Understanding the Photophysics of the Spinach-DFHBI RNA Aptamer-Fluorogen Complex to Improve Live Cell RNA Imaging

Chemical synthesis of DFHBI

DFHBI was synthesized from 4-hydroxy-3,5-difluorobenzaldehyde as a starting reagent following the procedure of Paige and coworkers (see the SI for details).¹¹

Preparation of Spinach RNA aptamer

DNA coding for the Spinach aptamer was cloned into the pET28a plasmid (Novagen). Spinach was subsequently transcribed in vitro from the T7 promoter using MEGAshortscript T7 Kit (Ambion) according to the protocol provided by the manufacturer. Then the transcribed RNA was buffer-exchanged with a P-6 micro bio-spin column (Bio-Rad) three times into storage buffer (10 mM Tris-HCl, pH 7.5, 0.1 mM EDTA and 10 mM KCl) to remove free, unincorporated nucleotides. The Spinach RNA was then folded in 40 mM HEPES buffer (pH 7.5) with 125 mM KCl by incubating at 90 °C for 2 minutes and slowly cooling down to room temperature. At 65 °C, MgCl₂ was supplemented to a final concentration of 5 mM, and cooling was resumed.¹¹ For experiments that require attachment of Spinach to surfaces, a biotin-labeled DNA oligonucleotide (5’-TTATCCGCTCACAATTCCCATTTG-biotin-3’, Integrated DNA Technologies) was annealed to Spinach 5’-end, by directly mixing the biotin-DNA with Spinach during folding as described above.

Surface immobilization of GFP and Spinach

A flow chamber was prepared on a PEG (polyethylene glycol) coated coverslip with a low density of biotin-PEG (Laysan Bio).¹⁴ Neutravidin (0.2 mg/mL, Thermo Scientific) was incubated for 5 min and washed away with 1× PBS buffer (Lonza #17-517Q). Freshly folded biotinylated Spinach aptamer (10 nM) was incubated for 10 min in an imaging buffer consisting of 40 mM HEPES (pH 7.5) with 125 mM KCl and 5 mM MgCl₂.¹¹ Then various concentrations of DFHBI (50 μL) in the same buffer was loaded to the chamber for the measurements. The surface density of the immobilized RNA was roughly 5–20 molecules per focal volume. In the case of fluorescent proteins, recombinant Aequorea coerulescens GFP (rAcGFP1) was purchased from Clontech and used without further purification. The spectral properties of rAcGFP1 (hereinafter simply GFP) are reported to be similar to those of enhanced GFP.¹⁵ To immobilize GFP to the PEG surface, a biotinylated rabbit polyclonal anti-GFP antibody (Rockland Immunochemicals) was incubated on Neutravidin-coated flow chambers for 5 min and then washed out. After incubating GFP (5 nM) in the flow chamber for 10 min, unbound excess proteins were flushed before measurements were taken.

Fluorescence intensity and lifetime measurements

We used a custom built confocal microscope to detect fluorescence of GFP and Spinach, unless specified otherwise. ¹⁶ A diode laser (473 nm, CNI laser) was used to illuminate the sample, and the illumination time and level of light intensity were controlled by acousto-optical filters (MT200-A0.5-VIS, AA Opto-Electronic). The fluorescence lifetime measurements were performed by the same microscope with an ultrafast laser (MaiTai HP, Spectra Physics) and a time correlated single photon counting module (SPC630, Becker Hickl).¹⁷ Please refer to the SI for more information.

Binding and unbinding kinetics measurements

The fluorescence increase was recorded with a fluorometer (Cary Eclipse) after rapidly mixing 40 nM of Spinach with different concentrations of DFHBI in a cuvette (80 μL). The excitation and emission wavelengths were 460 nm and 510 nm, respectively. The illumination intensity was estimated to be 1–5 W/cm². The fluorescence intensity followed a single exponential time course.

Live cell imaging in E. coli

Rosetta (DE3) competent cells (Novagen) were transformed with pET28c-tRNASpinach (provided by Prof. Samie Jaffrey) which encodes a Spinach aptamer inserted into the tRNA^Lys3 sequence. Cultures were grown overnight with shaking in Luria Broth (LB) containing 50 μg/mL kanamycin at 37 °C. Then aliquots of the static culture were added to 5 mL LB containing 1 mM IPTG to reach OD₆₀₀ = 0.4 and induced at 37 °C. After 2.5 h, 100 μL aliquots of cells were added to individual wells of 8-well Lab-Tek chamber (NUNC #155411) coated with poly-lysine and cells were allowed to adhere for 30 min at 25 °C. Cell-coated surfaces were gently washed two times with 200 μL modified minimal M9CA broth (Teknova #M8011) with glycerol supplemented with 1 mM IPTG, and then incubated for 30 min at 25 °C in 200 μL M9CA containing 1 mM IPTG and 100 μM DFHBI. We used a home-built epi-illumination microscope to image live cells with an oil immersion objective lens (NA = 1.4, UPlanSApo 100x, Olympus) and an EMCCD camera (iXon DU-897, Andor). The illumination time of 473 nm light was synchronized to clock signals of the camera and controlled by a custom program. Image analysis was performed using MATLAB (MathWorks) and ImageJ (NIH) software.

Fast fluorescence intensity decay and recovery of the Spinach-DFHBI complex

Figure 1b shows typical fluorescence intensity time traces of the surface immobilized fluorescent protein and Spinach at high density coverage under a focused laser beam. While periodically turning on and off blue laser light (473 nm) with 0.2 kW/cm² of excitation intensity (I_exc), the fluorescence signal was monitored by confocal microscopy at room temperature. As expected, the fluorescence intensity of GFP decreases slowly and does not recover in subsequent pulses. It has been well known that GFP’s signal drop comes from irreversible photobleaching¹⁸ and/or photoconversion to a long-lived dark state.¹⁹

In contrast, applying the same excitation scheme to Spinach in the presence of excess DFHBI (5 μM), the initially bright fluorescence of the Spinach-DFHBI complex shows fast decay to a significantly reduced fluorescence level (< 5% of the initial level) within 2 seconds. Remarkably, after switching off the excitation light for a few seconds, a subsequent pulse reveals that the fluorescence signal is recovered back to 85–95% of the initial signal. This indicates that the fluorescence loss of Spinach-DFHBI is not due to fast, irreversible photobleaching but instead is caused by reversible conversion to a nonfluorescent state and that the signal recovery occurs spontaneously without light.

Power-dependent fluorescence decay

We next examined how the excitation intensity I_exc affects fluorescence decay (Figure 2a). The fluorescence decay of GFP becomes faster at higher I_exc and its decay rate (k_off) increases linearly with I_exc (Fig. 2b). For Spinach, however, k_off exhibits weaker dependence on I_exc, i.e., it changes only two-fold, from 11.7 to 21.2 s⁻¹, when I_exc increases 25-fold, from 0.2 to 5 kW/cm², a range of I_exc commonly used in confocal microscopy and single molecule imaging (Figure 2b). The very weak dependence of fluorescence decay rate on I_exc suggests that fluorescence loss requires at least two steps, the first, light-dependent step already saturating at our typical I_exc values and the second, light-independent step.

In free solution, fluorescence decay would be difficult to detect because of replenishment of fluorescently active species from outside the excitation volume via diffusion. This was indeed the case for GFP (Figure 2c). For Spinach-DFHBI, however, photoconversion is very efficient, saturating even at moderate I_exc values (Figure 2a), and as a result, we could observe significant fluorescence decays even without immobilization (Figure 2c). To achieve the same degree of fluorescence decay for GFP, we had to use 1000-times higher I_exc (Figure 2c and Figure S1). This set of measurements in free solution further indicates that the weak dependence of fluorescence decay kinetics of immobilized Spinach-DFHBI on I_exc is due to a saturation of the highly efficient photoconversion process (please see the SI for further discussion).

Fluorescence lifetime of the Spinach-DFHBI complex

We examined the time-resolved fluorescence of the Spinach-DFHBI complex in two different ways to gain further insights. First, the fluorescence lifetime was measured when > 80 % of the complex is photoconverted after 8 seconds of illumination (Figure 3, top left). Second, the fluorescence lifetime was measured during 20 ms pulse of illumination time every 2 s to ensure collecting fluorescence mainly from the bright molecular complex before photoconversion (Figure 3, top right). Here we used 0.5 μM Spinach and 0.5 μM DFHBI free in solution. As depicted in Figure 3, both cases exhibit a single exponential decay with fluorescence lifetime τ_f = 4.0 ± 0.1 ns, that is, regardless of the extent of photoconversion, and it is 1.5-times longer than that of GFP (τ_f = 2.7 ns). This suggests that the photoconverted species does not accumulate to a high degree to be detected based on the fluorescence lifetime. The concentration of DFHBI does not greatly affect the fluorescence lifetime of the Spinach-DFHBI complex (Figure S2).

[DFHBI]-dependent fluorescence recovery

These characteristic behaviors of the Spinach-DFHBI complex may be attributed to (i) reversible photoconversion as previously proposed,²⁰ (ii) fluorogen exchange,¹¹ or (iii) a combination of both. To further clarify the mechanism, we next determined the fluorescence recovery rate as a function of fluorogen concentration. If the Spinach-DFHBI complex is photoconverted to a nonfluorescent state in which the fluorogen is still bound to the RNA aptamer, thermal or light-induced recovery to a fluorescent state would be independent of [DFHBI]. Instead, the recovery rate (k_on) displays a strong dependence on [DFHBI]; k_on changes from 0.08 to 2.0 s⁻¹ as [DFHBI] increases from 1 to 200 μM (Figure 4), in disagreement with a previous proposal²⁰ where the reported recovery rate was constant at ~ 0.0027 s⁻¹ independent of DFHBI concentration. This discrepancy results from the fact that we used immobilized RNA in excess of DFHBI whereas in the latter case, fluorescence recovery measured the replenishment rate of the fluorescent Spinach-DFHBI complex in a freely diffusing solution (please see the SI for detailed discussion). The sub-linear behavior of the recovery rate in Figure 4b might be attributed to altered spectral properties of the Spinach-DFHBI complex at high concentrations of DFHBI although its molecular mechanism remains unclear (Figure S3).

Binding-unbinding kinetics

Next we examined how fast fluorogen exchange occurs in the Spinach-DFHBI complex. Mixing the Spinach RNA with DFHBI in a quartz cuvette, the fluorescence change was monitored with very low light level (< 1 – 5 W/cm²) to minimize complications arising from photoconversion. Figure 5a shows that the fluorescence signal rises mono-exponentially, giving the observed rate constant (k_obs). A linear fit to k_obs vs. concentration of Spinach and DFHBI directly provides the binding rate constant, k_bind (slope), and the unbinding rate constant, k_unbind (intercept) in Figure 5b. The obtained k_bind = 6.2 ± 0.1 × 10⁴ M⁻¹s⁻¹ and k_unbind = 2.4 ± 0.1 × 10⁻² s⁻¹ yield dissociation equilibrium constant, K_D = 390 nM which is in good agreement with previous studies.^11,20 Here, k_bind of Spinach-DFHBI is well within the range of typical small molecule-RNA aptamer interactions, which vary from 10³ M⁻¹s⁻¹ to 10⁶ M⁻¹s^−1.21

Notice that the dissociation time (1/k_unbind) is very long, that is, once Spinach binds to DFHBI, it stays as a complex for > 40 s under low light level or in the absence of light. Also, [DFHBI]-dependence of the recovery rate in low concentrations gives a bimolecular reaction rate, 4.9 ± 0.3 × 10⁴ M⁻¹s⁻¹ that corresponds well to k_bind (Figure 4b inset). This confirms that the fluorescence recovery is mainly due to a bimolecular reaction between Spinach and DFHBI rather than a spontaneous recovery of the photoconverted complex.

Fluorescence signal gain with low-repetition illumination and its application to E. coli imaging

The limited time duration of fluorescence of the Spinach-DFHBI can be mitigated by low-repetition illumination. Here the fluorescence intensity was monitored under continuous wave (CW) illumination or under a pulsed illumination (50 ms pulse width) with a repetition rate varying from 2 Hz to 0.2 Hz (Figure 6a). A similar scheme has been applied in fluorescence microscopy to attain high fluorescence yield from organic dyes and fluorescent proteins where the repetition rates were from 0.1 kHz to 100 kHz to allow dark state relaxation.²²

To demonstrate the concept, we transformed E. coli with plasmids expressing Spinach and imaged them in the presence of 100 μM DFHBI (Figure 6b).¹¹ Time-lapse fluorescence microscopy shows that the fluorescence signal of E. coli under CW illumination rapidly drops to < 10 % of the initial intensity within 0.2 s (Figure 6c), whereas pulsed illumination at 0.2 Hz repetition rate retains the fluorescence signal much better by ensuring sufficient recovery time between successive pulses (Figure 6d). This phenomenon holds true for immobilized Spinach RNA aptamers as well (Figure S4). We further quantified the data by plotting the fluorescence time trajectories of a single cell (Figure 6e) and 40 cells averaged (Figure 6f). The accumulated fluorescence intensity, representing the total photon flux, is 5.7-fold higher at 0.2 Hz than with CW illumination for the same 2 s of total illumination time (Figure 6g). The steady state intensity is about 10 fold higher at 0.2 Hz compared to CW illumination (Figure 6e,f). Although the lower repetition rate gives higher fluorescence signal gain (Figure 6h), there is a tradeoff in imaging speed. Different RNA aptamers selected for higher k_bind may allow even higher imaging speed. Additionally, pulsed illumination with longer pulse widths is not effective in gaining higher signal, and needs more recovery time probably because shorter pulses (10 – 50 ms) minimizes the photoconversion process (Figure 6i).

Photophysical model of the Spinach-DFHBI complex

Based on our results, we can rule out the possibilities that Spinach undergoes reversible photoconversion without fluorogen exchange [model (i)] or efficient fluorogen exchange without photoconversion [model (ii)]. A study of a malachite green RNA aptamer has shown that the fluorogen can go through significant changes in their conformation and electronic structure upon illumination. ²³ In the case of photoswitchable proteins, photoisomerization is known to be strongly involved with the formation of the dark state,²⁴ where an accompanying change in the configuration of fluorescent proteins leads to the conversion of a (fluorescent) deprotonated state to a (nonfluorescent) protonated state.²⁵ We confirmed that after 405 nm illumination the absorbance of free DFHBI decreases with its spectrum showing a red-shifted band and an enhanced high-energy band (Figure S5a). This is very similar to the spectroscopic features of GFP chromophore and its analogs indicating that cis/trans photoisomerization occurs for DFHBI.²⁶ For these reasons, it is tempting to suggest that the photoconversion of the Spinach-DFHBI complex may also be related to the cis/trans photoisomerization although a definite conclusion should await further studies.^20,27

To account for the unique photophysical behavior of Spinach we propose an alternative model as follows (Figure 7): Upon illumination of the Spinach-DFHBI complex, the bound DFHBI undergoes photoisomerization with a fast kinetic rate (k_off*). It is possible that the photoisomerized complex before unbinding is already completely dark. However, because of the lifetime measurements shown in Figure 3, we can rule out the possibility that a partially quenched photoisomerized complex accumulates significantly under our experimental conditions. Photoisomerization results in fast unbinding (k_unbind*) of the photoisomerized fluorogen, in contrast to the long-lived binding of DFHBI under very low or no illumination. Such accelerated unbinding could be caused by two factors: first, since the RNA aptamer was selected based on its binding affinity to the ground-state of DFHBI, its binding affinity to a photoisomerized DFHBI may not be optimal. Second, the Spinach aptamer might also change its conformation to accommodate photoisomerized DFHBI, a phenomenon observed in photoconverting fluorescent proteins, which could greatly decrease or even abolish binding. Rapid dissociation of photoisomerized DFHBI allows Spinach to binds to a new ground-state DFHBI from solution, restoring fluorescence. This could also explain why we could not identify the photoisomerized product via fluorescence lifetime (Figure 3) and fluorescence excitation spectrum (Figure S5b). It is also possible that light-induced fluorescence decay we observed here may be related to the recently reported RNA folding problem of Spinach.²⁸