Photobleaching Analysis of Fluorescent Proteins in Two-Photon Microscopy at High Repetition Rates

Abstract

Severe photobleaching in two-photon excitation fluorescence microscopy (TPM) has limited its potential for long-term and quantitative imaging of live cells and tissues. One solution is to excite fluorophores with a high-repetition rate which reduces the peak intensity of irradiance. However, there is a lack of knowledge about the general utility of this strategy for fluorescent proteins. Here, using simple photobleaching assay, we studied the photobleaching of EGFP and tdTomato at 80 MHz and 640 MHz repetition rates. Our analysis shows that the impact of high repetition rate excitation on reducing photobleaching in TPM is highly dependent on the excitation wavelength and fluorophores. Our results are useful for selecting optimal parameters to minimize photobleaching and photodamage in TPM.

1. Introduction

Two-photon excitation fluorescence microscopy (TPM) is a powerful and indispensable tool for studying tissues and in vivo biological samples at the subcellular levels [1–3]. Its ability to deliver deep-tissue high-resolution imaging has transformed our understanding of molecular and structural characteristics as well as dynamic cellular processes [4, 5]. Recent advances in fluorophores [6], light sources [7] and adaptive optics [8] have allowed TPM to be widely used in neuroscience [9, 10]. However, the fluorescence signal loss due to severe photobleaching [11–14] has limited the full potential of TPM for quantitative long-term imaging [15–17]. In particular, imaging fluorescent proteins with minimal photobleaching remains a substantial challenge because of high-order photobleaching processes [18–21].

A plethora of approaches have been used to increase the photostability and signal of fluorescent proteins in TPM. For example, an adaptive illumination mitigated the problem by avoiding excessive laser exposure to the samples [22, 23]. Longer excitation wavelength has been shown to decrease the photobleaching rate of enhanced green fluorescence protein (EGFP) and DsRed2 by 50% and 90%, respectively [24–26]. Using different repetition rates for the laser other than 80 MHz has been another solution. For instance, a low repetition rate laser (<1 MHz) resulted in a 25-fold increase in the total fluorescence yield of EGFP by permitting triplet or dark states to be relaxed to the ground state before a subsequent pulse arrives.[27] It is rather challenging to apply this method in TPM, but it is possible to fulfill the condition with one or a few pulses per pixel by exploiting high-speed scanners [28, 29] or gating pulses [30, 31].

Another approach is to employ a high repetition rate laser (>80 MHz) [32–34]. The photobleaching rate is proportional to $P^n/{\left(\tau f\right)}^{n-1}$ where $P$ denotes the average power, $\tau$ the pulse duration, $f$ the repetition rate and $n$ the order of the photobleaching process [15]. Since $n$ is typically larger than two [11], one can alleviate photobleaching by increasing the pulse repetition rate to $N$-fold [32] while maintaining the excitation rate by increasing average power by $\sqrt{N}$-fold. Indeed, a 4-fold and 9-fold decrease in photobleaching has been demonstrated in fixed brain samples and live C. elegans labeled with EGFP [32]. However, a recent study with ~1.6 GHz femtosecond laser reported insignificant improvement in photobleaching of mKate [35] and thus it raises doubts about the general applicability of the high repetition rate excitation to lessen photobleaching in TPM.

Here we use a simple and robust photobleaching assay to characterize two-photon photobleaching of EGFP and tdTomato. Our analysis shows that the effect of high repetition rate excitation of TPM on reducing photobleaching strongly depends on the excitation wavelength as well as fluorophores. We also analyze other fluorescent probes to provide guidelines for optimal TPM imaging.

2. Experimental method

2.1. Two-photon imaging system

A femtosecond beam from an ultrafast laser system (InSight X3+, Spectra-Physics) was split into two optical paths for experiments with 80 MHz and 640 MHz repetition rates. In the first path (640 MHz), the beam was sent to an 8× pulse splitter [32, 36] consisted of three beam splitters (BSW29R, Thorlabs), three delay arms, a half wave plate (AHWP05M-980, Thorlabs) and a polarizing beam splitter (PBS102, Thorlabs). This configuration produces eight pulses separated by ~0.25 ns. In the second path (80 MHz), the beam was simply routed to a beam splitter which recombined the first and second beam. During the 640 MHz or 80 MHz experiments, the other beam path was blocked. We avoided to use flipping mirrors to secure the stability of imaging system. The recombined beams were spatially filtered by a pinhole (P30D, Thorlabs) and collimated before entering a microscope.

We used a custom built two-photon excitation fluorescence imaging system. Briefly, the collimated beam was reflected by a dichroic mirror (FF750-SDi02-25×36, Semrock) and sent to a dual-axis galvo mirror system (Saturn-5, ScannerMAX) with a relay system consisted of a scan lens (SL50-CLS2, Thorlabs) and tube lens (TTL200MP, Thorlabs). The collimated beam was focused by a water immersion objective lens (UPLSAPO60XW, 60x/1.20, Olympus). An optical isolator (Model 714, Conoptics) was installed to prevent the back reflected light from other optics. The illumination power was adjusted by a polarizing beam splitter and a half-wave plate mounted in a motorized rotation stage (PRM1Z8, Thorlabs). The illumination time was controlled by an optical shutter (SH05R, Thorlabs). We measured the pulse widths of beams before the objective using an autocorrelator (FSAC, Thorlabs), and confirmed that both beams at 80 MHz and 640 MHz showed similar pulse width. All the power values were measured right after the objective.

The fluorescence signal was collected by the same objective lens, passed through the same path as the excitation beam, transmitted though the dichroic mirror, filtered by a set of filters (FF01–525/50–25 and FF01–750/SP-25, Semrock) and focused onto a multimode fiber (M101L02, Ø150 μm, Thorlabs) by an imaging lens (f = 100 mm, Thorlabs). The multimode fiber was connected to an avalanche photodiode (SPCM-AQRH-13-FC, Excelitas Technologies) whose signal was read by a timer/counter board (PCI-6602, National Instruments). A three-axis piezo stage (NanoMax 300, Thorlabs) was used to hold and move samples. The galvo mirror system conjugated to the back focal plane of objective was used to scan the illumination beam and controlled by an analog output board (PCI-6733, National Instruments). All the measurements were controlled by a custom-made LabVIEW program (National Instruments). If necessary, we used 491 nm laser (Calypso, Cobolt) for one-photon excitation with an additional dichroic beam splitter (Di03-R405/488/561/635, Semrock).

2.2. Sample preparation

The stock solutions of EGFP (TP790050, OriGene), tdTomato (TP790045, OriGene), Alexa Fluor 488 (A20000, ThermoFisher) and Oregon Green 488 BAPTA-1 (O6806, ThermoFisher) were diluted with a 90% glycerol (49767, Sigma) in 1× phosphate buffered saline (PBS, pH 7.4, AM9625, ThermoFisher). The Oregon Green 488 BAPTA-1 solution included a 2 mM of CaCl₂ (21097, Sigma). At least 200 μL of solution was introduced into an eight-well chamber (LabTek II, 155409, ThermoFisher).

3. Results

3.1. Photobleaching assay

To investigate the effect of high repetition rate excitation in TPM, our experiments were designed to avoid (i) the compounding effect of triplet state relaxation resulting from scanning based assays [27, 32] and (ii) the possibility of denaturation of the surface-bound fluorescent protein whose fluorescent properties may be altered [37]. We measured the fluorescence intensity time traces of a freely diffusing EGFP solution (~4 μM) with a tightly focused femtosecond laser beam operating at 80 MHz or 640 MHz repetition rate through a water immersion objective lens (NA = 1.2). The 640 MHz pulse trains were generated by a passive pulse splitter [32, 36] whose inter-pulse time in the pulse bunch was ~0.25 ns. (Fig. 1a and 1b) To achieve the same excitation rate, the average excitation power at 640 MHz was set to be 2.8-times higher than that at 80 MHz. The pulse duration for both repetition rates was ~150 fs.

Fig. 1.

(a) Pulse schemes of conventional (80 MHz) and high repetition rate (640 MHz) excitation. For the higher repetition rate, the peak power was reduced to 35%. (b) Experimental setup for generating pulse trains with an 80 MHz or 640 MHz repetition rate. BS, 50:50 beam splitter; HWP, half-wave plate; PBS, polarizing beam splitter; PH, pinhole. (c) Fluorescence intensity time traces of EGFP in 90% glycerol measured at 80 MHz (black) and 640 MHz (red) repetition rate. $F_0$ represents the initial photon count. (d) Normalized fluorescence intensity time trace of EGFP subjected to light that was turned on and off. $R$ denotes the remaining fluorescence level after fluorescence decay.

We tested various percentage of glycerol in 1× PBS solution to slow down the diffusion time and minimize denaturation of fluorescent molecules (Fig. S1). We used 90% glycerol [38] for all the measurements although its viscosity is slightly higher than that of cells [39]. As shown in Fig. 1c, the fluorescence time traces reached equilibrium in less than 200 ms as photobleaching balances the replenishment of fluorescent molecules via diffusion. While periodically turning on and off the laser, the fluorescence signal is fully recovered (Fig. 1d), indicating that the amount of photobleached molecules is negligible during the measurements compared to the sample volume (200 μL).

The decay curves are well described by the steady-state approximation of dynamics of fluorescent molecules in the observation volume of TPM [40]. We used two observables for our analysis as indicated in Fig. 1, i.e., the initial fluorescence intensity ($F_0$) and the residual fluorescence level normalized to $F_0\left(R\right)$ as described by

$$F_0=\alpha N{k}_{exc}$$ (1)

$$R=\frac{k_d}{k_d+k_b}=\frac{1}{1+{\tau }_d{k}_b}$$ (2)

where $\alpha$ is a constant including the fluorescence quantum yield and detection efficiency, $N$ the number of fluorescent molecules in the excitation volume, $k_{exc}$ the excitation rate, $k_d$ the diffusion rate $(1/{\mathrm{\tau }}_{\mathrm{d}})$, $k_b$ the photobleaching rate [12, 40]. We neglected photobleaching in the initial fluorescence intensity.

3.2. Wavelength-dependent photobleaching

We examined wavelength-dependent photobleaching of EGFP at 80 MHz and 640 MHz with an average power of 6.5 mW and 18 mW, respectively (Fig. 2a). The initial fluorescence intensity curves corresponded well to the previous reports of two-photon absorption spectrum [6] and the molecular brightness spectrum [33] peaking at ~920 nm (Fig. 2b). Notably, at higher repetition rate it exhibited an increase of residual fluorescence levels at all wavelengths examined from 820 nm to 980 nm (Fig. 2c). The higher repetition rate yielded 1.2–1.4-fold higher photon counts (Fig. 2e), which is attributed to pulse saturation with the excitation power used at 80 MHz [12, 40]. To confirm this we analyzed the log-log plot of the initial fluorescence intensity over the excitation power [41]. It showed a slight deviation from quadratic power dependence with a slope of 1.8 ± 0.1 at 80 MHz (Fig. S2). However, non-negligible photobleaching in the initial photon count may also contribute to the higher $F_0$.

Fig. 2.

(a) Time traces of normalized fluorescence intensities for 4 μM EGFP in 90% glycerol at different excitation wavelengths with a repetition rate of 640 MHz. Wavelength dependent initial fluorescence intensity (b), residual fluorescence levels (c), and the corrected photobleaching probability (d) measured at 80 MHz and 640 MHz. Data are the mean ± standard error of mean from >10 measurements. (e) The ratio of the initial fluorescence count at 640 MHz to that at 80 MHz. (f) The ratio of the corrected photobleaching probability at 80 MHz to that at 640 MHz.

For accurately characterizing the wavelength dependence of photobleaching, we devised the corrected photobleaching probability (${q_b}^{\mathrm{*}}$) which reflects the excitation rate and the wavelength dependence of the observation volume and diffusion time (Fig. 2d) [See the Supporting Information for the details.]. To ensure this is a rational analysis, we measured the fluorescence time traces at 880 nm and 920 nm with adjusted powers that made the initial fluorescence intensities the same for both cases. Indeed, they displayed similar residual fluorescence levels as their corrected photobleaching probabilities (Fig. S3).

Based on our analysis for EGFP, we found that (i) the photobleaching probability at 640 MHz is 1.4–2.6-times lower than that at 80 MHz (Fig. 2f), which is in good agreement with the previous study [32], (ii) the lowest photobleaching occurred at ~880 nm where the photobleaching probability is 4–5-fold lower than at 980 nm, (iii) a higher repetition rate showed a pronounced effect of reducing photobleaching at <880 nm and (iv) wavelengths longer than 940 nm were not effective to reduce photobleaching. The latter is puzzling because the longer wavelength is known to be advantageous to reduce photobleaching [24–26, 42]. Other fluorophores that we studied below also favor the longer wavelengths. It is possible that >940 nm may facilitate a transition of EGFP to higher excited states, leading to severe photobleaching although further studies with transient absorption measurements are required.

At the two-photon absorption maximum of EGFP (920 nm), we measured the power dependence of fluorescence at 80 MHz and 640 MHz (Fig. S4). At high excitation powers (> 10 mW) both cases showed similar residual fluorescence levels. However, considering the enhanced fluorescence signal due to the reduced pulse saturation, the higher repetition rate showed 1.5–2.2-fold less photobleaching probability at the power levels tested.

3.3. Fluorophore-dependent photobleaching

Next, we examined the effect of high repetition rates on tdTomato, a bright red fluorescent protein. A 2 μM tdTomato solution was illuminated with an excitation power of 6.5 mW (80 MHz) or 18 mW (640 MHz) where a log-log plot of the initial fluorescence intensity over the excitation power showed a slope of 1.9. Similarly to EGFP the initial fluorescence level of tdTomato was well followed by the two-photon absorption spectrum in the wavelengths ranging from 760 nm to 1040 nm (Fig. 3a) [6]. We observed ~1.2-fold higher initial fluorescence level and 1.2–2-fold reduced photobleaching probability at 640 MHz than at 80 MHz (Fig. 3b and 3c). Interestingly, a higher repetition rate exhibited little impact on photobleaching near the two-photon absorption maximum. It only showed a decrease by 20–30%, similar to the previous report on mKate2 [35]. Since a recent study reported an extremely high order photobleaching process in tdTomato [21], we expected significant reduction of photobleaching at the high repetition rate. However, this discrepancy may be explained by the fact that in the previous study a ~40-times lower excitation intensity was used.

Fig. 3.

Wavelength dependent initial fluorescence intensity (a), corrected photobleaching probability (b), and the ratio of $F_0$ and ${q_b}^{*}$ (c) at 80 MHz to that at 640 MHz for tdTomato in 90% glycerol. (d) Ratios of ${q_b}^{*}$ at 80 MHz to that at 640 MHz for Alexa Fluor 488 (purple) and Oregon Green 488 BAPTA-1 with [Ca²⁺] = 2 mM (green).

Lastly, for comparison we also studied Alexa Fluor 488 and Oregon Green 488 BAPTA-1 which are commonly used in immunofluorescence and calcium imaging, respectively. They showed 2-fold and 2.3-fold improvement on the reduction of photobleaching across wavelengths (Fig. 3d and S5). Unlike EGFP and tdTomato, these organic dyes showed less wavelength dependence of photobleaching with high repetition rate excitation.

4. Conclusions

In summary, we investigated the effect of high repetition rate excitation in TPM on the reduction of photobleaching for two representative fluorescent proteins, EGFP and tdTomato. We confirm that higher repetition rate in TPM had the benefit of reducing the photobleaching rates and enhancing the fluorescence yield for fluorescent proteins and organic dyes. This effect strongly depends on the excitation wavelength and fluorescent probes, although we tested only an 8-fold higher repetition rate than the commonly used 80 MHz ultrafast system. There has been concern about using high-repetition rate excitation, primarily because a shorter interpulse time than the fluorescence lifetime may lead to excited-state absorption processes. However, our study, consistent with a previous report [32], suggests that the major photobleaching pathway of EGFP and other fluorophores in TPM is likely due to a successive multi-photon absorption process rather than sequential absorption at a singlet excited state, although the latter is not entirely negligible. Our simple photobleaching assay will provide useful guidelines of the optimal excitation wavelengths and the effectiveness of high repetition rate for many fluorescent proteins. This will be valuable when obtaining high signal to noise ratio images with minimal photodamage in TPM.