2.1.

Materials

The plasmids pECFP $\left(3.4\mathrm{Kb}\right)$ and pEYFP $\left(3.4\mathrm{Kb}\right)$ were purchased from Clontech (Mountain View, California). The sodium phosphate salts, sodium 1,4-bis(2-ethyl-1-hexyl) sulfo-succinate (AOT) and spectroscopic grade iso-octane were purchased from Merck (Darmstadt, Germany). The poly-L-lysine, isopropyl thiol galactoside (IPTG), and all other chemicals were purchased from Sigma (Saint Louis, Missouri). The E. coli strain BL21 (DE3) was obtained from Invitrogen (Carlsbad, California). Round coverslips ( $22\mathrm{mm}$ , #1) were obtained from Menzel-Glaser (Braunschweig, Germany).

2.2.

Production, Purification and Sample Preparation of Fluorescent Proteins

Plasmids with the CFP or YFP coding sequence were transformed into BL21 (DE3) and grown in $500\mathrm{mL}$ LB culture medium supplemented with $100\mu \mathrm{g}∕\mathrm{mL}$ ampicillin at $37°\mathrm{C}$ , $250\mathrm{rpm}$ . At ${\mathrm{OD}}_{600}=0.3$ , IPTG $\left(1\mathrm{mM}\right)$ was added. Cells were harvested $8\mathrm{h}$ after induction by centrifugation, ressuspended in $100\mathrm{mM}$ phosphate buffer (pH 8), and disrupted by sonication. Centrifugation at $\mathrm{12,000}\mathrm{g}$ , $4°\mathrm{C}$ , for $20\min$ separates the supernatant containing the fluorescent protein from cell debris. One protein batch of CFP was purified to electrophoretic homogeneity using two chromatographic steps carried out in an AKTA system. First, hydrophobic interaction chromatography was carried out in a phenyl-sepharose fast flow column (HiTrapTM HIC, $1\mathrm{mL}$ , GE Healthcare), and the protein was eluted at the end of the gradient from 20% to 0% ammonium sulphate (buffer $20\mathrm{mM}$ Tris-HCl, $1\mathrm{mM}$ EDTA, pH 8). Second, anionic exchange chromatography was carried out in a Resource Q column ( $1\mathrm{mL}$ , GE Healthcare), and the protein was eluted at 40 to 50% of the gradient from $0\text{to}0.5\mathrm{M}$ NaCl (buffer $20\mathrm{mM}$ Tris-HCl, pH 7). After two chromatographic steps, CFP was pure by biochemical standards since a single band with $27\text{-}\mathrm{kDa}$ size appears in the SDS-PAGE gel. YFP was not purified. The protein concentration was determined using the extinction coefficients of $26\times {10}^3{\mathrm{M}}^{-1}{\mathrm{cm}}^{-1}$ at $430\mathrm{nm}$ for CFP and $84\times {10}^3{\mathrm{M}}^{-1}{\mathrm{cm}}^{-1}$ at $514\mathrm{nm}$ for YFP.¹⁴

FPs were studied in solution ( $100\mathrm{mM}$ phosphate buffer, pH 7) and spread on coverslips previously coated with poly-L-lysine $(25\mu \mathrm{g}∕\mathrm{mL})$ . A spin coating machine (Laurell Technologies Corporation) was used to spread $100\mu \mathrm{L}$ of a solution containing $1\times {10}^{-5}\mathrm{M}$ of CFP (or CFP plus YFP) on coverslips under controlled conditions of vacuum and centrifugal force ( $1\times {10}^{-9}\text{moles}$ of CFP were actually spread). Reverse micelles at two $W_O$ values ( $W_O$ is the water-in-oil molar ratio, i.e., $\left[{\mathrm{H}}_2\mathrm{O}\right]∕\left[\mathrm{AOT}\right]$ ) were prepared by addition of the appropriate volume of protein aqueous solution to $1\mathrm{mL}$ of $100\mathrm{mM}$ AOT (or larger when specified) in iso-octane and further stirred until total transparency was achieved. CFP concentration in the water pool of reverse micelles was $2\times {10}^{-5}\mathrm{M}$ . FLIM images were obtained after spreading $20\mu \mathrm{L}$ of the reverse micellar solution on coverslips and letting the isooctane evaporate, meaning that $3.6\times {10}^{-12}$ and $14\times {10}^{-12}\text{moles}$ of CFP were actually deposited from reverse micelles with $W_O$ 5 and 20, respectively.

2.3.

Dynamic Light Scattering

Dynamic light scattering (DLS) measurements were carried out using a multiangle apparatus (Brookhaven Corp.) equipped with a He–Ne laser (Model 127, Spectra Physics) with $\lambda =632.8\mathrm{nm}$ and $35\text{-}\mathrm{mW}$ power. The diffused light was detected by a photomultiplier (placed at a fixed angle of $90\mathrm{deg}$ ) and analyzed with a 136-channel correlator (Model BI2030AT). CFP concentration in the water pool of reverse micelles was $5\times {10}^{-5}\mathrm{M}$ . All solutions were filtered through Millipore filters ( $0.2\text{-}\mu \mathrm{m}$ pore size) of hydrophobic polytetrafluoroethylene (PTFE). CONTIN software was used for size distribution information.¹⁵

2.4.

Steady-State Fluorescence Spectroscopy

Steady-state fluorescence measurements were performed at room temperature in a Cary Eclipse fluorescence spectrophotometer from Varian. Emission spectra were recorded using a $405\text{-}\mathrm{nm}$ excitation wavelength and then corrected using a curve supplied with the instrument. Considering the CFP quantum yield in water as 0.4 $\left(\Phi w\right)$ ,¹⁴ the refractive index of water as 1.33 $\left(n_w\right)$ ,¹⁶ and the refractive index of iso-octane as 1.39 $\left(n_m\right)$ , it is possible to calculate the CFP quantum yield in reverse micelles $\left(\Phi m\right)$ from:

Calculation of distances based on FRET measurements requires the knowledge of the distance at which FRET is 50% efficient $\left(R_0\right)$ . This distance for the CFP YFP pair in aqueous solutions is $4.9\text{to}5.2\mathrm{nm}$ .^{6, 17, 18} For reverse micelles, this distance was calculated taking into account small changes in absorption and emission that may occur in reverse micelles using the usual relationship:

Experimental FRET efficiencies were obtained from steady-state and lifetime measurements:

2.5.

Time-Resolved Fluorescence and FLIM Measurements

Time-resolved fluorescence and FLIM measurements were acquired with a Microtime 200 from Picoquant GmbH (Berlin, Germany) using the time-correlated single-photon counting (TC-SPC) technique. The instrument setup includes an inverted microscope (Olympus IX 71) and provides image resolution up to $50\mathrm{nm}$ per pixel. The sample holder is positioned on an XY stage E-710 Digital PZT Controller with scanning range of $100\times 100\mu {\mathrm{m}}^2$ with $1\text{-}\mathrm{nm}$ resolution. Spatial resolution of the setup was down to $220\mathrm{nm}$ as determined by using fluorescent beads from Molecular Probes. This value is close to the diffraction limit (about half the wavelength of the incident light). A more detailed description of the instrument is found elsewhere,^{19, 20, 21}

Lifetime measurements were performed with excitation light of $405\mathrm{nm}$ (pulsed picosecond laser diode set to $40\mathrm{MHz}$ , $0.4\mathrm{mW}$ , $54\text{-}\mathrm{ps}$ pulse width) and bandpass emission filters of $480∕30$ (for CFP) or $545∕35$ (for YFP). Detection could use a photomultiplier tube (Picoquant, Model PMA-182) or single-photon avalanche diodes (SPADs), depending on the fluorescence intensity. Mostly, the photomultiplier tube was used for measurements in solution and SPADs for measurements on coverslips. Data acquisition was performed in a PC equipped with a Timeharp 200 TC-SPC board, (Picoquant) with 4096 channels and a time increment smaller than $40\mathrm{ps}$ . Light from the laser beam was backscattered by the square base of $10\text{-}\mathrm{mm}$ optical path quartz fluorescence cuvette or coverslip and was directed to the detection system to obtain the instrumental response function (IRF).^{22, 23} Data analysis was performed by deconvolution using a nonlinear least-squares fitting program, based on the Marquardt algorithm. The goodness of the fit was evaluated by the usual statistical criteria and by visual inspection of the distribution of weighted residuals and the autocorrelation function.

Average lifetimes, ⟨ ⟩, were calculated from:

3.1.

CFP/YFP in Aqueous Solution

Time-resolved fluorescence decays were obtained for CFP and YFP in aqueous solution, pH 7 (Fig. 1 and Table 1 ). CFP displays two lifetimes and YFP a single lifetime, as described elsewhere.^{18, 24} CFP decays can be fitted with two exponentials even in mammalian cells.²⁵. The introduction of a second exponential to describe the decay of YFP slightly improves the fit (lower ${\chi }^2$ and better autocorrelation), as also documented elsewhere.²⁴ However, as the amplitude of the second lifetime is very low $(<5\%)$ and the fit improvement is not consistent for different measurements, we have considered that YFP decays according to a single lifetime. This is a reasonable approximation to the aim of our study and is well supported by the residuals and the ${\chi }^2$ parameter (Fig. 1 and Table 1). Purification of CFP to electrophoretic homogeneity has no effect on its fluorescence decay, showing that lifetimes of FPs are independent of the presence of other cell fluorophores (Table 1). Those that may eventually be excited at $405\mathrm{nm}$ such as heme proteins or flavin mononucleotides are either not fluorescent or present in such low amounts compared to overexpressed proteins in E. coli that no interference with fluorescence properties of FPs was detected. Therefore, all the results shown in the following report fluorescence of unpurified samples.

Fig. 1

CFP and YFP decays in $100\mathrm{mM}$ phosphate buffer, pH 7, and residuals for a double (CFP) and mono exponential decay (YFP).

Table 1

Fluorescence lifetimes and normalized amplitudes of fluorescent proteins in solution obtained upon deconvolution with the IRF.

In the experimental conditions used, no FRET was detected for the pair CFP/YFP in aqueous solution, as the lifetime of CFP does not decrease in the presence of YFP (Table 1). Calculations made to determine the possibility of energy transfer in solution show that $4\mathrm{mM}$ of the acceptor YFP is needed to detect FRET in solution,²⁶ and the YFP concentration was increased only up to $0.1\mathrm{mM}$ in this study.

3.2.

CFP/YFP Deposited on Cover Slips with Poly-L-Lysine

FPs were spread on coverslips previously coated with poly-L-lysine in order to promote protein adsorption. FLIM images were acquired and show large spots of FPs with no well-defined shape (Fig. 2 ). Distribution of lifetimes in these spots is shown in the histogram of Fig. 2. CFP alone displays an average lifetime distribution with a peak at $2.06\mathrm{ns}$ , shorter than in aqueous solution. In the presence of YFP (ratio 1:5 CFP/YFP), the lifetime distribution can be shifted to shorter or longer time scales, depending on the spot. This indicates that deposition of FPs is heterogeneous as regards both shape and proximity between CFP and YFP molecules, despite the use of a spin-coating machine. Most probably, FRET can occur for some areas whereas it does not occur for others, depending on the local concentration of YFP.

Fig. 2

FLIM images of (a) CFP and (b) CFP/YFP (1:5) spread on coverslips previously coated with poly-L-lysine (PLL). The histogram of lifetimes calculated from FLIM images is shown in (c) for CFP alone (red line, number 1) and for two different depositions of CFP/YFP (1:5) (blue line, number 2). Lifetime distribution for YFP is shown in black (number 3) after deposition of CFP/YFP (1:5). (Color online only.)

A fluorescence decay acquired as a point or an area measurement within the spot and deconvoluted with the instrumental response function reveals in more detail the characteristics of the CFP decay (Table 2 ). It is still a double exponential decay, as in aqueous solution. The average lifetime is shorter compared to the value in aqueous solution due to a decreased amplitude and lifetime of the long component. The fluorescence decay of CFP is sensitive to its immediate environment, and the average lifetime decreases with the increase on the refractive index of the medium.²⁴ When CFP and YFP (ratio 1:5) were jointly spread, the average lifetime of CFP does not change significantly, indicating no FRET. The heterogeneity of the deposition for different spots is also reflected on the large standard deviation observed for the average lifetime of CFP in the presence of YFP $(2.16\pm 0.28\mathrm{ns})$ . YFP in aqueous solution decays with a single component but when spread on PLL displays clearly two lifetimes of the same magnitude as the ones observed for CFP. However, the average lifetime is longer than that of CFP due to larger amplitude of the long component.

Table 2

Lifetimes and normalized amplitudes of fluorescent proteins spread on coverslips previously coated with poly-L-lysine (PLL) upon deconvolution with the IRF.

3.3.

CFP/YFP in Reverse Micelles

Decay of CFP microencapsulated in reverse micelles with a molar ratio of water/surfactant $\left(W_O\right)$ of 5 displays two lifetimes, as in aqueous solution (Fig. 3 ). However, the average lifetime $\left(1.82\mathrm{ns}\right)$ is shorter, as both components are shorter than in aqueous solution. In the presence of more water, i.e., with a water/surfactant molar ratio of 20, CFP decay is still double exponential, but the average lifetime increases to $2.09\mathrm{ns}$ , closer to the value observed in aqueous solution $\left(2.5\text{to}2.6\mathrm{ns}\right)$ . Fluorescence decays of CFP at low water content systems such as reverse micelles or deposited on PLL matrix are therefore shorter than in aqueous solution, mainly due to a decrease in the long component. Decays of YFP microencapsulated in reverse micelles can be fitted to double exponentials, as observed on coverslips coated with PLL (Fig. 3). The average lifetime of YFP is shorter than in aqueous solution, both for PLL-coated coverslips and reverse micelles, showing that low water content leads to decreased FP lifetimes, whether CFP or YFP. The same behavior was previously observed for the green fluorescent protein in reverse micelles.²⁷

Fig. 3

Decays of CFP, YFP, and CFP in the presence of YFP (CFP:YFP molar ratio is 1:5) microencapsulated in reverse micelles of AOT with $W_O$ 5. Lifetimes, amplitudes, and Chi-square values shown on the decay plot and residuals were obtained after deconvolution of the decay with the IRF.

Regarding FRET in reverse micelles with $W_O$ 5, the long component of CFP in the presence of YFP (ratio 1:5 CFP/YFP) decreases, leading to a significant decrease in the average lifetime from $1.82\text{to}1.54\mathrm{ns}$ (Fig. 3). This clearly indicates the occurrence of FRET when the pair CFP/YFP is microencapsulated in reverse micelles with $W_O$ 5. As expected, the lifetime of CFP decreases more significantly with the increase in the ratio acceptor/donor (Fig. 4 ). Efficiency of FRET increases linearly with the ratio acceptor/donor, reaching more than 20% at a ratio YFP/CFP of 10. Using the equation that relates efficiency of energy transfer with donor acceptor distance²⁶ and the value of $4.7\mathrm{nm}$ for $R_0$ measured in reverse micelles [Eq. 2], the distance between CFP and YFP was calculated. For a ratio YFP/CFP of 1 and 10, the average distance between CFP and YFP is 7.0 and $5.6\mathrm{nm}$ , respectively. FRET was confirmed through steady-state measurements (Fig. 4). Interestingly, FRET occurs only in the water pool of small reverse micelles ( $W_O$ 5), as there is no decrease in the fluorescence intensity or in the lifetime when FPs are encapsulated in large reverse micelles ( $W_O$ 20).

Fig. 4

Steady-state (a) and time-resolved (b) FRET measurements for the CFP/YFP pair (different molar ratios), microencapsulated in reverse micelles with $W_O$ 5 and 20. Inset in (b): Efficiency of energy transfer calculated from Eq. 3.

Dynamic light scattering was used to evaluate whether the different size of reverse micelles explains the absence of FRET at $W_O$ 20. Empty reverse micelles (no protein) show a single distribution with hydrodynamic radii larger than previously reported values^{28, 29} due to the use of nondried surfactant and solvent (Table 3 ). Microencapsulation of CFP in reverse micelles with $W_O$ 5 increases the average radius to $8\mathrm{nm}$ , as expected due to reorganization of reverse micelles to accommodate the protein^{28, 29, 30} but a single distribution is still observed (Fig. 5 ). CFP is a cylinder with radius of around $1.6\mathrm{nm}$ and half-height of around $2.5\mathrm{nm}$ . These dimensions justify the size of $8\mathrm{nm}$ for CFP-filled reverse micelles, which should account for the water plus the FP volume. At $W_O$ 20, microencapsulation of CFP results clearly in two size distributions. A bimodal distribution accounting for empty and protein-filled reverse micelles was also detected after microencapsulation of other proteins.³¹ Some reverse micelles remain empty, with hydrodynamic radii close to that measured in the absence of protein, but a new population of very large reverse micelles was detected. CFP molecules microencapsulated in reverse micelles containing more water ( $W_O$ 20) induce the appearance of very large structures, where FP molecules are far enough apart to prevent the occurrence of FRET.

Fig. 5

Illustration depicting microencapsulation of CFP in reverse micelles of two different sizes ( $W_O$ 5 and 20). Hydrodynamic radii $\left(R_H\right)$ were determined by dynamic light scattering (DLS). Microencapsulation of CFP in reverse micelles with $W_O$ 5 increases the size of reverse micelles from $5.5\text{to}8.0\mathrm{nm}$ , with $R_H$ $8\mathrm{nm}$ being the weighted average size of filled and empty reverse micelles, as DLS has no resolution to discriminate between similar sizes. Microencapsulation of CFP in reverse micelles with $W_O$ 20 leads clearly to a two-size distribution of empty and filled structures containing probably several CFP molecules. The shape of this large structure is undefined (shown by a dotted line), as its size is too large in comparison to common reverse micelles, defined as a thermodynamically stable system composed by spherical structures under dynamic equilibrium.

Table 3

Hydrodynamic radius (RH) at the peak of distribution of AOT reverse micelles with no protein and with CFP as measured by dynamic light scattering.

At $W_O$ 5, CFP is sequestered in small reverse micelles. The size of CFP-filled reverse micelles at $W_O$ 5 is not large enough to allow simultaneous encapsulation of both CFP and YFP. The distance between donor and acceptor calculated from FRET efficiency also supports this statement, as the size of filled reverse micelles cannot account for two FP molecules separated from $5.6\text{to}7\mathrm{nm}$ (Fig. 5). Therefore, FRET should result from either micelle encounter, which is diffusion limited, or micelle fusion/fission. Indeed, reverse micelles can collide and fuse their contents temporarily,¹⁰ and transient oligomerization was described.³² Exchange rates of reverse micelles are in the order of ${10}^7\text{to}{10}^8{\mathrm{M}}^{-1}{\mathrm{s}}^{-1}$  ³³. This value times the reverse micellar concentration ( $1.4\times {10}^{-3}\mathrm{M}$ for $W_O$ 5) gives a rate constant significantly smaller than the fluorescence decay rate constant, meaning that during the lifetime of CFP, the reversed micellar system is close to static. FRET results then from micelle proximity and/or from reverse micelles caught in fusion/fission processes. At $W_O$ 20, reverse micelles are very large, and CFP and YFP molecules incorporated should remain quite far apart. The reverse micelle concentration is significantly lower, and diffusion of reverse micelles must be slower, especially regarding protein-filled reverse micelles. Therefore, the micelle proximity and fusion does not allow the FRET process to occur.

3.4.

CFP/YFP Deposited on Coverslips from a Reverse Micellar Solution

FRET measurements used in cell biology studies reveal the spatial distribution of CFP and YFP in the cell, which in most of the cases grow attached to a surface. Assessment of FRET on surfaces is thus of particular interest to the study of cell biology, but as we showed earlier, deposition on a matrix of PLL is heterogeneous and does not allow characterization of the CFP/YFP pair. Since FRET was detected for this pair microencapsulated in small reverse micelles, we have deposited the reverse micellar solution on coverslips and let the organic solvent evaporate. The spherical shape of the water pool of a reverse micelle at $W_O$ 5 allows deposition of FPs as round spots with well-defined contours, a diameter of around $1\text{to}3\mu \mathrm{m}$ , and well separated from each other (Fig. 6 ). By keeping the $W_O$ value constant and increasing the surfactant concentration, which is equivalent to keeping the structure and increasing the number of reverse micelles, it is possible to increase the number of spots on the coverslip. Probably this increase is not linear as single spots should result from reverse micellar coalescence upon solvent evaporation (see the following). At $W_O$ 20 and low concentration of reverse micelles ( $100\mathrm{mM}$ of AOT), spots tend to be more diffuse since they result from very large micelles, as proved by the DLS data shown earlier. Again, by increasing the number of reverse micelles, the number of spots increases. Networks of spots were even obtained when the number of reverse micelles is too large ( $300\mathrm{mM}$ of AOT), indicating spot coalescence. Measurements of lifetimes reveal FRET from CFP to YFP within the spots obtained at $W_O$ 5 (Fig. 7 and Table 4 ). CFP alone display a lifetime distribution peaked at $1.96\mathrm{ns}$ . In the presence of YFP, the peak distribution decreases to $1.74\mathrm{ns}$ , indicating 11% of FRET efficiency and an average distance of $6.6\mathrm{nm}$ between CFP and YFP. The spherical shape of the water pool of reverse micelles in solution remains upon deposition on coverslips, sequestering CFP and YFP in microdroplets that allow FRET to occur. These round spots should result from coalescence of different water pools, as they are larger than the diffraction limit. Fusion of water pools that allow FRET to occur in liquid reverse micelles may be transformed into static entities upon deposition, and the average distance between CFP and YFP remains approximately constant through the deposition process and solvent evaporation. Deposition from large reverse micelles with $W_O$ 20 does not induce FRET, as observed in solution (Table 4). Deposition of FPs on coverslips from reverse micellar solutions reproduces the FRET pattern observed in solution and constitutes a new controlled way to deposit proteins on surfaces.

Fig. 6

Fluorescence intensity images of CFP spread on coverslips from a reverse micellar solution with $W_O$ 5 (upper row) and 20 (lower row). The $W_O$ parameter was kept constant, but the surfactant concentration (AOT) was increased from $100\text{to}300\mathrm{mM}$ to increase the number of spots. Small picture shows a zoom of one spot obtained from a reverse micellar solution at $W_O$ 5 and $100\mathrm{mM}$ of surfactant. All the figures were normalized to have the same contrast to allow comparison of spot intensities.

Fig. 7

FLIM images $(5\times 5\mu \mathrm{m})$ of (a) CFP, (b) CFP:YFP (1:5), and (c) YFP spread on coverslips from a reverse micellar solution with $W_O$ 5. The histogram of lifetimes calculated from FLIM images is shown in (d) for CFP alone (red line, number 1), CFP:YFP (1:5) (blue line, number 2) and YFP (black line, number 3). RGB colors were used to scale lifetimes (red for longer and blue for shorter lifetimes) with predominantly blue and green pixels in A and B, and predominantly yellow pixels in C. (Color online only.)

Table 4

Average fluorescence lifetimes of cyan and yellow fluorescent proteins deposited on coverslips from a reverse micellar solution at two WO values of 5 and 20, obtained from lifetime histograms of FLIM images.