Blue ﬂ uorogenic probes for cell plasma membranes ﬁ ll the gap in multicolour imaging †

a Blue ﬂ uorescent probes for cellular imaging are poorly developed and rarely used because of strong cell auto- ﬂ uorescence at these wavelengths. However, multi-colour imaging needs blue probes, such as ubiquitous nucleus markers DAPI and Hoechst, because they can be readily combined with common green and red markers based on dyes and ﬂ uorescent proteins. Cell plasma membrane is an important target for imaging, but membrane probes that absorb and emit light between 400 and 500 nm are missing. Here, using 3-methoxychromone dyes we designed two blue membrane probes exhibiting >100-fold ﬂ uorescence turn-on ( ﬂ uorogenic response) on membrane binding, large Stokes shift (70 – 90 nm) as well as high brightness and photostability. These unique properties enabled cellular imaging at low probe concentrations (20 – 50 nM) with minimal background from cell auto- ﬂ uorescence and from free probe. RGB multicolour imaging was successfully realized using these probes in combination with common green and red markers. As the new probes enable high-quality imaging of cell plasma membranes in the poorly explored blue spectral region, they may become popular tools that ﬁ ll the gap in multi-colour microscopy.


Introduction
Recent years have seen a tremendous expansion of uorescence techniques and tools for cellular research. In addition to genetically encoded uorescence proteins, a number of molecular probes for monitoring cellular life have been developed. [1][2][3][4] Notably, membrane probes underwent rapid development during the last years. [5][6][7][8] The rst reason is that plasma membrane plays a key role in cell functions and it is the rst barrier that molecules and ions need to cross to enter the cells. Secondly, uorescent probes able to concentrate within the conned space of lipid plasma membrane permit not only to clearly delimit the cell surface, but also to characterize its biophysical properties and monitor cell internalization of molecules. 5,[9][10][11][12][13] Finally, the hypothesis of lipid ras formed by sphingomyelin/cholesterol rich domains 14 stimulated chemists to develop appropriate imaging tools. 6,[15][16][17] Nevertheless, the performance of membrane probes is limited, so that biologists generally use membrane proteins bearing uorescent protein tags or uorescently labelled membrane binding-proteins, such as wheat germ agglutinin (WGA). 18 Among the existing molecular membrane probes, two large families, namely uorescently labelled lipids and specially designed probes, should be mentioned. Labelled lipids are ideal for model membranes, but their use for living cells requires special lipid delivery systems, such as cyclodextrins. 19 The second class is based on specially designed uorescent probes, which can spontaneously stain lipid membranes without using delivery agents. [5][6][7][11][12][13]15 To achieve specic staining of cell plasma membranes with minimal internalization, uorescent dyes are usually modied with a polar head group, as it was done for FM4-64, 20 TMA-DPH, 21 di-4-ANEPPDHQ, 22 C-Laurdan, 23 Mem-SQAC, 24 push-pull glycoconjugates, 12 and oligothiophene amphiphiles. 11 We should also mention F2N12S 25 and NR12S, 13 bearing an amphiphilic anchor group that enables specic staining of the outer membrane leaet with minimal internalization and ip-op between the leaets. 13 However, the membrane probes developed so far, being highly useful for studying biomembranes, 5 overlap with the 500 to 650 nm spectral window of common molecular probes and uorescent proteins, 26 which limit their applications for multi-colour imaging. Exceptions are TMA-DPH and C-Laurdan and their analogues, but these probes require ultraviolet excitation ($360 nm), which is harmful for the cells. Therefore, there is a strong need for powerful blue probes for cell plasma membranes, which, similarly to DAPI and Hoechst, the blue stains of nucleus, could become ubiquitous cell imaging tools. These blue probes should be excitable by violet light ($400 nm), emit efficiently around 450-500 nm and exhibit large Stokes shi in order to minimize the contribution of cell auto-uorescence, commonly observed in the blue spectral region. Moreover, they should be uorogenic, i.e. non-uorescent in water but highly uorescent in lipid membranes, which would enable imaging without background noise from the non-bound probes. 27,28 As prospective building blocks for preparation of blue membrane probes, we considered 3-methoxychromones, because they were recently shown to be bright and photostable, and to exhibit large Stokes shi and uorogenic response to solvent polarity. 29 In the present work using 3-methoxychromone dyes, we developed the rst high-performance blue membrane probes, characterized by high brightness and photostability as well as by suitable absorption and emission wavelengths, allowing their combination with green and red markers. We showed that these probes could be used at concentrations as low as 20-50 nM and that their high brightness, large Stokes shi and efficient membrane binding with uorogenic response provided excellent signal-to-noise ratio in cellular imaging.

Lipid vesicles
Dioleoylphosphatidylcholine (DOPC) and cholesterol were purchased from Sigma-Aldrich. Bovine brain sphingomyelin (SM) was from Avanti Polar Lipids (Alabaster, AL). Large unilamellar vesicles (LUVs) were obtained by the extrusion method as previously described. 32 Briey, a suspension of multilamellar vesicles, prepared from the hydrated lm of lipids, was extruded with the Lipex Biomembranes extruder (Vancouver, Canada). The size of the lters was rst 0.2 mm (7 passages) and thereaer 0.1 mm (10 passages). This generates monodisperse LUVs with a mean diameter of 0.11 mm as measured with a Malvern Zetamaster 300 (Malvern, U.K.). LUVs were labelled by adding aliquots (generally 2 mL) of probe stock solutions in dimethyl sulfoxide to 1 mL solutions of vesicles. Since the probe binding kinetics is very rapid, the uorescence experiments were performed a few minutes aer addition of the aliquot. 20 mM phosphate buffer, pH 7.4, was used in these experiments. Concentrations of the probes and lipids were generally 0.4 and 200 mM, respectively, unless indicated.
In uorescence spectroscopy experiments, HeLa cells were detached by trypsinization. DMEM medium was rst removed from the culture dish, and cells were washed two times with DPBS. Trypsin 1Â (LONZA) solution in DPBS was added to the cells and the cells were incubated at 37 C for 4 min. The solution of trypsinized cells was then diluted by DPBS, transferred to Falcon tubes and centrifuged for 5 min. The washing procedure was repeated one more time with HBSS solution. To stain the cell suspension with the probes, an appropriate aliquot of their stock solution in DMSO was added to 0.5 mL of HBSS buffer, and aer vortexing, the solution was immediately added to 0.5 mL of the cell suspension to obtain a nal probe concentration of 40 nM (<0.25% DMSO) and a cell concentration of 5 Â 10 5 to 10 6 cells per mL. It should be noted that only freshly prepared solutions of the probes in HBSS should be used (<1 min) for cell staining, because of the slow aggregation of the probe in water. Before measurements, the cell suspension with the probe was incubated for 7 min at RT in the dark.
For microscopy studies with the probes, attached HeLa cells were washed two times by gentle rinsing with HBSS. Then, a freshly prepared solution of F2N12S, F2N12SM or FC12SM in Opti-MEM (or HBSS) was added to the cells to a nal probe concentration of 50 nM (<0.25% DMSO volume) and incubated for 7 min in the dark at RT. The obtained samples were imaged directly without washing.
For tri-colour uorescence imaging, HeLa cells were transfected by a plasmid coding mCherry. 3 Â 10 5 HeLa cells were seeded in 35 mm glass coverslips (m-Dish IBIDI, Germany) in DMEM supplemented with 10% FBS and antibiotics (penicillin 100 UI mL À1 , streptomycin 100 UI mL À1 ) and kept at 37 C in a 5% CO 2 atmosphere. Transfection was performed at 24 hours post-seeding with 1 mg of pcDNA3.1 mCherry plasmid using jetPEI™ (PolyPlus transfection, France) according to supplier's recommendations. All observations were done between 16-24 hours post DNA transfection. To stain lysosomes in green, LysoTracker® Green DND-26 (Life technologies) was added to the cells at 50 nM nal concentration and incubated for 30 min at 37 C. Then F2N12SM was added as described above.

Fluorescence spectroscopy and microscopy
Absorption spectra were recorded on a Cary 4 spectrophotometer (Varian) and uorescence spectra on a Fluorolog (Jobin Yvon, Horiba) spectrouorometer. Fluorescence emission spectra were systematically recorded at 410 nm excitation wavelength at room temperature, unless indicated. All spectra were corrected for the uorescence of the corresponding blank (suspension of cells or lipid vesicles without the probe) and wavelength-dependent sensitivity of the detector. Fluorescence quantum yield were measured using 4 0 -(dimethylamino)-3-hydroxyavone in methanol as a reference (27%). 33 Confocal microscopy experiments were performed by using a Leica TCS SPE-II microscope with HXC PL APO 63Â/1.40 OIL CS objective. The excitation was provided by 405 and 488 nm laser and the images were processed with the Image J soware.

Design and synthesis
F2N12SM is an analogue of F2N12S, where the 3-hydroxy group is methylated (Fig. 1). It was prepared by direct methylation of F2N12S with methyl iodide in acetonitrile (Fig. S1 in ESI †). To prepare FC12SM (Fig. 1), 6-(diethylamino)benzofuran-2-carbaldehyde 30 was condensed with 5-chloromethyl-2-hydroxyacetophenone 31 in basic conditions and then oxidized in the presence of hydrogen peroxide into the corresponding 3hydroxychromone derivative (Fig. S2 in ESI †). It was then reacted with N-methyl-dodecylamine and the obtained tertiary amine was quaternized with 1,3-propanesultone. The 3-hydroxy group of the chromone was then methylated to obtain the FC12SM probe. The nal probes were puried by thin layer chromatography and their structure was conrmed by NMR and mass spectrometry.

Fluorescence spectroscopy in model membranes
F2N12SM and FC12SM showed very poor uorescence intensity in aqueous media (QY < 0.5% for both probes), but they were highly emissive in organic solvents (Table 1). On addition of large unilamellar vesicles (LUVs), the uorescence intensity of both probes grew rapidly with lipid concentration and reached stable values at lipid/probe ratios $125 (Fig. 2), indicating that above this probe/lipid ratio nearly all dye molecules were bound to lipids. These data indicate an efficient binding of the probes to lipid membranes. Remarkably, the uorescence intensity increased >200-fold and the emission band shied by 30 and 65 nm to the blue for F2N12SM and FC12SM, respectively (Fig. 2). The position of the emission band also stabilized at lipid/probe ratios $125, conrming that optimal binding was achieved in these conditions. The obtained quantum yields in DOPC vesicles were 33 and 37%, for F2N12SM and FC12SM, respectively, which conrmed their efficient binding to the lipid vesicles. The uorogenic response of F2N12SM and FC12SM to membrane binding can have two mechanisms. On one hand, their 3methoxychromone moieties, owing to the charge transfer character, exhibit efficient quenching in polar media, particularly in water. 29 Therefore, the transfer of the probes from polar aqueous medium to apolar lipid membrane should drastically increase the uorescence efficiency of the probes. On the other hand, our works on Nile Red-based probe NR12S 13 and its analogue, 34 presenting similar structure, as well as other recent reports 35,36 suggest that membrane probes can be self-quenched due to aggregation in water and then disaggregate aer binding to lipid membrane. We expect that both environment-sensitivity and disaggregation of F2N12SM and FC12SM contribute to their remarkable uorogenic character.
In lipid vesicles composed of dioleoylphosphatidylcholine (DOPC), F2N12SM and FC12SM showed an absorption maximum around 400 and 450 nm, respectively (Fig. 3). Thus, both probes were compatible with the 405 nm laser source, commonly used in uorescence microscopy. Moreover, they exhibited a single emission band, centred at 480 and 544 nm (Fig. 3), conrming that methylation of the 3-hydroxy group prevented the formation of the tautomer form observed for F2N12S. 25 According to our earlier studies of the corresponding 3-methoxychromones in organic solvents, 29 these emission maxima indicate a relatively apolar environment similar to the one of ethyl acetate-dichloromethane (dielectric constant 3 ¼ 6-9), suggesting the insertion of the uorophores of both probes into apolar ester region of the lipid bilayer. In comparison with  a l abs and l uo are positions of the absorption and uorescence maxima, respectively (nm); QY is the uorescence quantum yield (%).  the commercially available blue dyes of the coumarin (aminomethylcoumarin acetate, AMCA) or pyrene (Cascade Blue) families, our probes exhibit signicantly larger Stokes shis, which are of key importance for minimizing contribution of the auto-uorescence in microscopy applications. These large Stokes shis originate from the dipolar nature of these chromone derivatives, which ensures signicant solvent relaxation of the uorophores. 37 As these dyes are solvatochromic, we checked their sensitivity to changes in the lipid composition. Surprisingly, the new dyes showed relatively small variation of their emission maximum for different lipid compositions corresponding to liquid crystalline (DOPC), liquid disordered (DOPC/cholesterol) and liquid ordered (sphingomyelin/ cholesterol) phases (Fig. 4). This poor sensitivity to lipid composition is in clear contrast with F2N12S or other solvatochromic membrane probes based on Nile Red 13 and Laurdan. 23,38 Nevertheless, this property is of interest for standard imaging and FRET applications, which require stable position of the emission maximum. Then, we evaluated the photostability of the new probes bound to lipid membranes in comparison to their parent analogue F2N12S. Aer 1 h of illumination (light ux of $1 mW cm À2 ), the uorescence decreased only by 16% and 14% for F2N12SM and FC12SM, respectively, whereas for F2N12S the corresponding uorescence loss was as high as 88% (Fig. 5). This strongly improved photostability is in agreement with our previous report on 3-methoxychromones in organic solvents 29 and is explained by the absence in these dyes of the excited state intramolecular proton transfer generating the less photostable tautomeric form.

Application for cellular imaging
Aer addition to suspensions of HeLa cells, the new probes exhibited a single emission band similar to that observed in model membranes composed of DOPC/cholesterol mixture (Fig. 4). The uorescence intensity and the band shape stabilized within <5 min. The new probes were then added to adherent HeLa cells and studied by confocal uorescence microscopy. A clear membrane staining was observed for both new probes. They exhibited excellent uorescence contrast even at low probe concentrations (50 nM), with a signal-tobackground ratio of 15-20, while for our reference membrane probe F2N12S it was only 3-4 ( Fig. 6B-D). Fluorescence imaging at different concentrations of probes revealed that already at 20 nM of F2N12SM we obtained high-quality images, in contrast to F2N12S that needed concentrations >100 nM (Fig. S4 †), in line with our earlier studies of F2N12S. 25 To the best of our knowledge, these are the rst blue dyes that can be used at such low concentrations for cell membrane staining. Despite these low concentrations, the cell auto-uorescence remained negligible, which is due to the efficient binding of the probes to cell plasma membranes as well as their high brightness (extinction coefficient Â quantum yield $ 30 000 Â 0.4) and large Stokes shi allowing detection far from the excitation wavelength. Importantly, the blue emission region is very convenient, as biologists use extensively green and red emission channels with uorescent proteins and organic dyes. Particularly interesting in this respect is F2N12SM, because its absorption maximum matches perfectly with the 405 nm laser excitation, while its emission does not overlap with the green channel. We evaluated the compatibility of this probe with markers representative of the most commonly used colours. The rst one is LysoTracker® Green DND-26, which labels lysosomes and uses the same instrumental settings as uorescein, AlexaFluor 488, and eGFP. The second one is mCherry, which stains the cytoplasm and corresponds spectrally to common Rhodamine and Cyanine 3 dyes. From the obtained multi-colour images (Fig. 6E-H), it can be seen that F2N12SM clearly stains the plasma membranes in the presence of the other markers (Fig. 6E). Moreover, its uorescence is not detectable in the green (Fig. 6F) and red (Fig. 6G)  channels, indicating the absence of cross-talk between the channels. Thus, F2N12SM is perfectly compatible with both LysoTracker® Green DND-26 and mCherry, allowing RGB imaging. Therefore, this new probe could be used to stain the cellular contour with a complementary blue colour, similarly to Hoechst or DAPI used for staining nucleus.
In conclusion, we describe uorescent membrane probes with convenient absorption in a blue region allowing their combination with common green and red cellular markers for multicolor imaging. To overcome the problem of auto-uorescence common for blue dyes, our probes were developed with several key features. Firstly, these probes are uorogenic, so that they turn-on their uorescence >100-fold aer nearly quantitative binding to the lipid membranes. Secondly, they exhibit high brightness and large Stokes shi, which enable their use at low concentrations (20 nM), applying a simple staining protocol. We expect that these new probes may become as common for multicolour cellular staining as blue nucleus staining dye Hoechst or DAPI.