Visible-wavelength two-photon excitation microscopy for fluorescent protein imaging

Abstract. The simultaneous observation of multiple fluorescent proteins (FPs) by optical microscopy is revealing mechanisms by which proteins and organelles control a variety of cellular functions. Here we show the use of visible-light based two-photon excitation for simultaneously imaging multiple FPs. We demonstrated that multiple fluorescent targets can be concurrently excited by the absorption of two photons from the visible wavelength range and can be applied in multicolor fluorescence imaging. The technique also allows simultaneous single-photon excitation to offer simultaneous excitation of FPs across the entire range of visible wavelengths from a single excitation source. The calculation of point spread functions shows that the visible-wavelength two-photon excitation provides the fundamental improvement of spatial resolution compared to conventional confocal microscopy.


Introduction
Simultaneous observation of different types of biological molecules is essential for investigating complex biological phenomena because most biological events are based on ensembles of different molecules or proteins. 1 Fluorescence microscopy techniques allow the simultaneous visualization of different biological molecules by spectral separation of fluorescence emission. In typical confocal microscopy, multicolor excitation is used to excite fluorescent proteins (FPs) with different emission wavelengths. 2 However, simultaneous irradiation of multiple excitation lasers may restrict the usable wavelength band for fluorescence detection and introduce chromatic aberration that can interfere with detecting colocalization or the relative distributions of different FPs. Switching the excitation wavelength is another approach; however, it causes a problem in monitoring molecules acting in temporal and spatial harmony. Since it is clear that simultaneous visualization of multiple targets provides much more information about complex biological phenomena, 3,4 increasing the number of FP in simultaneous excitation and detection is in great demand in the wide range of biological investigations. 5 For simultaneous multicolor imaging, use of multiphoton excitation has been intensively investigated. 6,7 Since many kinds of fluorophores exhibit two-photon excitation spectra blueshifted or broadened from the single-photon counterpart, a single laser line can excite more fluorophores compared to single-photon excitation. [6][7][8][9][10] As another approach, multiple excitation lines in the near-infrared (NIR) region have been utilized for exciting multiple proteins, 4,11,12 which allows the use of the whole visible wavelength range for detecting FPs, whereas multiwavelength excitation with single-photon excitation does not.
In this paper, we present a new concept of multiphoton excitation microscopy for multicolor imaging, where two-photon excitation with a single visible wavelength laser light is used for concurrent excitation of multiple FPs. Two-photon excitation is already well characterized in the NIR/IR region and has been applied to the observation of the inside of a thick sample. [13][14][15][16] Although fluorescent properties of FPs under two-photon excitation with visible light have been studied, 17,18 the feasibility of using visible-wavelength two-photon excitation of FPs in laser-scanning confocal microscopy have been restricted to theoretical investigation and the resultant imaging property remains virtually unknown. In addition to the capability of multicolor imaging, our concept offers an improved spatial resolution in three dimensions due to the nonlinear excitation using the shorter-wavelength laser.
As described in the following section, many FPs can be excited at a wavelength range of 250 to 300 nm. To access this deep UV (DUV) range for simultaneous multicolor imaging, we propose the use of a visible, in particular green, wavelength laser light and induce two-photon excited fluorescence from the FPs. As shown in Fig. 1(a), the simultaneous absorption of two photons in the visible range can excite a fluorophore in a manner equivalent to that of a DUV wavelength. The use of fluorescence excitation in the UV region has been applied in autofluorescence imaging of intrinsic fluorophores, such as amino acids, biotin, nicotinamide adenine dinucleotide (NADH), and serotonin by single-photon excitation, 19 multiphoton excitation with a visible light 20,21 and an NIR light. 6,[22][23][24] Two-photon fluorescence imaging of exogenous fluorescence dye 4',6-diamidino-2-phenylindole, dihydrochloride using visible light has also been reported recently. 25 For imaging FPs, a similar concept, which utilizes the S n state for multicolor imaging, has been demonstrated; however, in this case, FPs were excited by two-photon absorption at the NIR region. 10 In particular, two-photon excitation where the effective absorption wavelength is in the DUV range (<300 nm) energy level has been demonstrated with the observations of amino acids, 26,27 serotonin, 21 and a glass substrate, 28 but no demonstration of imaging biological cells using such visible-wavelength two-photon excitation of FPs or exogenous fluorescent probes has been reported.

Simultaneous Multicolor Imaging by Visible-Wavelength Two-Photon Excitation
Previously, the fluorescence property of FPs under excitation by using a DUV wavelength was presented only for enhanced cyan fluorescent protein (ECFP), enhanced green fluorescent protein (EGFP), and enhanced yellow fluorescent protein (EYFP). 29 We compared the absorption and excitation spectra of FPs at the DUV to the visible range as shown in Figs. 1(b) and 1(c). From these spectra, we can confirm that various types of FPs can emit fluorescence by absorbing light at ∼280 nm. Especially, FPs with shorter emission wavelengths, such as Sirius, 30 EBFP2, 31 and mseCFP, 32 show high excitation efficiency at the DUV region. Although two-photon excitation and absorption spectra are not, in general, identical to those by single-photon processes (e.g., usually blueshifted or broadened), this result indicates that the FPs can all be excited at a single wavelength in the DUV region, and this sets up the next step of attempting to excite all of them by two-photon excitation.
For a demonstration of multicolor FP excitation, a 525 nm pulsed laser light was used for simultaneous two-photon excitation of HeLa cells labeled with Sirius, mseCFP, mTFP1, 33 and EGFP, 34 in mitochondrial matrix, nucleosome, Golgi apparatus, and nucleoli, respectively. We selected these FPs for imaging because they do not have strong single-photon absorption at 525 nm. The fluorescence emission from each protein was partially separated by using dichroic mirrors and was detected by four photomultiplier tubes [ Fig. 2(a)]. In this optical system, we introduced a confocal pinhole to remove the fluorescence from the optics and the adhesion agent excited by two-photon absorption of the visible light. Figures 2(b)-2(e) show the fluorescence images obtained simultaneously by the two-photon excitation at 525 nm. Figure 2(f) is an image reconstructed by a linear spectral unmixing of each protein distribution. 35 The stained organelles are clearly recognizable with the high contrast presumably due to the nonlinear excitation. Imaging of HeLa cells with other combinations of FPs and excitation wavelengths was also demonstrated as shown in Figs. 2(g)-2(i).
In our imaging scheme, autofluorescence from DNA, amino acid residues, and endogenous molecules 36 may interfere with the fluorescence imaging. Therefore, as shown in Fig. 3, we compared fluorescence intensity from FPs with autofluorescence from the HeLa cell itself and then confirmed that the intensity of fluorescence emitted by FPs was ∼30 to 450 times higher than that of autofluorescence in the region of 410 to 500 nm. Autofluorescence intensity measured from fixed cells was stronger than that in living samples. This is due to fluorescence products generated by aldehyde fixatives reacting with amines and proteins. Other intrinsic emitters, such as NADH and flavins, are reported to be excited by multiphoton NIR excitation, 37 but do not appear with the multiphoton visible excitation used in our experiments.

Excitation and Emission Properties of FPs Under Visible-Wavelength Two-Photon Excitation
If the fluorescence emission from the FPs in the visible region is generated by two-photon excitation, the relationship between the excitation intensity and emission intensity should be quadratic. As seen in Figs. 4(a)-4(d), the fluorescence signal from all the FPs presents a clear quadratic response to the excitation intensity of 525 nm light, proving that the fluorescence signal is indeed generated by an interaction of two photons. We also measured the fluorescence spectra under two-photon excitation at 525 nm and compared them with those under the singlephoton excitation at the maximum absorption in the visible wavelength range. We performed the fluorescence measurement by using the same spectrophotometer (Acton Research Co., Spectra-Pro 320i) equipped with an electron multiplying (EM)-CCD camera (Andor Technology, iXon DU888E) for the single-and two-photon excitations. Interestingly, in the results shown in Fig. 4(e), the fluorescence spectra of Sirius appear similar for both single-photon and two-photon excitation; however, the spectra of other FPs differ. In particular, in the fluorescence spectrum of mseCFP, the subpeak usually appearing as a shoulder of the main peak has risen substantially under the two-photon excitation. Since the single-photon excitation for both visible and DUV wavelengths shows the same spectral shape, which has been confirmed by our experiment (data not shown), these results imply that the fluorescence   , and EGFP at mitochondria, nucleosome, Golgi apparatus, and nucleoli, respectively. For images (b) through (e), a mode-locked Ti:sapphire laser (Spectra Physics, Tsunami) was used for the light source and the frequency-doubled light generated through a beta-barium borate (BBO) crystal was used for two-photon excitation at 525 nm. Fluorescence was observed by four photomultiplier tubes detecting wavelength ranges of (b) 410 to 440 nm, (c) 455 to 475 nm, (d) 475 to 490 nm, and (e) 490 to 500 nm. Low-pass filtering was applied to all images. The excitation power was 0.6 mW at the object plane. (f) An image reconstructed by overlaying the unmixed fluorescence distributions. (g) to (i) Images of HeLa cells with other combinations of fluorescent proteins (FPs) and excitation wavelengths, in which linear unmixing was not applied. For obtaining the images (g) through (i), an optical parametric oscillator (OPO) system (Coherent, Mira OPO) seeded by a mode-locked Ti:sapphire laser (Spectra Physics, Tsunami) and a BBO crystal were used to produce the visible pulsed laser. Pixel dwell time: 1 ms. Pixel numbers of the images: 560 × 980 for (b) through (e); 1024 × 1024 for (g) through (i). 1.4 and 1.49 NA objective lenses were used for the observations of (b) through (f) and (g) through (i), respectively. Scale bars: 5 μm. emission pathways are different between the single-photon and the visible-wavelength two-photon excitation schemes.
In order to understand the excitation property of the FPs, we have measured the excitation spectra of the FPs as shown in Fig. 4(f). To ensure that the spectra were measured under two-photon excitation, we confirmed the quadratic response of fluorescence at the excitation wavelengths for each FP used in this plot. In this measurement, we found that the excitation efficiency increased with decreasing excitation wavelength. As a similar result was also reported in Ref. 18, our findings indicate that the FPs can be excited by resonant two-photon excitation when the excitation wavelength is close to the single-photon absorption band. In our measurement, mseCFP, mTFP1, and EGFP show a linear fluorescence response to the excitation intensity at wavelengths <525 nm, presumably due to singlephoton excitation. On the other hand, Sirius shows the quadratic response to the excitation within the entire wavelength range used for the measurement, indicating that the use of a shorter excitation wavelength may further improve the fluorescence signal from Sirius.

Spatial Resolution and Image Contrast
Nonlinear excitation offers the potential to improve the spatial resolution due to the use of the nonlinear fluorescence response.
A key point that governs the actual resolution improvement by nonlinear excitation is whether the target can absorb light at shorter than usual wavelengths. As shown above, Sirius can be excited via two-photon absorption at 525 nm and emits fluorescence at ∼424 nm. From Fig. 4(f), we can expect that shorter excitation wavelengths will also allow sufficient nonlinear absorption and provide further improvement of the spatial resolution. Therefore, we calculated the effective point spread functions (PSFs) under several different excitation conditions to compare the spatial resolution. To obtain the effective PSFs, the square of the light distribution at the laser focus was multiplied by the detection PSF at the emission wavelength 38-40 with a 0.5 Airy pinhole size that gives a spatial resolution close to the theoretical limit of confocal microscopy. 39 In the calculation results, we confirm that the effective PSF for the lateral direction shows strong dependence on the polarization of the excitation light. This is because the quadratic nature of two-photon excitation emphasizes the small difference of the light distribution between the two polarizations. The highest resolution and axial resolution were given by the linear excitation polarization as shown in Fig. 5(a). However, in practical conditions, it would be reasonable to use circular polarization in order to obtain a spatial resolution isotropic to the x and y directions, where the resolution is limited to 122 nm. One can also try to use a detection range <400 nm; however, the autofluorescence from the endogenous molecules may interfere with the image.   The result shows that a spatial resolution higher than that in a typical confocal microscopy observation can be expected in observations using visible-wavelength two-photon excitation. This calculation also confirms that the visible-wavelength two-photon excitation provides higher spatial resolution in the axial direction than conventional confocal microscopy using a similar excitation and emission wavelength.
To experimentally investigate the spatial resolution in the visible-wavelength two-photon excitation imaging, we observed fluorescence beads with a diameter of 100 nm as shown in Fig. 6. The fluorescence beads can be excited at the DUV  range and emit fluorescence at ∼440 nm. We excited the fluorescent beads by using a pulsed laser light at 500 nm and detected the fluorescence in the wavelength range of 417 to 477 nm. As shown in Figs. 6(a), 6(c), and 6(d), the spatial resolutions and polarization dependence similar to the calculation results are observed in the experiments. The axial resolution presented in Figs. 6(b) and 6(e) also shows a value close to the theoretical result. We also confirmed that the fluorescence imaging in Figs. 6(a) and 6(b) were performed by two-photon excitation by measuring the fluorescence and emission relation with the excitation at 500 nm and the capability of fluorescence excitation in the DUV range as shown in Figs. 6(f) and 6(g).
In addition to the fundamental improvement of the spatial resolution, we can expect that visible-wavelength two-photon excitation provides high image contrast and suppression of the background. Figure 7 shows the comparison of images taken by confocal single-photon and visible-wavelength twophoton excitation. mseCFP expressed in the nucleosomes of a living HeLa cell was imaged by 405 nm for single-photon excitation [Figs. 7(a) and 7(b)] and 525 nm for visible-wavelength two-photon excitation [Figs. 7(c) and 7(d)]. The image contrast itself is significantly improved in the visible-wavelength two-photon excitation scheme, presumably due to the localization of the excitation volume within the laser focus. The localization of the excitation itself is key since even a very small pinhole cannot eliminate the out-of-focus signal completely. Although the theoretical resolutions are similar in both observations, the higher optical sectioning capability with the nonlinear excitation significantly contributes to enhancing the image contrast and reveals cellular interiors more distinctly.

Time-Lapse Imaging and Photostability
Since labeling a sample with FPs is useful in the investigation of cells under physiological conditions, we also examined the capability of visible two-photon excitation in the observation of living cells. HeLa cells expressing Sirius at mitochondria were excited by pulsed laser light at 525 nm wavelength. The cells were kept at 37°C in an incubation chamber on the microscope stage in order to maintain the cell viability during the observation. As shown in Fig. 8, the dynamics of mitochondria expressing Sirius in a HeLa cell was clearly visualized. In this experiment, we did not observe a rapid morphological change in the cell during the image acquisition. However, the damage seems to have accumulated to some degree with irradiation, since some change in the shape of mitochondria toward a round or ring form was observed after the image acquisition. The change of the mitochondria shape after imaging was more evident under excitation with higher power.
We also examined the photobleaching properties of the FPs used in the above experiments. For this purpose, we investigated the decay of fluorescence intensity during fluorescence imaging of a living HeLa cell expressing single FP, as shown in  during imaging as seen in the result using Sirius and mTFP1 due to the movement of the stained organelles. The photobleaching efficiency is stronger in the FPs with a longer excitation wavelength. This might also be explained by the high excitation efficiency due to the resonance to the single-photon absorption. This experimental result also confirms that it is important to consider the bleaching property of FPs for performing multicolor time-lapse imaging, and the choice of FPs is still important even in single wavelength excitation.
To demonstrate the stability of the imaging and resilience to photobleaching of out-of-focus regions, we also performed volumetric imaging of a HeLa cell expressing Sirius, EGFP, and mTFP1 in the mitochondrial matrix, nucleoli, and nucleosomes, respectively [Figs. 10(a)-10(c)]. In this observation, x − y images of each sample were recorded for each slice in the stack, starting from the bottom. To reconstruct a three-dimensional (3-D) image, we used volume rendering of the x − y image stacks. Figures 10(a)-10(c) are composed of 19, 33, and 42 x − y images, respectively. The resulting images clearly show how this method enables volumetric observations with high optical sectioning capability derived from nonlinear excitation.

Simultaneous Single-and Two-Photon
Excitation Imaging The number of FPs for simultaneous observation can be increased by combining conventional single-photon excitation.
There are a variety of FPs that can be excited by single-photon absorption at ∼525 nm wavelength, such as Venus, 41 DsRed, 42 and mCherry. 43 Therefore, simultaneous excitation of FPs by using both single-and two-photon absorption is practical for expanding the usable emission wavelength range for FP imaging. As a demonstration of this approach, we observed a living HeLa cell expressing mTFP1, EGFP, Venus, and DsRed at Golgi apparatus, nucleoli, actin filaments, and mitochondria, respectively, as shown in Fig. 11. We used an optical setup shown in Fig. 11(a) with a 525 nm pulsed laser light in this observation. Figures 11(b)-11(e) show fluorescence images obtained by each detection channel in Fig. 11(a). Figure 11(f) was reconstructed by applying the spectral unmixing 35 to Figs. 11(b)-11(e). This demonstration confirms that the organelles stained with FPs can be concurrently visualized by using a single excitation wavelength.

Discussion
In this paper, we reported the first use of visible-wavelength two-photon excitation of multiple FPs for simultaneous multicolor fluorescence imaging. The quadratic fluorescence response against the excitation intensity shown in Figs. 4(a)-4(d) clearly demonstrates that FPs were excited via two-photon absorption. DUV excitation has not been an obvious choice for FP imaging due to phototoxicity. By exploiting visible light for two-photon excitation, we attain the benefits of DUV excitation without the phototoxicity induced by direct DUV irradiation and the necessity of special optics designed for DUV light. Due to the short-wavelength nonlinear excitation, the two-photon excitation at the visible wavelength provides 3-D imaging capability with high spatial resolution. The use of nonlinear excitation additionally suppresses the background signals from out-offocus planes, resulting in a substantial improvement of image  contrast. 44,45 Despite the fact that using a short excitation wavelength compromises the imaging depths offered by NIR light, which provides a major advantage for conventional two-photon excitation microscopy, our work clearly demonstrates that the visible-wavelength two-photon excitation can be utilized to allow multichannel imaging of subcellular structures. One of the major advantages in the visible-wavelength twophoton excitation is the simultaneous multicolor imaging capability with less chromatic aberration. In the visible-wavelength two-photon excitation mode, we need to use FPs that have almost no light absorption at the excitation laser wavelength. This requirement limits the number of simultaneously observable targets. However, this limitation can be overcome by combining conventional single-photon excitation of FPs, as demonstrated in Fig. 11.
Another important advantage of the technique is the high spatial resolution due to the nonlinear excitation. As discussed with Figs. 5 and 6, the visible-wavelength two-photon excitation mode realizes the spatial resolution as higher than conventional confocal microscopy using a visible wavelength for excitation. The improved resolution enables us to observe the 3-D structures in cellular organelles that are difficult to visualize in conventional confocal microscopy. The suppression of background fluorescence by a combination of nonlinear excitation and confocal detection also effectively contributes to the ability to recognize small structures near the diffraction limit with high image contrast. 44,45 In our experiment, we found that the single-and two-photon excitations of FPs at DUV region exhibit different emission spectra. While some reports of a difference in emission spectra between single-and two-photon excitation have emerged, the mechanism has not been well understood. 46 For the fluorescence response observed in our experiments, two different photophysical pathways can be proposed. One is Förster resonance energy  transfer (FRET) from two-photon excited amino acid residues, such as tryptophan, tyrosine, and phenylalanine. The other pathway is initiated by the direct excitation of a fluorophore to its upper excitation state. Visser et al. have investigated the excitation and emission pathways of ECFP, EGFP, and EYFP (Ref. 29) under DUV single-photon excitation through fluorescence lifetime measurement experiments and concluded that both pathways could be taken for the fluorescence emission. We consider, in our visible-wavelength two-photon excitation mode, that the fluorescence emission might be caused predominantly by direct fluorophore excitation and not by FRET. As reported in Ref 18, when the excitation wavelength is close to the single-photon absorption band, the enhancement of two-photon excitation by the resonance effect can be expected. Since it is unlikely that FRET causes such a spectrum difference, our experiment may also indicate that the direct excitation is the main pathway for the fluorescence emission under visible-wavelength two-photon excitation. Our experimental result may then show that the emission spectrum is not necessarily independent of the excitation pathway, which may give a new insight for understanding the photophysical mechanism of fluorescence excitation and emission. In addition, if the direct DUV excitation of fluorophore is possible, the visible-wavelength two-photon excitation can be applied to fluorescence observation using other types of probes. As an example, we have confirmed that ATTO488 dyes can produce fluorescence emission through two-photon excitation with a 560 nm pulsed laser light as shown in Fig. 12.
To apply the presented technique to various types of live cell imaging, the phototoxicity given by DUV excitation can be an issue. Although it was possible to observe damage in some cases, compared to single-photon excitation at DUV, two-photon excitation produces less damage since absorption is essentially localized to within the focal volume and is negligible in out-of-focus planes. In addition, as also discussed above, provided the DUV excitation of fluorescence excitation is presumably enhanced by resonant two-photon absorption, the DUV excitation efficiency of the endogenous molecule would be much lower than that of FPs, which can explain our observation of a weak autofluorescence background. Optimization of excitation power and exposure time allows imaging of protein dynamics in living cells for sufficient observation times without observing morphological changes in intracellular structures.  Using detectors with higher quantum efficiency, such as an EM-CCD camera, together with multifocus excitation [47][48][49] can further reduce the light exposure and extend the application of the technique to a wide range of biological samples. The opening up of a new multiphoton regime for multicolor fluorescent protein imaging, which allows simultaneous singleand two-photon imaging, should continue to be optimized with further work and lead to a large number of simultaneous targets in fluorescence imaging. This multiplexing limitation is probably the primary drawback of fluorescence microscopy in the current state of the art, since recent resolution advances in imaging now allow exquisite spatial detail. The cellular feature specificity that is the key role of probe chemistry is quite limited by the ability to imaging multiple concurrent targets, and the method we demonstrate here shows a practical method for breaking this limitation. Although some issues still remain in our multicolor imaging technique, the high spatial resolution and the low chromatic-aberration characteristic are additionally advantageous for cellular imaging. The advantages in our technique can also boost confocal microscopy based techniques, such as fluorescence-lifetime imaging microscopy, fluorescence correlation spectroscopy, and fluorescence cross-correlation spectroscopy. On top of that, the visible-wavelength two-photon mode can be combined with laser scanning based super-resolution techniques, such as stimulated emission depletion 50 and saturated excitation microscopy. 51