Two-photon lifetime-based photoconversion of EGFP for 3D-photostimulation in FLIM

Enhanced green fluorescence protein (EGFP) is a fluorescent tag commonly used in cellular and biomedical applications. Surprisingly, some interesting photochemical properties of EGFP have remained unexplored. Here we report on two-photon-induced photoconversion of EGFP, which can be permanently converted by intense IR irradiation to a form with a short fluorescence lifetime and spectrally conserved emission. Photoconverted EGFP thus can be distinguished from the unconverted tag by the time-resolved detection. Nonlinear dependence of the two-photon photoconversion efficiency on the light intensity allows for an accurate 3D localization of the photoconverted volume within cellular structures, which is especially useful for kinetic FLIM applications. For illustration, we used the two photon photoconversion of EGFP for measurements of redistribution kinetics of nucleophosmin and histone H2B in nuclei of live cells. Measurements revealed high mobility of fluorescently tagged histone H2B in the nucleoplasm and their redistribution between spatially separated nucleoli.


Abstract
Enhanced green fluorescence protein (EGFP) is a fluorescent tag commonly used in cellular and biomedical applications. Surprisingly, some interesting photochemical properties of EGFP have remained unexplored. Here we report on two-photon-induced photoconversion of EGFP, which can be permanently converted by intense IR irradiation to a form with a short fluorescence lifetime and spectrally conserved emission. Photoconverted EGFP thus can be distinguished from the unconverted tag by the time-resolved detection. Nonlinear dependence of the two-photon photoconversion efficiency on the light intensity allows for an accurate 3D localization of the photoconverted volume within cellular structures, which is especially useful for kinetic FLIM applications. For illustration, we used the two photon photoconversion of EGFP for measurements of redistribution kinetics of nucleophosmin and histone H2B in nuclei of live cells. Measurements revealed high mobility of fluorescently tagged histone H2B in the nucleoplasm and their redistribution between spatially separated nucleoli.

Introduction
Fluorescent proteins (FPs) are widely used for imaging of dynamic cellular processes, monitoring of gene expression, protein tracking, and localization [1,2]. FPs frequently exhibit complex photochemical behavior. Emission properties of a number of them can be light-shifted between various emitting states [3][4][5].
Photoactivatable fluorescent proteins can be photoconverted from dark to bright emitting state, photoswitchable FPs permanently shift from one fluorescent state to another, and reversibly switchable FPs can be repeatedly light-toggled between spectrally distinct fluorescent states by proper illumination [5]. New photo-manipulable variants of FPs have often opened new application fields ranging from cellular protein tracking and rapid cellular dynamics to multicolor and superresolution imaging [5][6][7].
Fast proliferation of FP mutants and their applications often cause a lag between their usage and thorough physical and photochemical characterization [8].
For this reason, users of such molecules should rather cautiously interpret their data. Sometimes even notoriously known FPs like EGFP can reveal overlooked properties that, unrecognized, can seriously bias experimental data. It could be e.g. unwanted spectral photoconversion during photobleaching known as redding [9,10], greening [11], or lifetime-based photoconversion [12]. However, with proper knowledge, these drawbacks can be turned to advantages in new applications [12].
EGFP is an extensively used fluorescent tag. It exhibits good quantum yield, large extinction coefficient and absorption and emission compatible with common excitation sources and microscopes, efficient folding, low oligomerization, and high photostability [13,14]. EGFP can be also efficiently excited by the two-photon excitation (TPE) with the cross section σ 2 close to 40 GM [15]. All these features are desirable for quantitative microscopy. Over the years, EGFP has been well characterized [e.g. 13, [16][17][18][19][20][21][22] and validated for use in the cell research [12,23]. The tag is widely spread, easily available, and remains the green FP of choice for many researchers [24]. Recently we have revealed that EGFP can be permanently photoconverted to a form with a shorter fluorescence lifetime. While under gentle illumination EGFP has been found fairly photostable and resistant to photodestruction [1,25,26], strong blue illumination to the first singlet absorption band causes a lasting reduction of its excited state lifetime without noticeable spectral shifts [12]. Though there are number of FPs photoconvertible between spectrally distinct forms [2,27,28], the emission rates can be light-manipulated only for a few of them. In addition to EGFP, this property has also been reported for ECFP and Cerulean [29]. This group of lifetime-convertible FPs is extremely desirable and useful for fluorescence lifetime imaging (FLIM) [30].
Compared to the conventional single-photonexcited laser scanning FLIM, its two-photon variant exhibits a number of advantages in biomedical and cellular research [31][32][33]. A quadratic dependence of the two-photon absorption rate on the light intensity causes excitation only at the focal point where the photon flux is large enough to induce fluorescence [34]. This implies an inherent sectioning capability permitting elimination of the photon-wasting confocal pinhole in the detection light path and improved photon efficiency [33]. Since the absorption occurs only at the small femtoliter-sized volume, the overall photodamage of the sample is reduced, which is highly desirable in experiments requiring prolonged sample illumination [15]. Infrared excitation better penetrates the scattering samples and allows for deep tissue imaging [35,36]. Due to the antistokes emission, scattering and autofluorescence of the sample can be efficiently suppressed.
An intriguing task of cellular biophysics is an evaluation of the diffusion and redistribution of macromolecules in live cells [37], especially between tiny cellular compartments [38,39]. A conical singlephoton excitation beam results in a photodamage, photoactivation or photoconversion along the whole excitation light path and subsequent loss of the ability to accurately 3D-position the photoconverted volume. Importantly, due to the nonlinear intensity dependence, the two-photon excited or photoconverted spot can be well localized within the 3D sample and allows for more accurate targeting of the subcellular structures. Two-photon-induced photoconversion (TPPc) is therefore well suited for the investigation of 3D-diffusion and transport processes on the subcellular scale [32]. Unlike FRAP, TPPc does not require complete photodestruction of fluorophores in the activated area. As a consequence, mixing of signals from the unconverted and photoconverted protein can be followed in time.
Here we present a two-photon photoconversion of EGFP resulting in a fluorescent EGFP form with reduced fluorescence lifetime and spectrally conserved emission. The photoconverted protein is easily separable from the unconverted form by lifetime imaging and serves as a suitable light-inducible lifetime highlighter for FLIM. We show that unlike the single-photon conversion, the strong dependence of the twophoton photoconversion on the light intensity allows for accurate 3D localization of the photoconverted volume within the cell. This is especially useful for kinetic FLIM applications, e.g. for exploration of cellular diffusion, protein redistribution, etc [40]. Importantly, only a single spectral channel is needed for the detection while the other ones stay free for multispectral imaging. Moreover, the inherently ratiometric nature of emission lifetimes allows for elimination of artifacts commonly accompanying any intensity-based measurements. As an application example of the TPPc-FLIM technique, we present visualization of the redistribution of EGFP-tagged nucleophosmin (EGFP_NPM) between nucleoli of live HEK-293T cells. The second application example shows TPPc-FLIM measurement of redistribution kinetics of EGFP-labeled histone H2B (EGFP_H2B) in the nucleus of live HeLa cells, which documents diffusive communication between separated nucleoli and rapid exchange of H2B between them.
Specifically, the plasmids were amplified in E. coli and purified with PureYield Plasmid Miniprep System (Promega). The cells were seeded to 1.5 × 10 5 ml −1 cell density, incubated for 24 h and transfected using jetPrime transfection reagent (Polyplus Transfection) according to the manufacturer's protocol. The transfected cells were further grown for 20-40 h prior to experiment.

Cell fixation
EGFP-producing cells grown on a glass-bottom Petri dish were fixed with 4% paraformaldehyde and permeabilized by 0.5% Triton X-100 as described earlier [12]. The permeabilization was used merely for compatibility with the original protocol that ensures faster equilibration of the fixed cells. Finally, the cells were submerged in the sterile PBS and stored in 4°C. Experiments with fixed cells were done at room temperature.

Fluorescence lifetime imaging
FLIM measurements were performed on the apparatus described in detail elsewhere [12]. Briefly, we used an inverted IX83 microscope equipped with a FV1200 confocal scanner (Olympus, Hamburg, Germany), a cultivation chamber, and FLIM add-ons from Pico-Quant (Berlin, Germany). Single photon emission was excited by the LDH-DC-485 pulsed laser emitting at 482 nm with 20 MHz repetition rate (PicoQuant, Berlin, Germany). Laser output coupled to the microscope by a single-mode optical fiber was reflected to the sample by the 488 nm long-pass dichroic mirror (Olympus). Emission was collected through the 520/ 34 bandpass filter (Semrock, New York, USA).
Two-photon imaging was done on the same microscope using a free-space-coupled IR beam from the Ti:Saphire laser (Chameleon Ultra2, Coherent, Santa Clara, California) running at 80 MHz. The group-delay dispersion was pre-compensated by a pulse compressor (Thorlabs, Dachau, Germany) and the IR light was reflected to the sample by the RDM 690 dichroic mirror (Olympus). Depending on the experiment, fluorescence was directed to the cooled GaAsP hybrid PMTs (PicoQuant) either through a liquid light guide mounted on the non-descanned port or via a multimode optical fiber mounted on the descanned port. EGFP fluorescence was spectrally isolated using the 520/34 bandpass filter (Semrock) and the residual IR excitation was blocked by the 680/SP filter (Semrock). Time-resolved data were acquired by the TimeHarp 260-PICO TCSPC card and processed by the SymPhoTime64 software (PicoQuant). To minimize pile-up artifacts [43], the data collection rate for the brightest pixels was kept below 5% of the excitation frequency. Typically, we used UPLSAPO60XW water immersion objective (NA 1.2) or UPLSA-PO30XS silicone oil immersion objective (NA 1.05) (Olympus). Experiments with fixed and live cells were done at 25°C and 37°C, respectively.
For the rapid FLIM imaging, the end of photoconversion was synchronized with the start of the FLIM acquisition. A series of several hundreds of FLIM images was immediately collected with the FLIM frame rate of 2 frames/s. The obtained multiframe FLIM stack allowed for construction of the lifetime evolution in any selected ROI of the image.

Two-photon photoconversion
Photoconversion was performed by scanning a selected ROI with live monitoring of the emission intensity. Depending on the excitation wavelength, type of the experiment and a particular sample, the mean power of the IR laser was set between 10-40 mW at the focal point of the objective. Typically, we finished the photoconversion when the emission intensity of EGFP decreased to about 20%-30% of its initial value. For time-undemanding experiments the TPPc was done gradually with standard scanning and lower light power within 2-4 min. For the rapid FLIM experiments the photoconverted area and the photoconversion light intensity was adjusted to complete TPPc within few seconds using the 'turbo bleach' mode of the microscope.
2.5. Data processing As described earlier [12], FLIM images were constructed by the 'fast-FLIM' approach using the Sym-PhoTime64 software. Briefly, pixel lifetimes were determined by the method of moments [44]when τ avg is a difference between the barycentre of the emission decay and the time-offset of the steepest growth of the decay, t offset : where I i stands for the decay intensity at time t i . FLIM data were further processed and visualized in the Fiji software [45]. More accurate analyses of the cumulative decays from larger ROIs were done by the leastsquares reconvolution. Fluorescence was assumed to decay multiexponentially according to the formula: where τ i and α i are the fluorescence lifetime components and corresponding amplitudes, respectively. Typically, two decay components were sufficient for acceptable fits with χ 2 below 1.3. The intensityweighted mean fluorescence lifetime was calculated as: The deconvolution was performed either using the SymphoTime or the FluoFit software (PicoQuant, Berlin, Germany). Both packages allow for the socalled 'cyclic excitation correction', which corrects for the incomplete emission decay between two successive excitation pulses [46,47]. We used the correction for the deconvolution of TPE decays excited at 80MHz when excitation pulse separation of ∼12.5 ns did not allow for a sufficient decay of EGFP with ∼2.5 ns lifetime. As a result, unbiased fluorescence lifetimes were extracted from the incomplete decays collected with high repetition rate excitation, which is typical for the two-photon imaging.

Results
EGFP undergoes photoconversion to a spectrally indistinguishable form with shorter fluorescence lifetime upon intense illumination to the first singlet absorption band [12]. Here we tested whether the similar effect can be achieved with the two-photon excitation and utilize its better 3D-localization for kinetic and protein-tracking experiments using twophoton fluorescence microscopy. Figure 1 demonstrates the TPPc experiment on HEK-293T cells producing free EGFP. To prevent EGFP diffusive mixing of the photoconverted and unconverted molecules, we fixed the sample by paraformaldehyde, which immobilized the fluorophore and allowed for better visualization of the photoconverted volume. TPE fluorescence intensity and FLIM images are presented in the first and the second column, respectively. The first row shows the sample before the TPPc. From the intensity images we clearly recognize cytoplasm and nuclei of the stained cells. As expected, the mean EGFP fluorescence lifetime τ mean = 2.52 ns is mostly homogeneous across the cells both in the XY-and XZ-planes (panel B and H, respectively). Then the area marked by the white rectangle was scanned by strong 920 nm light. In order to increase illuminated volume in the Z-direction, five confocal scans with the focal plane shifted by 0.2 μm in the Z-direction were sequentially performed. The process was repeated until fluorescence from the irradiated area decreased to about 30% of the initial value, panel C. The TPE-FLIM image acquired immediately after the photoconversion is shown in figure 1(D). We can clearly distinguish the photoconverted area in the lifetime image where τ mean of EGFP decreased from 2.52 ns to 2.32 ns upon the irradiation. Confocal X-Z sections corresponding to panels A-D are shown in panels G-J. The position of the X-Z section is marked by the red line. We can see spatially well-localized volume of photoconverted EGFP. To evaluate reversibility of TPPc, the sample was stored at dark in 4°C and reimaged 24 h later. Figures 1(E), (F) document that the photoconverted area is still distinguishable with only insignificant drift in lifetimes. The corresponding X-Z sections are shown in figures 1(K), (L). Intensity and lifetime profiles across the photoconverted nucleus before and after TPPc are presented in figures 1(M), (N). The drop in the lifetime profile is clearly visible and correlates with change of the intensity profile. Figure 2 shows differences between the spatial localization of the single-and two-photon photoconversion performed side-by-side on two nucleoli of a fixed cell with EGFP-tagged nucleolar protein nucleophosmin. Compared to figure 1, we aimed at much smaller objects, since nucleoli are sub-nuclear structures. The X-Z FLIM-sections of the nucleoli reveal that the focused single-and two-photon lightbeams targeted into the nucleoli make different imprints in the Z-direction of the originally uniform lifetime distribution. As seen from the lifetime profiles in figure 2(E), the single-photon process induces photoconversion that extends almost across the whole nucleolus (FWHM 3.4 μm) compared to the better axially localized TPPc (FWHM 1.6 μm). This TPPc property allows for photoconversion of smaller volumes and targeting smaller cellular structures.
Further we investigated the two-photon excitation efficiency of the photoconverted EGFP for different excitation wavelengths. Figure 3 shows TPE confocal images of the fixed sample where EGFP in the right cell was subjected to TPPc (white rectangle). The left cell served as a control. Visual inspection of panels A and B reveals stronger TPE emission of TPPc-EGFP when it is excited by 1050 nm compared to the 850 nm excitation. Intensity profiles across the converted and control cells shown in figure 3(C) document that the two-photon excitation efficiency of TPPc-EGFP increases at longer excitation wavelength. This is further supported by the ratio of emission intensities of the photoconverted and unconverted EGFP (I TPPc-EGFP /I EGFP ) shown in figure 3(D) and overlaid with the two-photon absorption spectrum of EGFP [15]. The I TPPc-EGFP /I EGFP ratio exhibits a minimum close to the two-photon absorption peak of EGFP and increases toward longer wavelengths. Data indicate that the TPE spectrum of TPPc-EGFP is likely red-shifted compared to the one of the unconverted protein.

Exchange of NPM between nuclear structures
We examined suitability of TPPc for investigation of diffusive kinetic processes in live cells. Nucleophosmin (NPM) is an abundantly expressed multifunctional nucleolar phosphoprotein participating e.g. in cellular stress response via complex interaction network [48]. In healthy cells NPM exhibits nucleolar localization [49,50], which may be altered by drugs or pathogenic mutations [51,52]. Diverse NPM roles and cellular interactions require free trafficking of NPM in the nucleus and cytoplasm. Nevertheless, not much is known about NPM mobility and dynamics [53,54]. Figure 4 shows typical redistribution of EGFP_NPM in live HEK-293T cell upon perturbation of its equilibrium by localized TPPc. The shorter τ mean of TPPc-EGFP allows its tracking by FLIM. Figures 4(A), (B) show the equilibrium intensity and FLIM images of EGFP_NPM before photoconversion. As expected, EGFP_NPM is localized preferentially in spatially separated nucleoli, the NPM level in the nucleoplasm outside nucleoli is dramatically lower, and fluorescence from the cytoplasm is undetectable, figure 4(A). For the convenience, the nuclear area of one cell is marked by the yellow dashed line. The FLIM image in figure 4(B) reveals τ mean being 2.26 ns in the nucleoli. The slight lifetime difference between the nucleoli and the nucleoplasm results from different local environments of EGFP [55], e.g. from the different local refraction index [12,56,57]. The localized TPPc was done by a brief illumination of a small nucleolar volume by intense IR light and after few minutes the sample was reimaged, figures 4(D)-(F). From the FLIM image we can see that TPPc-EGFP_NPM with shorter τ mean spreads from the illuminated spot over the whole nucleus and appears in all spatially separated nucleoli of the photoconverted cell. The separation can be seen from the X-Z section images shown in figures 4(G), (I). TPPc-FLIM data suggest rapid exchange of TPPc-EGFP_NPM among all nucleoli of the treated cell, as documented by the decreased τ mean in the whole volume of the untreated nucleoli caused by mixing of the converted and unconverted protein. All NPM redistribution was completed within minutes.

Rapid nuclear dynamics of histone H2B
Nucleosomes are dynamic structures carrying diverse regulatory functions including gene expression, epigenetic silencing, packaging and condensing the genome, see e.g. [58,59] for review. Histone H2B is a part of this disk-shaped structure formed by the heterooctameric complex assembled from two H2A-H2B and H3-H4 histone hetero dimers wrapped by 146 base pairs of DNA [60]. Due to the tight association of histone H2B/H2A with chromatin-bound nucleosomes, their mobility is reduced and their redistribution kinetics can be followed by time-lapse FLIM experiments. Figure 5 presents rapid TPPc-induced redistribution kinetics of EGFP-labeled H2B (EGFP_H2B) in the nucleus of a single live HeLa cell. In the intensity image we can see the nucleus with several nucleoli accumulating EGFP_H2B, which exhibit uniform emission lifetime in all nucleoli. The fractional volume positioned at the equatorial plane of the largest nucleolus was briefly photoconverted by 870 nm light and a multi-frame series of FLIM images synchronized with the photoconversion pulse was immediately acquired. From figure 5(C) we can notice time-dependent mixing of the converted and unconverted fractions of EGFP_H2B in the illuminated nucleolus. The equilibration of the photoconverted protein between different nucleoli is visually less prominent, nevertheless, as documented in figure 5(D), it is measurable and significant. Figure 5(D) presents time-evolution of the τ mean in the photoconverted focal plane of the light-treated and control nucleoli. The time-dependent lifetimes in the photoconverted nucleolus were fitted by the biexponential rising model, which revealed two characteristic redistribution times of 6 s and 150 s. The fast rising component documents rapid initial mixing of the converted and unconverted proteins, which can be associated mainly with the redistribution of the mobile fraction of TPPc-EGFP_H2B within the photoconverted nucleolus. The longer component reflects slower processes likely associated with a dynamic exchange of bound photoconverted histones for the unconverted ones. As already mentioned, the control nucleolus also exhibits photoconversion-induced change in the mean EGFP lifetime, which speaks for free diffusion of TPPc-EGFP_H2B all around the nucleus resulting in a gradual enrichment of other nucleoli for the photoconverted protein and decreased τ mean . Fitting the data in the control nucleus by the biexponential model revealed characteristic exchange times of about 5 s and 250-500 s. Interestingly, the fast mixing time closely resembles the 6s rising component in the photoconverted nucleolus and the long components are also reasonably comparable. The result suggests that a significant fraction of the tagged H2B is highly mobile. Besides the internal nucleolar mixing, TPPc-EGFP_H2B escaping from the photoconverted nucleolus very rapidly diffuses through the nucleoplasm being almost instantly trapped in other nucleoli where it equilibrates. The bi-phasic equilibration kinetics in the control and photoconverted nucleolus seem to mimic, within the experimental uncertainty, each other, which is consistent with a significant fraction of highly mobile H2B in the nucleus.

Discussion
A strong blue illumination photoconverts EGFP to a form with shorter fluorescence lifetime [12]. It can be therefore used as an optically activable highlighter in kinetic FLIM applications. Since the single-photon photoconversion does not exhibit any distinct intensity threshold, the overall light dosage seems to determine the photoconversion efficiency. As a consequence, the photoconverted volume in the axial Z-direction of the microscopic sample is poorly defined. It spans essentially the whole sample thickness, even though, due to the focusing the efficiency of the conversion is not uniform throughout the sample. This limitation prevents accurate 3D positioning of the photoconverted volume within the sample, which is critical for accurate targeting of small cellular compartments in tracking applications [12,25,63]. For instance, targeting of the interior of nucleoli, whose dimensions typically do not exceed 2 μm in live cells, is hardly possible with the single-photon excitation, as seen from figure 2(E). Here we have shown that EGFP can be permanently photoconverted to the short-lived form also by the two-photon absorption process. Since this effect is nonlinear and it is supposed to depend at least quadratically on the illumination intensity, the photoconverted volume is better defined and considerably shrinks in the axial Z-direction. Consequently, it can be better localized in the examined structures, as shown in figure 2. When the TPPc is followed by the time-lapse lifetime-based imaging, it can reveal kinetic properties of both converted and unconverted proteins by simultaneous visualization of their diffusion. Since both the converted and unconverted proteins are fluorescent, mixing of their signal can be followed in time by FLIM. Such prolonged time-lapse FLIM experiments can therefore reveal also the rate of the freshly synthesized FPs, which might negatively affect standard FRAP experiments [24]. Unrecognized lifetime-based photoconversion in the vicinity of the photobleached volume, which is inevitably accompanied by changes in fluorescence intensity, may also bias quantitative analysis of both singleand two-photon FRAP with EGFP. Our approach offers milder FLIM-based alternative to the twophoton FRAP, which usually requires complete photodestruction of the fluorophore in the range of interest [36]. Importantly, TPPc is fully compatible with the two-photon imaging with all its advantages, e.g. deep imaging of thick scattering samples or elimination of artefacts rising from difference in the objective's focal length for IR photoconversion and following single photon imaging [37,64]. Unlike intensity-based FRAP, the lifetime-based FLIM is internally ratiometric and does not suffer by common artifacts associated with intensity measurements [30].
Nuclear proteins are known to be highly mobile [65]. Histones are expected to be an exception, because they bind to the chromatin as constituents of nucleoplasm-localized nucleosomes [66]. Nevertheless, consistently with literature, we found ectopically expressed EGFP_H2B histone accumulated in nucleoli [67], since their binding sites on the chromatin are occupied by the endogenous protein with a slow exchange rate. In the nucleolus, EGFP_H2B appears to be retained by entropic forces and nonspecific electrostatic interactions with charged nucleolar components. Our TPPc-induced EGFP_H2B redistribution measurements revealed highly mobile nucleolar fraction of EGFP_H2B with characteristic exchange time of 6 s, and a slower fraction with characteristic time of ∼150 s. Results qualitatively agree well with literature FRAP data where slow and fast fractions were also identified [67]. While the whole nucleolus was bleached in the FRAP experiments, we photoconverted only an equatorial layer of the nucleolus. The TPPc-FLIM data therefore reflect not only the exchange of H2B between the nucleolus and nucleoplasm but also the 3D diffusion inside the nucleolus. Newly we also documented fast exchange of EGFP_H2B between nucleoli. The almost instant appearance of TPPc-EGFP_H2B in the unconverted nucleoli suggest very high mobility of EGFP_H2B in the nucleoplasm where the protein rapidly scans the whole nucleus until it is trapped by the proper target.
In conclusion, we have revealed permanent lifetime-based photoconversion of EGFP induced by a two-photon absorption process. Besides potential FLIM artifacts, which can be caused by an unintentional photoconversion, controlled application of TPPc turns EGFP to a valuable tool for kinetic measurements. In combination with FLIM it offers higher information content compared to FRAP. Applicability of TPPc was demonstrated on measurements of a nuclear diffusion of NPM and histone H2B, which revealed high mobility of these fluorescently tagged proteins in the nucleoplasm.