Mechanism Behind the Apparent Large Stokes Shift in LSSmOrange

Nov 3, 2015 - LSSmOrange is a fluorescent protein with a large energy gap between the absorption and emission bands (5275 cm–1). The electronic stru...
2 downloads 19 Views 4MB Size
Download PDF
Article pubs.acs.org/JPCB

Mechanism Behind the Apparent Large Stokes Shift in LSSmOrange Investigated by Time-Resolved Spectroscopy Eduard Fron,† Herlinde De Keersmaecker,§ Susana Rocha,† Yannick Baeten,† Gang Lu,† Hiroshi Uji-i,† Mark Van der Auweraer,† Johan Hofkens,† and Hideaki Mizuno*,§ †

Molecular Imaging and Photonics, Department of Chemistry, KU Leuven, Celestijnenlaan 200F, 3001 Heverlee, Belgium Laboratory of Biomolecular Network Dynamics, Biochemistry, Molecular and Structural Biology Section, Department of Chemistry, KU Leuven, Celestijnenlaan 200G, bus 2403, 3001 Heverlee, Belgium

§

S Supporting Information *

ABSTRACT: LSSmOrange is a fluorescent protein with a large energy gap between the absorption and emission bands (5275 cm−1). The electronic structure of the LSSmOrange chromophore, 2-[(5-)-2-hydroxy-dihydrooxazole]-4-(p-hydroxybenzylidene)-5-imidazolinone, is affected by deprotonation of the p-hydroxybenzylidene group. We investigated LSSmOrange by time-resolved spectroscopy in the femtosecond and nanosecond range. The ground state chromophore was almost exclusively in the neutral form, which had a main absorption band at 437 nm with a small shoulder at 475 nm. The absorption at a wavelength within the former band promoted the protein to the excited state where excited state proton transfer (ESPT) could lead to deprotonation in 0.8 ps. Following ESPT, the chromophore emitted fluorescence with a maximum at 573 nm and a decay time of 3500 ps. Although deprotonation by ESPT occurs, we unexpectedly found a slow accumulation of the anionic form in the ground state upon repeated high intensity excitation. This accumulation of the anionic form was accompanied by a shift of the absorption band to 553 nm without changing the emission band. MALDI-MS revealed that this shift is accompanied by decarboxylation of E222, which is interacting with the imidazolinone ring of the chromophore. We concluded that the photoinduced decarboxylation induced a conformational change that affected local environment around the hydroxyl group, resulting in a stable deprotonated form of the chromophore.

1. INTRODUCTION On the basis of their large gap between absorption and emission peaks, a peculiar series of fluorescent proteins has been developed and used as noninvasive labels for multicolor observation of proteins in living cells.1−4 The well separated absorption and emission spectra (>100 nm) offers rational and great advantages in multicolor fluorescence microscopy as a single laser wavelength can simultaneously excite multiple fluorescent proteins by either one- or two-photon process.2,5 This property has already been found in wild-type Aequorea victoria green fluorescent protein (avGFP), which upon excitation in its major absorption peak at 395 nm yields emission peaking at 508 nm.6 Isoleucine substitution for T203 reduces a minor absorption of avGFP at 475 nm, and for this type of proteins the term “LSS, large Stokes shift” fluorescent proteins has been introduced. An Anthozoa fluorescent protein that possesses this property, Keima, has been engineered from a nonfluorescent chromoprotein isolated from a stony coral and its extensive applicability in dual-color fluorescence imaging such as fluorescence cross-correlation spectroscopy has been reported.4 The term “LSS” has propagated in the field of biology among the newly designed Anthozoa fluorescence © 2015 American Chemical Society

proteins obtained via random and rational mutagenesis with emission in the 550−650 nm region of the visible spectrum; LSSmKate1 and LSSmKate2 with excitation/emission maxima at 463/624 and 460/605 nm, respectively, have been developed for multicolor intravital imaging of tumor cell migration.2,3 An outstanding bright fluorescent protein (quantum yield of 0.45) named LSSmOrange (excitation/emission maxima at 437/572 nm) has also been developed.7 LSSmOrange allows, among others, several multicolor applications using a single excitation wavelength while in combination with mKate2 allowed observing biochemical activities in a living cell via intracellular imaging. Members of this class were further used as FRET donors to develop new biosensors with advanced performances in observing apoptotic activity and Ca2+ fluctuations compared to the commonly used CFP-YFP pair.7 In spite of the evident applicability of Anthozoa LSS fluorescent proteins, information related to the mechanism at the basis of such unusual absorption and emission characterReceived: September 21, 2015 Revised: October 31, 2015 Published: November 3, 2015 14880

DOI: 10.1021/acs.jpcb.5b09189 J. Phys. Chem. B 2015, 119, 14880−14891

Article

The Journal of Physical Chemistry B

2.4. Stationary UV−Vis Absorption and Fluorescence Experiments. The stationary absorption and fluorescence spectra were recorded with respectively a spectrophotometer (Lambda 40, PerkinElmer) and a fluorimeter (SPEX Flurolog3, Horiba Jobin Yvon) corrected for the wavelength dependence of the detection system.16 The optical density at the absorption maximum of all samples was kept below 0.1 in a 1 cm cuvette. 2.5. Picosecond Fluorescence Time-Correlated Single Photon Counting (TCSPC) Experiments. The fluorescence decay times were determined by TCSPC using the setup that has been described in detail previously.17 A time-correlated Single Photon Timing PC module (SPC 830, Becker & Hickl) was used to obtain the fluorescence decay histogram in 4096 channels. The decays were recorded with 10 000 counts in the peak channel, in time windows of 15 ns corresponding to 3.7 ps/channel and analyzed globally as a sum of exponentials,

istics is limited. It has been reported that the large energy gap of Keima is due to excited state proton transfer (ESPT).8,9 Also for LSSmOrange, the contribution of ESPT is suggested from the isotope effect observed on stationary emission spectra at 77 K.7 Since the emission occurs from a species (presumably anionic) that has an electronic state differing from the one reached after absorption of the photon, the nomenclature used for this class of proteins “large Stokes shift” intrigued readers from the field of photochemistry and photophysics, as the term “Stokes shift” should only be used for absorption and emission transitions involving the same electronic states.10 Moreover, other structural and conformational processes, which lead to dramatic changes in protein photophysics and limit or even distort the fluorescence-based experiments, can accompany ESPT. Since applications of fluorescent proteins greatly rely on their photophysical properties, a deep understanding of the mechanism is not only a prerequisite of fundamental interest from a photochemical and photophysical point of view but also for a successful application of these proteins as fluorescent probes, especially considering their use in super-resolution microscopy applications.11−13 In this study, we explored the excited state dynamics of this class of proteins by investigating LSSmOrange. Due to the short distances between the relevant protein constituents, many of the intramolecular processes that are important in the excited state dynamics are expected to be governed by ultrafast processes and can therefore only be probed by ultrafast spectroscopy. Employing femtosecond fluorescence up-conversion, transient absorption with a resolution of 100 fs, and time-correlated single photon counting (TCSPC), a systematic investigation on the optical properties and excited state dynamics in solution upon excitation with different wavelengths was performed to unravel the possible relaxation pathways following excitation of LSSmOrange. Through investigation of the excited state dynamics, we unexpectedly observed a slow photoconvertible property of LSSmOrange, characterized by a change in the absorption peak from 437 to 553 nm. Here we also report the molecular basis behind this photoconversion.

I (t ) =

⎛ t⎞ ⎟ ⎝ τi ⎠

∑ Ai exp⎜− i

with time-resolved fluorescence analysis (TRFA) software.18,19 The full width at half-maximum (fwhm) of the instrument response function (IRF) was typically in the order of 40 ps. The quality of the fits was judged by the fit parameters χ2 (600 ps was regarded as being identical to the 3500 ps component observed by TCSPC, which was the component with the largest amplitude in the 550−650 nm region and was attributed to emission of the anionic chromophore. The 49 and 3.8 ps (the latter cannot be retrieved in the TCSPC experiments) components corresponded to two different species responsible for the low 14884

DOI: 10.1021/acs.jpcb.5b09189 J. Phys. Chem. B 2015, 119, 14880−14891

Article

The Journal of Physical Chemistry B

Figure 5. Decay traces and corresponding fits obtained by femtosecond fluorescence up-conversion for unconverted LSS-mOrange in a normal (A) and a deuterated (B) buffer in a 50 ps time window, (λexc = 400 nm, λdet = 570 nm, pH 7.4). The analysis of the decays as sum of four exponentially decaying components allowed to recover following decay times: A in a normal buffer, 0.8, 3.8, 49 ps (fixed) and >600 ps (fixed); B in a deuterated buffer, 2.87, 3.8 ps (fixed), 49 ps (fixed), and >600 ps (fixed).

Figure 6. 2D (A, C) and 3D (B, D) femtosecond transient absorption spectra obtained for unconverted LSSmOrange in a 50 ps time window (A, B) and 6 ps time window (C, D); λexc = 400 nm, λdet = 480−640 nm, pH 7.4.

involvement of ESPT process. To confirm the involvement of ESPT process, we measured the isotope effect on the decay time of this component (Figure 5). The increase of the decay time from 0.8 ps in an aqueous buffer to 2.8 ps in a deuterated buffer is apparently in line with this hypothesis. For a better understanding of the dynamics, transient absorption experiments were performed in a 6 ps and a 50 ps time window following the excitation at 400 nm (Figure 6). The results obtained for the 50 ps time window revealed the presence of a strong stimulated emission band (negative signal) ranging from 540 to 640 nm with a maximum at 573 nm accompanied by a positive signal between 480 and 540 nm. The negative signal can be attributed to induced emission from the species decaying with a 3500 ps time constant observed by TCSPC/up-conversion. Since the time dependence of the positive signal was similar to that of the negative signal, the

intensity component emitting around 506 nm in the steadystate experiment. The 0.8 ps component (which was not possible to be observed in the TCSPC experiments) had an intriguing wavelength dependence; this 0.8 ps component had a large positive amplitude in the 510−540 nm region and its sign changed in the 540−590 nm region, suggesting that the excited state of the anionic chromophore is formed from the state emitting at 510−530 nm with a time constant of 0.8 ps. The spectral features of the 0.8 ps component is compatible with blue-green shoulder observed at the low wavelength side of the spectra shown in Figure 3B. As this spectrum is a mirror image of the short wavelength absorption band shown in Figure 2A, it is possibly originates from the neutral form. Hence the wavelength dependence of the amplitude of the 0.8 ps component, suggests that the neutral form is a precursor of the excited state of the anionic form, which indicates the 14885

DOI: 10.1021/acs.jpcb.5b09189 J. Phys. Chem. B 2015, 119, 14880−14891

Article

The Journal of Physical Chemistry B positive signal can be ascribed to the excited state absorption (S1−Sn) of the same species. When the same experiments were performed in a 6 ps time window, a clear progression of the stimulated emission signal was observed. The features of the spectra indicated that the excited state anionic form was being populated within a time interval of about 5 ps. No ground state depletion corresponding to the induced emission of the anionic form, which would produce a negative signal between 480 and 560 nm, was observed. This suggests that the formation of the yellow-emitting species occurs via the excited state (Supporting Information Figure 2). Furthermore, one can observe in Figure 6C that the maximum of the negative signal shifts from 550 nm at 1.49 ps to 565 nm at 6 ps. This shift could be related to further excited state relaxation of the anionic form following the ESPT. The signal attributed to the excited state absorption does not change after 1.49 ps, hence the time constant for the rise of this signal must be 1 ps or less. This suggests that this signal is due to both the precursor and the final excited species. It can also not be excluded that this shift is due to the disappearance or blue shift of S1 → Sn absorption of the intermediate species. 3.4. Photoconvertible Property of LSSmOrange. During experiments in which time-resolved spectra of LSSmOrange were carried out by irradiating with 400 nm fs pulses, we visually noticed a gradual change in color of the sample from yellow to pink (Figure 1B). While checking the integrity of the solution by recording the stationary absorption spectrum, we observed that the band centered at 437 nm was reduced and a new absorption band with maximum at 553 nm appeared (Figure 1A). We refer to LSSmOrange after the laser irradiation as photoconverted LSSmOrange in this study. At pH 7.4, the emission spectrum of the photoconverted LSSmOrange exited at 550 nm was identical to that of unconverted LSSmOrange excited at 440−480 nm (Figure 3B and D). In acidic buffers (pH ≤ 5), the absorption band at 553 nm was no longer observed while an absorption band at 437 nm showed up (Figure 2), indicating an anionic character of this species. From the pH dependence of the peak at 553 nm an apparent pKa value 6.6 was estimated, which is much lower than that of the unconverted LSSmOrange (>11). Since the 553 nm peak also appeared for unconverted LSSmOrange in basic buffers (pH ≥ 10), the photoconversion is also associated with deprotonation of the chromophore. In order to explain the change in pKa, this deprotonation should be accompanied by a photoinduced modification in the acidity of the hydroxyl group of the chromophore. This would suggest a photoinduced change of the environment of this chromophore. To investigate the structural basis of the photoconversion, LSSmOrange was trypsinized and analyzed by MALDI-MS (Figure 7 and Supporting Information Figure 3). Tryptic peptides covering 77.5% and 76.3% of the amino acid sequences were observed for unconverted and photoconverted LSSmOrange, respectively. Most of the fragments were consistent with the theoretical mass number, and found to be identical for photoconverted and unconverted LSSmOrange. However, for the photoconverted LSSmOrange, a tryptic peptide corresponding to L205-R223 (numbering follows that for eGFP, note that the amino acid corresponding to 216 is not existing in LSSmOrange) was detected at m/z = 2180.9180, which was 44 Da smaller than the theoretical value (m/z = 2225.0357). The counterpart of the unconverted LSSmOrange (m/z = 2224.8645) was consistent with the theoretical value. The analysis of MS/MS fragments revealed that the 44-Da decrement happened at E222, suggesting decarboxylation of the

Figure 7. MALDI-MS spectra of LSSmOrange. (A) Amino acid sequence of LSSmOrange. Amino acid numbers are indicated on both sides. Note that the numbering follows that for eGFP. 56* means the amino acid residue following to 56 but missing in eGFP. Arrowheads indicate trypsin cleavage sites. The chromophore is shown in green. The fragment containing E222 is in cyan with E222 in bold and indicated with an asterisk. Green and magenta lines indicate the tryptic fragments of unconverted and photoconverted LSSmOrange, respectively, detected by MALDI-MS. (B) Mass spectra of unconverted (green) and photoconverted (magenta) tryptic peptides containing E222. (C) MS/MS fragment ions of the tryptic peptides. The m/z values of C-terminal fragment ions (y1+, y2+) and N-terminal fragment ions (b16+, b17+) for unconverted (green) and photoconverted (magenta) LSSmOrange are shown.

glutamate residue. This is essentially the same mechanism as suggested for the critical step in the phototransformation of avGFP and PA-GFP, where also a deprotonation of the chromophore resulting in the red shift of the absorption spectrum was proposed.25 From these results, we concluded that LSSmOrange possesses a photoconvertible property shifting the absorption maximum from 437 to 553 nm upon irradiation with a 400 nm laser. This shift is caused by the 14886

DOI: 10.1021/acs.jpcb.5b09189 J. Phys. Chem. B 2015, 119, 14880−14891

Article

The Journal of Physical Chemistry B

Figure 8. Depicted pathways for the excited state processes suggested to occur in unconverted LSSmOrange. The position of the S0-levels is chosen for the clarity of the scheme and does not reflect their energetic content.

fraction of photoconverted proteins (signal 480−640 nm). The absorption spectrum of the photoconverted fraction (Figure 1A) appeared as a mirror image of the emission spectrum for either the unconverted or the photoconverted form (Figure 3B, D) with a main peak at 553 nm and a vibronic shoulder at 520 nm. Excitation spectra (Figure 3C) showed that only this band yielded the fluorescence at 573 or 650 nm. When exciting the sample with 480−550 nm light the emission spectra were similar to the ones obtained for the unconverted form. Considering these facts, we suggest that the bright species emitting at 573 nm and corresponding to the anionic form of the chromophore is characterized by a ground state absorption with the maximum at 553 nm and thus has a Stokes shift of 631 cm−1, typical of chromophores related to GFP.23,24 The TCSPC experiments revealed that on a nanosecond time scale the fluorescence decays of the converted LSSmOrange recorded upon 550 nm excitation could be fitted to a biexponential decay with decay times of 660 and 3380 ps, which are independent of the detection wavelength (Table 1). These two decay components can be related to the yellowemitting species. The decays were dominated by the 3380 ps component, which is equivalent to the 3500 ps component observed for the anionic form of the unconverted protein as well as for the anionic form of various other fluorescent proteins.23,24 The importance of this finding resides in the fact that both emitting species, which resulted from excitation of the unconverted form in the violet and from exciting of the

change in acidity of the chromophore hydroxyl group by decarboxylation of E222 resulting in chromophore deprotonation at neutral pH. In an attempt to observe structural changes accompanying the photoconversion, we recorded stationary IR absorption spectra of the protein before and after photoconversion (Supporting Information Figure 4). The spectra in the 1000− 1800 cm−1 region revealed a clear difference between the unconverted and photoconverted forms. Two peaks (1542.0 and 1559.8 cm−1) noticed for the unconverted form of the protein were no longer found in the photoconverted sample. Assuming that this vibrational mode is characteristic of one of the CN stretching within the chromophore, this could indicate a change in protonation of the imidazolinone nitrogen, in addition to that of the phenolic oxygen. The region 1700− 1800 cm−1 is characteristic to the CO stretch vibrations, and the peak at 1724.9 cm−1 appearing after photoconversion could be related to a change in the hydrogen bond network linking to the D165 side chain, which includes a carboxylate group before the photoconversion; upon photoconversion it receives a proton and is converted to carboxylic acid which can absorb at 1724.9 cm−1 (Figure 1C, D). 3.5. Spectra of Photoconverted LSSmOrange. The steady-state absorption spectrum of the photoconverted LSSmOrange consists of two bands (Figure 1A). This spectrum can essentially be considered as a superposition of a fraction of residual photodamaged proteins (signal 350−400 nm) and a 14887

DOI: 10.1021/acs.jpcb.5b09189 J. Phys. Chem. B 2015, 119, 14880−14891

Article

The Journal of Physical Chemistry B

formation of a species whose excited state decays nonradiatively. Fischer et al. modeled the proton translocation in avGFP as a series of activated hopping processes with localized proton steps between neighboring sites along the hydrogenbond network extending from the imidazolinone moiety to E222.26 They suggest that ESPT favors the 3-step model with E222 as final acceptor. Based on these arguments, we cannot exclude that other nonequilibrium protein states are involved in the ESPT reaction of LSSmOrange. Actually, the fact that the rise of the induced emission of the anionic state, which also shifts as a function of time to longer wavelengths, is slower (around 5 ps) than the initial decay of the exited protonated state (0.8 ps) is in agreement with such a multistep model. As de-excitation to the ground state may occur at any state in the ESPT reaction, such a state could escape our observation. The excited state of the anionic form (I2) decays to the ground state with a time constant of 3500 ps. Assuming that is formed with a 100% quantum yield we can calculate krad and knr (krad = 1.28 × 108 s−1, knr = 1.85 × 108 s−1). As a result, the emission is preceded by a transition that occurs in electronic excited states, and hence “Stokes shift” is not an appropriate term to describe the energy gap between the absorbance and the emission of LSSmOrange. Taking the fact that the abbreviation “LSS” has widely been used to describe this series of fluorescent proteins into account, we propose to use this term as an acronym of “Light-induced Spectral Shift” instead of “Large Stokes Shift”. Since the ESPT relies on the increase of the chromophore acidity in the excited state and considering the fact that the acidity is lowered again upon the transition to the ground state, the formation of the deprotonated chromophore is transient and the neutral form of the chromophore is regenerated in the ground state. The hydroxyl group of p-hydroxybenzylidene is facing to the hydroxyl group of S148 that links the hydrogen network to the deprotonated carboxyl group of D165. These side chains are thought to play a crucial role for efficient deprotonation and reprotonation by stabilizing the neutral form of the chromophore in the ground state and serving as the proton acceptor in the excited state. Considering the transient absorption spectra in the region 540−640 nm, where the observed signal is exclusively due to stimulated emission, practical applications can be envisioned for super-resolution microscopic methods. In this study, we revealed the photoconvertible property of LSSmOrange upon exposure to light at 400 nm. The absorption band of LSSmOrange shifted from 437 to 553 nm whereas the emission band stayed the same. This mode of irreversible photoconversion had not been reported with Anthozoa fluorescent proteins until recently. Besides the Anthozoa fluorescent protein, avGFP (Aequorea fluorescent protein) and its mutant PA-GFP show this mode of photoconversion. Although PA-GFP is widely used as a photoactivatable fluorescent protein upon 400 nm illumination, PA-GFP is actually already fluorescent before photoactivation. While the absorption band shifts from 400 to 488 nm during the photoactivation process the emission maximum remains at 511 nm.27,28 The photoconversion of avGFP consists of decarboxylation at E222 by a Kolbe-type reaction followed by deprotonation of the hydroxyl group of the chromophore through a hydrogen bond network.25 The decarboxylation at E222 is also observed as a common photoactivation process in Anthozoa photoactivatable fluorescent proteins such as PAmCherry.29 In the case of PAmCherry, however, the hydrogen network of the chromophore hydroxyl group is not

photoconverted form at 550 nm, possess similar photophysical properties. The minor component of 666 ps (