1. Introduction

Fluorescent proteins (FPs) are biosensors that have revolutionized biomedical research [1]. They have found numerous applications in fields such as neuroscience [2–4], cancer research [5], and immunology [6], to name just a few. Both cultured cells and biological tissues in vivo can express FPs. Besides structural information acquired through various imaging modalities, cell dynamics can also be captured benefiting from the development of genetically encoded calcium and voltage indicators using FPs [2,7].

To visualize samples that express FPs, multiphoton microscopy (MPM) [8], combining subcellular spatial resolution, deep-tissue penetration, 3D sectioning and functional imaging capabilities [9–11], is widely adopted. MPM is based on the principle of nonlinear optics: multiple excitation photons are absorbed simultaneously by the fluorophores, yielding fluorescent photons for signal detection and image reconstruction. Compared with 1-photon fluorescence imaging technique also capable of 3D sectioning such as confocal microscopy, the most notable advantage of MPM is its deep-tissue penetration. In the realm of MPM, 2-photon fluorescence (2PF) [3], 3-photon fluorescence (3PF) [10,12] and 4-photon fluorescence (4PF) [13] imaging have all been demonstrated.

3PF imaging excited at the 1700-nm window (roughly covering 1600 nm to 1840 nm) fully exploits the deep-tissue penetration capability of MPM [10,14]. Harnessing the genetic modification capability of FPs, 3PF excited at the 1700-nm window enables visualization of neurons expressing red fluorescent proteins (RFPs) in the mouse hippocampus in vivo without damaging overlying layers for the first time [10]. In comparison with the commonly adopted 2PF imaging at other wavelengths, the principles underlying deeper 3PF imaging of RFPs are: (1) excitation at the 1700-nm window suffers from less attenuation upon propagating inside biological tissue such as the brain [10,15]; (2) 3PF better suppresses background fluorescence at the surface [10,15]; (3) it is believed that red fluorescence penetrate biological tissue better compared with shorter wavelengths, due to the less tissue scattering and intrinsic absorption [2,16].

Despite its success in neuron imaging and promising prospect in visualizing other structures and dynamics, 3PF imaging in FPs excited at the 1700-nm window has not found as many applications as 2PF imaging of green fluorescent proteins (GFPs). From the material perspective, this is largely due to the lack of detailed characterization of 3-photon excitation and emission data for a variety of FPs suitable for excitation at this window. The corresponding 1-photon fluorescence (1PF) excitation wavelength for 3-photon excitation at the 1700-nm window lies above 530 nm, suggesting that (mainly) RFPs can be excited. To the best of our knowledge, 3PF excitation spectra of only two RFPs have been measured [17], while no 3PF emission spectra have been reported for excitation at the 1700-nm window.

From the perspective of applications, the lack of 3-photon excitation and emission spectra hampers the selection of FPs and the exact excitation wavelength within the 1700-nm window, and the optimization of the MPM system for deeper and faster imaging. From the perspective of fundamental photochemistry, the lack of 3-photon emission spectra specifically poses impediment to whether Kasha’s rule [18–20], which dictates fluorescence emission from fluorophores, is still valid in 3PF from FPs.

According to Kasha’s rule, the fluorescence of polyatomic molecules always occurs from the lowest excited state, independent of which state was initially excited. Although there are indeed violations [20], this rule is generally accepted by the photochemistry community. Kasha’s rule for 1PF can be interpreted as: the emission spectrum is independent of excitation wavelength. Kasha’s rule also has the following implications for MPM: (1) the emission spectrum is independent of excitation modalities, i.e., multiphoton emission spectrum should be the same as 1PF emission spectrum. This notion is widely embraced by the photochemistry and MPM community [19] which is based on, however, limited measured results mainly comparing 2PF and 1PF emission spectra [21–24]. We also note that there are limited examples of deviation from Kasha’s rule, such as Bodipy TRoad and propidium iodide according to [23]. (2) 3PF emission spectrum is also independent of excitation wavelength. To the best of our knowledge, there has been no reported measurement and comparison of 3PF for PFs. Consequently, it is not clear whether Kasha’s rule is still valid for this higher (compared with 2PF) excitation modality.

Targeting deep-tissue 3PM applications in biomedical research, and experimental verification of the fundamental principle in photochemistry, here we demonstrate systematic measurement of 3-photon excitation and emission spectra, for selected FPs excitable at the 1700-nm window. The measured results reported here will provide guidelines for MPM using FPs, and extend the applicability of Kasha’s rule.

2. Methods

2.1 3PF excitation spectrum measurement

Relevant to the application of MPM [21,25], the multiphoton excitation property of fluorophores is characterized by the wavelength-dependent multiphoton action cross section ησ_n(λ), where η is the quantum efficiency, ησ_n(λ) is the wavelength (λ)-dependent n-photon absorption cross section. As a result, 3PF excitation spectrum is given in ησ₃(λ) throughout the paper.

The method for ησ₃(λ) measurement is the same as that in [17]. It is based on fluorescence photon counting referenced to a known standard sulforhodamine 101 (SR101), excited at 1680-nm [13]. Briefly, a tunable femtosecond laser based on intra-pulse stimulated Raman scattering [15] generated excitation pulses for 3PF covering 1600 nm to 1840 nm. 3PF signals from the selected FPs were measured by a gallium arsenide (GaAs) photomultiplier tube (h7422p-50, Hamamatsu) connected to a photon counter (SR400, Standford Reseach Systems). The counted 3PF photons were compared with 3PF photons from the reference SR101 excited at 1680 nm, from which ησ₃(λ) of the FPs can be derived, covering the entire 1700-nm window.

2.2 3PF emission spectrum measurement

The experimental setup for measuring the 3PF emission spectrum is shown in Fig. 1. Based on the measured 3PF excitation spectra of the measured FPs, they could be efficiently excited by the 1600-nm pulses. So experimentally we tuned the soliton wavelength to 1600 nm to measure the 3PF emission spectra from all FPs, unless comparison among 1600-nm, 1700-nm and 1800-nm were made for certain FPs. The solitons were generated from a 1.1-m large-mode-area (LMA) fiber (DC-200/40-PZ-Si, NKT Photonics), pumped by a 1550-nm high-energy, 500-fs fiber laser (FLCPA-02CSZU, Calmar) running at 5 MHz. A 1575-nm long-pass filter (1575lpf, Omega Optical) was used to remove the residual. Since the LMA fiber is polarization maintained, a half-wave plate was added in front, in order to align the polarization of the (linearly-polarized) laser to the main axis of the fiber. Our fiber-coupled EMCCD spectrometer (SpectraPro HRS-300, Princeton Instruments) was integrated to a multiphoton microscope (MOM, Sutter). The microscope was used to deliver excitation light onto the sample and to epi-collect the 3PF signals from the FPs. A water immersion objective lens (XLPLN25XWMP2-SP1700, Olympus) with deuterium oxide (D₂O) instead of water immersion [26] was used to focus excitation light into the sample and to collimate the epi-propagated 3PF signals, which were then coupled through a multimode fiber (QP600-1-VIS-NIR, Ocean Optics) into the spectrometer for spectrum measurement. During acquisition, we used 100x zoom with a 3 μm x 3 μm field of view, in order to achieve virtually a static focus inside the sample. We note that for 3PF emission spectrum measurement, the multiphoton microscope may not be necessary. We used it since our spectrometer was already integrated into it.

 

Fig. 1 Experimental setup for measuring 3PF emission spectrum. λ/2 plate: half-wave plate; L1: f = 30mm lens; L2: f = 75mm lens; L3: f = 25.4mm lens; LPF: 1575-nm long-pass filter; LMA fiber: large-mode-area fiber; MM fiber: multimode fiber.

Download Full Size | PDF

2.3 1PF emission spectrum measurement

In order to test Kasha’s rule, we also performed measurement of 1PF emission spectrum of the FPs, using the same EMCCD spectrometer. To excite 1PF from FPs, a 450-nm diode laser (CPS450, Thorlabs) was used for mBanana, mOrange, dsRed, tdTomato, and mStrawberry FPs. A long-pass filter (LP02-514RU-25, Semrock) was used before coupling the 1PF signals into the multimode fiber, to remove the residual excitation light. To excite 1PF from all other FPs including eRFP, mRaspberry, mKate, and mPlum, a 532-nm diode laser (CPS532, Thorlabs) was used. Due to our lack of proper long-pass filter to remove the residual 532-nm laser, 1PF emission spectrum acquisition was limited to the wavelength above 550 nm. The FPs for all 1PF emission spectrum measurement were exactly the same samples as those for 3PF emission spectrum measurement, facilitating fair comparison of the measured 3PF and 1PF emission spectra.

2.4 Sample preparation

Our reference sample for 3PF excitation spectrum measurement was 10-μM SR101 (S7635-100mg, Sigma-aldrich) dissolved in phosphate buffer saline (PBS), whose measured ησ₃(λ = 1680 nm) = 65.5x10⁻⁸⁴ cm⁶(s/photon)² [13]. The purified FPs (Creative BioMart) we used were either stock solutions (tdTomato), or diluted in PBS (all other measured FPs). A total of nine FPs were measured, including: 25-μM mBanana, 25-μM mOrange, 21-μM dsRed, 3.8-μM tdTomato, 25-μM mStrawberry, 25-μM eRFP, 25-μM mRaspberry, 25-μM mKate and 25-μM mPlum. The SR101 and FP solutions were prepared on glass slides and sealed with cover glass.

2.5 1PF excitation spectrum data

Although it is not the purpose of this paper to compare 3PF excitation spectrum and 1PF excitation spectrum, we still list the latter. The (normalized) 1PF excitation spectra for the FPs are from the online database www.fpbase.org, except for eRFP which was measured by us using a spectrophotometer (UV1780, Shimadzu). We caution that the 1PF excitation spectra from this database may not be the same as the specific FPs we purchased, but merely a reference. Those interested in 1PF excitation spectrum of FPs may also refer to [19] for a detailed review with absolute values.

3. Results and discussion

3.1 Verification of 3PF from FPs

In order to measure 3PF excitation and emission spectrum, the prerequisite is to verify that the fluorescence generated from the FPs are indeed 3PF. This was accomplished by the standard procedure: measuring the fluorescence photon counts vs excitation power, fitting the experimental data on log scale, and checking the slope to see if it was close to 3 on log-log scale [21]. Representative measured and fitted results are shown in Fig. 2, for mStrawberry and mKate, excited by 1600-nm, 1700-nm and 1800-nm pulses.

 

Fig. 2 3PF signal vs excitation power on the sample and the linear fitting on log scale for mStrawberry (A) and mKate (B). The excitation wavelengths are 1600 nm, 1700 nm and 1800 nm as indicated in the figure. Symbols: measured data; solid lines: linear fitting. The fitted slopes are also indicated.

Download Full Size | PDF

Figure 2 clearly shows the 3PF origin from both mStrawberry and mKate, by fitting the measured data on log scale: the slopes are quite close to 3.0 for the six data sets. For all the FPs at the representative excitation wavelengths (1600 nm, 1700 nm and 1800 nm) we measured, the fitted slopes (k) for 3PF signal vs excitation power fell within the range of 2.76 (for dsRed excited at 1600 nm)≤k≤3.12 (for mBanana excited at 1700 nm). This verifies that the selected nine FPs can generate 3PF excited at the 1700-nm window.

3.2 Measured 3PF excitation spectra for the FPs

Having established that the fluorescence from these FPs are 3PF, next we performed systematic measurement of the 3PF excitation spectra, given by the wavelength-dependent 3PF action cross section ησ₃(λ). We note that ησ₃(λ) for tdTomato has been measured before [17], so here we show measured results for the other eight FPs in Fig. 3, together with normalized 1PF excitation spectra. In plotting 1PF excitation spectra, we chose the wavelength span from 533 nm ( = 1600/3 nm) to 613 nm ( = 1840 nm/3), corresponding to exactly 1/3 of the wavelength span covered by our 3PF excitation wavelengths (1600 nm to 1840 nm), for better comparison between 3PF and 1PF excitation.

 

Fig. 3 Measured 3PF excitation spectra (red data, bottom abscissa and left ordinate), and normalized 1PF excitation spectra (blue line, top abscissa and right ordinate) for the selected FPs. The arrows are a guide to the eye indicating different ordinates for either 3PF or 1PF excitation spectrum.

Download Full Size | PDF

From Fig. 3 we can see that: (1) all these FPs can be excited to generate 3PF at the 1700-nm window. (2) The general trend is that toward 1840 nm, 3PF excitation becomes much less efficient. (3) We cannot directly predict 3PF excitation spectrum from 1PF excitation spectrum, as not all of them overlap. Within the measured window, only mBanana, mOrange and dsRed show overlapping between 3PF and 1PF excitation spectra (after factoring the 3 times difference in excitation wavelength). mStrawberry also shows a secondary 3PF peak (1720 nm) that coincides with the 1PF peak (575 nm). (4) Among all these measured FPs, mKate has the largest ησ₃(λ), while mBanana and mPlum are least efficiently excited at the 1700-nm window.

3.3 Test for Kasha’s rule I: comparison between 3PF and 1PF emission spectrum

Next we test whether Kasha’s rule is still valid for the excitation modality of 3PF. The measured normalized 3PF and 1PF emission spectra are shown in Fig. 4, for all the nine FPs. We note that in the measured 1PF emission spectra, the small peaks and discontinuities preceding the emission peak for mBanana, mOrange, dsRed, tdTomato, and mStrawberry, are due to the residual 450-nm excitation laser cut off by the long-pass filter. The small peak at ~800 nm for eRFP, mRaspberry, mKate, and mPlum, are due to the residual 532-nm excitation laser. These residuals were verified by measuring the spectrum from a pure PBS solution with no FP.

 

Fig. 4 Measured normalized 3PF emission spectra (red) and 1PF emission spectra for all the nine FPs. 3PF emission spectra were measured with 1600-nm excitation.

Download Full Size | PDF

From Fig. 4, it can be seen that for all these FPs, despite there is some discrepancy in the secondary peak (mBanana), slight shift of peak position (mOrange, dsRed, tdTomato, and mStrawberry), and slight spectral narrowing (dsRed, mStrawberry and eRFP), both 3PF and 1PF emission spectra show good overlapping with each other. This proves that Kasha’s rule is still obeyed when comparing different excitation modalities of 3PF and 1PF.

3.4 Test for Kasha’s rule II: comparison of 3PF emission spectrum with different excitations

Kasha’s rule in its most common understanding is that, the emission spectrum is independent of excitation wavelength. While typically tested for 1PF, we expect that this is also valid for 3PF. So we tested 3PF emission spectra with 1600-nm, 1700-nm and 1800-nm excitation. The measured 3PF excitation (Fig. 3) spectra indicate that 3PF excitation becomes rather inefficient toward 1800 nm. Consequently, we could not get measurable 3PF emission spectra with all the three excitation wavelengths for all the nine FPs. In the following we only show 3PF emission spectra from mRaspberry and mKate, excited by all the three wavelengths. It can be clearly seen that in these FPs, all the three excitation wavelengths yield identical 3PF emission spectra. This verifies that Kasha’s rule, in the sense that it is independent of excitation wavelength, is also valid for 3PF in the measured FPs (Fig. 5).

 

Fig. 5 Measured normalized 3PF emission spectra for mRaspberry (A) and mKate (B), with 1600-nm (black), 1700-nm (red) and 1800-nm (blue) excitation.

Download Full Size | PDF

4. Conclusion

In conclusion, through systematic measurement, we can provide complete 3PF excitation and emission properties of the selected FPs, excited at the 1700-nm window suitable for deep-tissue MPM. The measured 3PF excitation spectra will provide guidelines for FP and excitation wavelength selection, in order to optimize generated 3PF signals in MPM. The measured 3PF emission spectra, from the perspective of fundamental photochemistry, verifies Kasha’s rule in the following aspects: (1) emission spectra are quite similar for the excitation modalities of 3PF and 1PF; (2) emission spectra are also independent of excitation wavelength for 3PF. From the application perspective, these measured 3PF emission spectra will help collection system design in the multiphoton microscope, specifically the optimization of the signal collection system, as proper optics [14] and photodetectors [26] can be chosen targeting the desired emission spectra for most efficient signal collection from FPs.