Time-resolved microspectrofluorometry in live cells, based on time- and space-correlated single-photon counting, is a novel method to acquire spectrally resolved fluorescence decays, simultaneously in 256 wavelength channels. The system is calibrated with a full width at half maximum (FWHM) of 90 ps for the temporal resolution, a signal-to-noise ratio of 106, and a spectral resolution of 30 (Δλ/Λ). As an exemple, complex fluorescence dynamics of ethidium and cyan fluorescent protein (CFP) in live cells are presented. Free and DNA intercalated forms of ethidium are simultaneously distinguishable by their relative lifetime (1.7 ns and 21.6 ns) and intensity spectra (shift of 7 nm). By analysing the complicated spectrally resolved fluorescence decay of CFP, we propose a fluorescence kinetics model for its excitation/desexcitation process. Such detailed studies under the microscope and in live cells are very promising for fluorescence signal quantification.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Owing to its very low invasive properties and the emergence of the green fluorescent protein (GFP) tool (Tsien, 1998), fluorescence microscopy is a widespread method in biology. Several methodologies measuring dynamics and interactions of biomolecules are based on fluorescence, such as fluorescence recovery after photobleaching (White & Stelzer, 1999; Reits & Neefjes, 2001), fluorescence resonance energy transfer (Bastiaens & Squire, 1999; Selvin, 2000; Tramier et al., 2002) and fluorescence correlation spectroscopy (Berland et al., 1995; Gennerich & Schild, 2000; Goedhart et al., 2000). To use these methodologies effectively, quantitative analysis of the fluorescence has to be performed. Because of cellular autofluorescence, photobleaching, photoconversion processes or unknown probe concentration, the experimental conditions applied to cellular studies do not always allow us easily and automatically to quantify the fluorescence signal. To achieve this, an experimental approach to the fluorescence phenomenon has to be performed with as much detail as possible, looking at both the temporal and the spectral properties.
Fluorescence microscopy images are based on a spectral shift between excitation and emission that permits the discrimination of fluorescence signal. In this case, some information contained in the fluorescence spectrum properties is lost. By recording the complete emission spectra, characterization and quantification of fluorescence signal is made easier. Microspectrofluorometry is now an easy-to-use technique with currently available commercially systems (Dickinson et al., 2001).
Of particular interest is the possibility to monitor simultaneously fluorescence spectral and lifetime properties, which allows precise characterization and quantification of cellular fluorescence. Acquisition of fluorescence dynamics on the picosecond time-scale of very weakly emitting sources from subcellular areas in living cells is carried out by using the time-correlated single-photon counting (TCSPC) method (O’Connor & Phillips, 1984). Based on time- and space-correlated single photon counting (TSCSPC), a delay line detector (DL) was previously used in microscopy (Kemnitz et al., 1997; Tramier et al., 2002). In combination with a polychromator in front of the DL detector, the space information of the DL detector corresponded to wavelength information (Kemnitz, 2001). In this paper, we present the design and calibration of this time-resolved microspectrofluorometry system. Calibration of the system was carried out by taking advantage of the fluorescence properties of ethidium. The accuracy of spectrally resolved fluorescence decays was demonstrated by measurement of the fluorescence lifetime and spectra of free and intercalated ethidium in DNA in living cells. This technique was further applied to the analysis of fluorescence dynamics of cyan fluorescent protein (CFP).
By taking advantage of simultaneous acquisition of time and space information by a TSCSPC detector (DL, EuroPhoton GmbH, Berlin, Germany) (Fig. 1) (Kemnitz et al., 1997), fluorescence decays were established by counting and sampling single emitted photons according to (i) the time delay between their arrival and the laser pulse, (ii) their coordinates along the detector axis (one-dimensional space resolution), which corresponded to a slit entrance of 0.2 mm width and 20 mm height placed just before the photocathode. In this detector, the incident photon produces a cloud of electrons at the output face of the second microchannel plate that hits a conductive line at the same spatial position of the original photon. The electric pulse is split into two pulses, amplified, discriminated, and used as the start and stop pulses in a time-to-amplitude converter (TAC). The travel time difference indicates the x-position. Simultaneously, a second TAC was used for time correlation between one of the DL pulses and signal from a fast photodiode triggered by the pulsed excitation source. The time difference calculated by the two TACs (TAC1 and TAC2, see Fig. 1) corresponded to:
where x is the position of the incident photon along the line, L the length of the conductive line, V the conduction rate and Δt the correlation time between the incident photon and the signal from the fast photodiode triggered by the pulsed excitation source. By combining information from the two TACs, time and space parameters could be independently characterized. Thus,
TAC1 = ax + b and TAC2 − 1/2TAC1 = Δt + b′
where a, b and b′ are constant. The two-dimensional (2D) multichannel analyser, equipped with transputer and local memory, combines the two TAC outputs, resulting in the acquisition of a single counted photon with digitized time and space coordinates.
By using a transmission imaging polychromator, based on a holographic grating, sandwiched between two prisms, fluorescence was linearly dispersed according to the emission spectra with a transmission efficiency of better than 50%. The wavelength region is 380–780 nm and the resolution is 5 nm with a 80-µm slit width. In our application, we did not used the spatial resolution of the imaging polychromator (slit height). The selected sample fluorescence was imaged by the microscope at the slit entrance of the polychromator. The DL detector was directly coupled to the output port of the polychromator. The direction of the slit entrance of the DL was crossed with the polychromator slit entrance and corresponded to the direction of wavelength dispersion. Therefore, the wavelength distribution is converted into space information and is sent to the DL for temporal resolution (Fig. 2). The resulting 2D histogram of photons is displayed as a pseudocolour map with 256 and 2048 channels for the wavelength and time dimensions, respectively. The set-up is shown schematically in Fig. 3.
A picosecond mode-locked titanium sapphire laser (Millennia 5W/Tsunami 3960-M3BB-UPG kit, Spectra-Physics, France) was tuned between 760 and 980 nm to obtain excitation wavelengths of 380–490 nm after frequency doubling. The repetition rate was 4 MHz after a pulse-picker was inserted (Spectra-Physics 3980-35, France). The laser beam was expanded and inserted into an inverted epifluorescence microscope (Leica DMIRBE, France) to obtain wide-field laser illumination (Tramier et al., 2002). The subcellular region of interest was selected by closing the field diaphragm of the microscope. By using a 100× objective (NA = 1.3), the 2-µm diameter of the selected subcellular region, which was 200 µm wide, was imaged directly at the slit entrance of the polychromator (80 µm) settled on the outlet port of the microscope. The fluorescence was collected by the DL detector placed directly after the polychromator such that the two slits are crossed (Fig. 2). The two DL electronic signals were amplified and resolved. The signal from the fast photodiode triggered by the pulsed excitation source was also determined. These three signals coming from constant fraction discriminators were used as start and stop pulses in the two TACs. Finally, the TAC outputs were directed towards the 2D multichannel analyser driven by a computer. Acquisition of time-resolved emission spectra on the picosecond timescale, simultaneously in 256 wavelength channels, was performed after accumulation of sufficient single-photon events.
For living cells studies, Vero cells were cultured on glass coverslips in Dulbecco's modified eagle medium (Life Technologies, Cergy Pontoise, France), supplemented with fetal calf serum (10%), at 37 °C in a 5% CO2 atmosphere. Cells were either incubated with 10 µg mL−1 ethidium bromide (Molecular Probes, Leiden, The Netherlands) for 30 min at 37 °C or transfected with pECFP-C1 from Clontech using FuGENE™ 6 Transfection Reagent (Roche Molecular Biochemicals, Meylan, France), according to the manufacturer's instructions. The coverslips were mounted in a special holder and placed on the stage of the inverted microscope.
Fluorescence data were analysed to recover the different discrete lifetimes that could be detected in the decay. To obtain lifetimes from fluorescence decays, experimental measurements were fitted by the convolution product of a multi-exponential theoretical model with the instrument response function (IRF):
where ai is the relative contribution of fluorescent species i, characterized by fluorescence lifetime τi and IRF is the experimental measurement of the pulsed excitation (by acquiring the refraction of the laser beam). Data were analysed by a Marquardt non-linear least-squares algorithm using the Globals software package developed at the Laboratory for Fluorescence Dynamics at the University of Illinois at Urbana-Champaign. Fitting procedures were carried out following two different methods: (i) analysis of a single decay corresponding to the whole emission spectrum, which was obtained by gathering data over all the wavelength channels, and (ii) global analysis with linked lifetimes over the different decays corresponding to different emission wavelengths, which were obtained by gathering data over five contiguous wavelength channels via addition of blocks of five sets of counts. In the latter case, decay-associated spectra (DAS) were obtained, i.e. the relative contribution of both lifetimes at different wavelengths allows the emission spectrum of both species to be characterized.
Calibration of time-resolved microspectrofluorometer
The IRF is shown in Fig. 4 (FWHM of 90 ps). The shape of the curve is a result of the fluctuation in the travelling time of the electron avalanche in the microchannel plate. Note the presence of some prepulses, which are attributed to shorter optical paths of the photons and which are characteristic of the TCSPC method carried out under microscope. Indeed, the calculated lengths of these short paths corresponded to the direct refraction of part of the laser light by the dichroic mirror and the bottom of the objective. Nevertheless, this phenomenon is negligible (less than 1/1000 photons).
Signals from the DL were thresholded and electronic noise coming from the detector was partially removed. This noise is not time-correlated, in contrast to the fluorescence. This was verified by acquisition of flash lamp photons by the detector alone (Fig. 5). The fit of this background, with a reduced χ2 (< 1), revealed no time correlation. These properties of the TCSPC method explain the high dynamic range of the detection system, with a signal-to-noise ratio of 106. This represents an invaluable advantage for biological studies where the fluorescence signal is weak and where a low level of excitation is necessary to not perturb the living state (Delic et al., 1991).
In the present work, the count rate of the complete system was between 500 and 2000 cps. This count rate limitation is not related to the DL detector itself, for which the count rate can be up to 4 × 104 cps (see Discussion) but comes from two other sources: (i) microscope/laser coupling for wide-field illumination, which gives a maximum excitation density as low as a few mW cm−2 at the sample (for maximum laser power of 0.25 mW at 490 nm and 2.50 mW at 440 nm); (ii) collection efficiency of ∼10% for the polychromator/DL detector coupling system. Nevertheless, because the excitation level is very low, it is possible to acquire over a period of several minutes fluorescence coming from the cell without any photobleaching or perturbation of the living state.
The wavelength resolution of the polychromator/DL system was determined by measuring the spectral response of the laser at 480 nm. The measured FWHM of 16 nm gives a spectroscopic resolution, Δλ/λ, of 30. Each of the 256 wavelength channels corresponds to 1.35 nm and the FWHM of 16 nm corresponds to 12 channels. It is thus possible to gather five consecutive channels without reduction of the wavelength resolution as is necessary for global analysis in measuring DAS. To take into account the variation of the detection sensitivity according to the wavelength, all the steady-state spectra and DAS were corrected with the spectral response of white light (halogen lamp).
Fluorescence dynamics of ethidium bromide
To test the ability of our system to determine simultaneously fluorescence lifetime and spectra of different species in living cells, fluorescence properties of ethidium in the nucleus of proliferative mammalian cells were analysed. We have previously shown that the intercalation of ethidium is strongly restrained in nuclear DNA of living proliferative mammalian cells (Tramier et al., 2000), giving two different populations of ethidium: free and intercalated in DNA in the nucleus. Free and intercalated ethidium could be resolved as a result of their fluorescence lifetime values, 1.7 ns and 22 ns, respectively. Indeed, in an aqueous environment, ethidium fluorescence is quenched owing to proton transfer from ethidium to water molecules (Olmsted & Kearns, 1977). The fluorescence enhancement and the increase of fluorescence lifetime upon ethidium intercalation into nucleic acids is due to a restricted access of water and/or other proton acceptors to the intercalating sites of the double helix. The very low steady-state fluorescence intensity of ethidium in the nucleus of the cell displayed in Fig. 6 is a result of the very low level of intercalated ethidium, confirming the integrity of the living cell state as demonstrated previously (Delic et al., 1992; Coppey-Moisan et al., 1996; Durieux et al., 1999; Tramier et al., 2000).
By contrast, a high steady-state ethidium fluorescence is seen in the nucleolus and the mitochondria because intercalation of ethidium in nucleic acids (DNA/RNA) is not restrained in these organelles. Figure 6(B) shows the spectrally resolved ethidium fluorescence decay at 490 nm excitation wavelength from the nuclear area of the Vero cell presented in Fig. 6(A) (red circle). Global analysis of the data revealed two lifetimes, 21.6 ns and 1.7 ns. These values are in good agreement with the lifetimes of the intercalated and free ethidium determined previously (Olmsted & Kearns, 1977; Tramier et al., 2000). The DAS of these two lifetimes (21.6 ns and 1.7 ns) are shown in Fig. 6(C). These spectra correspond well with those obtained with ethidium-free solution and intercalated in λ-phage linear DNA. Intercalated ethidium into DNA gives a blue-shifted spectrum in comparison with that of ethidium-free DNA (maximum blue shift from 600 nm to 593 nm).
Fluorescence dynamics of cyan fluorescent protein
Fluorescence spectra of the CFP have an absorption maximum at 434 nm and an emission maximum at 476 nm with a red shoulder for both spectra (Tsien, 1998). In addition, CFP fluorescence decay can be modelled with two different lifetimes as previously shown (Tramier et al., 2002). These results suggest that the two close peaks of the CFP emission spectrum correspond to two species for which the desexcitation kinetics are different. In order to verify this assumption, we performed time-resolved microspectrofluorometry on living cells expressing CFP; a representative result of several experiments is shown in Fig. 7. Two days after transfection, a subcellular area of a Vero cell expressing CFP was chosen (Fig. 7A) to record spectrally resolved fluorescence decay at 439 nm excitation wavelength (Fig. 7B). The DAS of the two fitted lifetimes (3.13 ns and 0.75 ns) represent the result of the global analysis (Fig. 7C). Surprisingly, only a slight difference between these two spectra was observed.
Because the fluorescence excitation spectrum of CFP contains two close peaks, measurements of spectrally resolved fluorescence decays were taken at different excitation wavelengths. Ti : Sa laser was tuned to obtain an excitation wavelength of 424–449 nm after the pulse-picker and a measurement was performed every 5 nm. The corresponding fluorescence decays and spectra are presented Fig. 8. As shown by the logarithmic representation of the different fluorescence decays, CFP fluorescence kinetics depend on the excitation wavelength, the kinetics being faster in the blue and slower in the red part of the spectrum. Results of individual analyses of these decays are presented in Table 1. The two lifetime values obtained for each excitation wavelength are similar and confirm the existence of the two fluorescent species. However, the proportion of the two lifetime species varies according to the excitation wavelength. This result suggests that these two species exhibited different absorption properties. The two close peaks that comprise the excitation spectrum of the CFP correspond to the two species, i.e. the blue side and red side for short (0.75 ns) and long (3.13 ns) lifetimes, respectively.
Table 1. CFP fluorescence kinetics parameters as a function of excitation wavelength.
Excitation wavelength (nm)
ai, the pre-exponential factor, is the relative contribution of the τi lifetime species.
In contrast to the proportion of the two lifetimes, CFP fluorescence emission spectra are not affected by the excitation wavelength (Fig. 8C). This result confirms that emission spectra of both species are identical, in agreement with the result presented in Fig. 7.
The picosecond TCSPC method in microscopy is advantageous for the quantification of fluorescence lifetimes in living cells even in the case of several and close lifetime values. The very high sensitivity of this technique allows (i) extremely low excitation intensity to be used, thereby avoiding photobleaching, and (ii) very low levels of fluorescent molecules to be detected. In addition, combination of polychromator with DL detector allows us to measure, in living cells, spectrally resolved fluorescence decay in the picosecond timescale.
In TCSPC, a limitation of the count rate comes from the requirement for single photon counting statistics. Indeed, the acquisition system is not able to process more than one emitted photon for each pulse excitation. Where two emitted photons after one pulse excitation are detected, only the first is processed, inducing artefactual faster fluorescence kinetics and shorter lifetime. In optimal use, with a 4-MHz repetition rate of the laser, the count rate of the DL detector is up to 4 × 104 cps (ratio of detected photons to pulse excitation less than 1%) without inducing micro channel plate saturation or TAC saturation. The count rate in the present study (500–2000 cps) is below the DL detector limitation and comes from low laser excitation (wide-field illumination of a few mW cm−2) and the collection efficiency of the polychromator/DL detector sytem (∼10%), which mainly arises from coupling of the microscope/polychromator/DL detector. This can be further improved by: (i) optimization of the geometry of the observed volume; (ii) an increase in width of the polychromator slit up to 150 µm, but with a concommitant decrease in the wavelength resolution; (iii) dispersion of the full observed wavelength range (200 nm) over the total portion of the photocathode of the DL detector instead of just a small portion, resulting in dispersion of the entire 380–780 nm range to the photocathode. Nevertheless, under the conditions used here, we were able to preserve the sample in an appropriate environment, even over a very long acquisition period. Although the count rate limitation means that at present we are unable to investigate fast biological process in the cell, it was possible to measure simultaneously the lifetime and the intensity spectra of different fluorescent species.
Other technical aspects in recording spectral and lifetime information would be (i) to use simultaneously multiple individually spectral filtered detectors or (ii) to use a single detector with sequential filter activation. In the first of these, taking advantage of multiple detectors, the count rate will be higher. Nevertheless, the ratio of counted photons to emitted photons will be worse than the ratio obtained with the polychromator/DL system, because some emitted photons would be lost during spectral filtering. In the second solution, for the polychromator/DL system, the simultaneous spectral acquisition led to a constant maximum count rate. By contrast, for the single detector with sequential filter activation, the maximum count rate must be reached only at the point of maximum emission. During activation of other wavelength filters, by using the same excitation intensity, the count rate will be lower. The mean count rate would then be lower than the maximum count rate.
To demonstrate the accuracy of the measurements, ethidium fluorescence dynamics were analysed in the nucleus of living mammalian cells. The ability to distinguish two different states of ethidium (free and intercalated in nuclear DNA) was demonstrated from simultaneous determination of the known fluorescence lifetimes and emission spectra corresponding to both species. A similar study of ethidium fluorescence was carried out earlier via microspectrofluorometry (Favard et al., 1997). In microspectrofluorometry, the measured spectrum is analysed by decomposition with known species and their respective spectra. In our current approach, the time-resolved microspectrofluorometry allows the fluorescence lifetime and spectra to be discriminated simultaneously. Moreover, for the two ethidium fluorescence states in the nucleus of living mammalian cells, because the two spectra are very similar (wavelength shift from 600 nm to 593 nm between free and intercalated ethidium), the lifetime information becomes particularly useful, especially when very few ethidium molecules are intercalated. By contrast, lifetime determination without spectral information should not be sufficient to discriminate between fluorescent species of interest and autofluorescence, especially for signals originating from regions of very low intensity, as is the case for ethidium fluorescence in the nucleus. In this example, the particular shape of the DAS of the two lifetime species allows us to demonstrate the presence of free and bound ethidium in the nucleus without any autofluorescence artefacts.
Until now, CFP fluorescence lifetime and spectral properties have been investigated separately. The fluorescence decay corresponds to two distinct lifetime species (Tramier et al., 2002) and excitation and emission spectra both present a red shoulder (Tsien, 1998). Picosecond time-resolved microspectrofluorometry of CFP in living cells gives precise details about the connection between lifetime and spectral properties. Surprisingly, the two fluorescence lifetime species of CFP show different excitation and similar emission properties. This experimental approach allows us to propose a kinetic model shown schematically in Fig. 9. Two different species A and B with different excitation spectra (λ1max and λ2max, respectively) could exist for CFP and their excitation give rise to the two bands of the CFP excitation spectrum (λ1max < λ2max). Then the two corresponding excited species, A* and B*, return to equilibrium via an intermediate excited species I* with distinct rates k1 and k2, respectively (k1 > k2). Finally, I* would relax to a final state (I) by a radiative process characterized by a unique emission spectrum and a unique rate k3. In this proposed model, the two possible excitation/desexcitation paths (AA*I*, BB*I*) would have different excitation spectra, different lifetimes [respectively 1/(k1 + k3) and 1/(k2 + k3) with 1/(k1 + k3) < 1/(k2 + k3)] and identical emission spectra.
Fluorescence resonance energy transfer (FRET) measurement provides a useful tool to detect interactions between fluorescently tagged proteins and is now used to follow the dynamics of interactions between specific proteins in relation to biological processes directly in the living cell (Verveer et al., 2000; Li et al., 2001; van Der Wal et al., 2001). We recently improved picosecond FRET microscopy by acquiring the fluorescence decay of a donor fluorophore in living cells (Tramier et al., 2002). However, the quantification of FRET is difficult to carry out under the microscope when the stoichiometry of the donor–acceptor pair cannot be ensured. Interestingly, the approach presented here would have several advantages in the FRET quantification. First, the precise spectroscopic characterization of both donor and acceptor fluorophores in live cells, as carried out in this paper for CFP, presents several advantages to the use of these fluorophores in FRET experiments. Second, this technique allows us to measure simultaneously donor and acceptor fluorescence decays. A similar study using a phase and modulation method was recently presented (Hanley et al., 2002). Time-resolved microspectrofluorometry in living cells is a promising tool to discriminate between short lifetime coming from autofluorescence and coming from non-radiative energy transfer processes.
We are indebted to Dr Valentina Emiliani for critical reading of the manuscript and to Michel Thomas for technical help. This work was supported by European Union Grant BIO4 CT97 2177; the Association pour la Recherche sur le Cancer Grants 9222 and 5632; the Groupement des Entreprises Françaises de Lutte contre le Cancer; and the region Iles-de-France (SESAME). M.T. was supported by the Ligue Nationale Contre le Cancer.