. supervised the study. R.B.L., K.S., and W.H.R. wrote the final version of the manuscript.

Editor: Frederick Heberle.