Pixel Team, IBS, Institut de Biologie Structurale Jean-Pierre Ebel, CEA, CNRS, Université Joseph Fourier, 41 rue Jules Horowitz, Grenoble, France
Institut de Recherches en Technologies et Sciences pour le Vivant, iRTSV, Laboratoire de Physiologie Cellulaire et Végétale, CNRS/CEA/INRA/UJF, Grenoble, France
Pixel Team, IBS, Institut de Biologie Structurale Jean-Pierre Ebel, CEA, CNRS, Université Joseph Fourier, 41 rue Jules Horowitz, 38027 Grenoble, France and Institut de Recherches en Technologies et Sciences pour le Vivant, iRTSV, Laboratoire de Physiologie Cellulaire et Végétale, CNRS/CEA/INRA/UJF, Grenoble 38054, France
Fluorescent proteins (FPs) have become indispensable tools to investigate the synergy between cell structure, function, and dynamics. FPs fold as 11-stranded β−barrels enclosing an endogenous 4-(p-hydroxybenzylidene)-5-imidazolinone (p-HBI) chromophore that only needs oxygen as an external cofactor to mature (1). The discovery of phototransformable FPs (PTFPs) from Anthozoa species, the fluorescence properties of which can be modified by actinic light, triggered a revolution in the field, notably by enabling single-molecule localization-based super-resolution microscopy (2, 3). Three types of phototransformations are distinguished: photoactivation and photoconversion are nonreversible processes that result from covalent modifications of the FP matrix, whereas photoswitching refers to the reversible transition between a fluorescent on-state and a nonfluorescent off-state. Excellent reviews cover PTFPs in general, as well as their applications in the biological microscopy field (4, 5). Photoswitching in organic dyes in the context of nanoscopy has also been recently reviewed (6). Here, we focus on reversibly photoswitchable FPs (RSFPs), highlighting mechanistic aspects. We also touch on blinking, the transient and stochastic loss of fluorescence observable in essentially all fluorophores.
Reversible Photoswitching Over the Years
Reversible photoswitching in FPs has first been observed on yellow derivatives of Aequora victoria GFP (YFP) at the single-molecule level (7). Single immobilized YFP proteins excited by light of alternate cyan (488 nm) and violet (405 nm) colors underwent repeated cycles of off/on switching on a timescale of several seconds. However, this behavior was not observable in bulk studies. Reversible photoswitching at the ensemble level was demonstrated at ultra-low temperature on GFP (8), and some red-shifted derivatives (9, 10) using spectral hole burning experiments. Switching appeared to involve only a subpopulation of proteins, and no data was presented on the achievable on/off fluorescence contrast. Bulk photoswitching at room temperature was then evidenced in GFP derivatives, notably CFP (11) and several yellow variants such as EYFP, Citrine (11), E2GFP (12), and YFP-10C (13). In all cases, the achieved photoswitching contrast remained limited, probably stemming from the coexistence of noninterchangeable populations of molecules over experimental timescales, only some of them being photoswitchable. Reversible light-induced cis-trans isomerization of the chromophore and/or chromophore protonation was hypothesized as the most likely reasons for the observed switching.
The discovery of FPs from Anthozoa species such as corals or anemones triggered a considerable boost in FP research. Efficient photoswitching was initially described as reversible “kindling” in the weekly fluorescent and tetrameric asFP595 from the sea anemone Anemonia sulcata. The first reversible optical highlighter adequate for biological experiments was obtained upon engineering a Pectiniidae coral FP, yielding the well-known RSFP Dronpa (14). The creation of Dronpa was followed by the introduction of other RSFPs exhibiting improved properties such as better fluorescence brightness, enhanced switching contrast, tunable switching quantum yields, increased fatigue-resistance, or red-shifted emission color (Table 1). Mechanistic investigations of some of these RSFPs, generally based on a combination of crystallographic, spectroscopic, and molecular dynamics approaches (15–21) suggested photoswitching mechanisms similar to those hypothesized for Hydrozoan FPs, based on coupled cis-trans isomerization and protonation changes of the chromophore.
Table 1. Properties of well-characterized RSFPs developed to date
Recently, new Hydrozoan GFP-based RSFPs were introduced, confirming that high-performance photoswitching is not restricted to Anthozoan FPs. Single-mutation of the strictly conserved Glu212 (GFP numbering) into Gln conferred high-contrast switching properties to variants of YFP (22), while a combined rational and random mutagenesis approach yielded the fatigue-resistant rsEGFP (23). Despite insightful studies by vibrational spectroscopy (24), it should be noted that crystallographic evidence for a trans chromophore in Hydrozoan GFP remains yet to be established. In 2011, Brakemann et al. (25) very elegantly took advantage of the complex photoswitching behavior previously noticed on YFP-derivatives (13, 26) to elaborate Dreiklang, an RSFP exhibiting decoupled fluorescence excitation and switching, based on an entirely new mechanism.
Using RSFPs in Fluorescence Microscopy
Reversibly photoswitchable FPs offer a vast panel of potential applications, including super-resolution microscopy, photochromic FRET (22, 27), optical lock-in detection (28), and biotechnological developments such as rewritable high-density optical storage media (29, 23).
Super-resolution microscopy techniques based on point-scanning, saturated structured illumination or single molecule-based wide-field approaches take advantage of a nonlinear response of fluorescence emission upon excitation. As photoswitching constitutes such a response, it is in principle suitable for all these nanoscopy schemes. The first applications used Dronpa in PALM nanoscopy (30, 31). Dronpa-like RSFPs emitting in the green allow combination with for example a photoconvertible PTFP such as EosFP, emitting in the orange/red, to achieve two-color imaging with minimal crosstalk. Another interesting application was based on the use of the multitransformable PTFP mIrisFP to achieve pulse-chase nanoscopy with reversible switching in both green and red channels (32). Employing reversible switching in single-molecule localization based nanoscopy presents advantages and drawbacks as compared to photoconversion or photoactivation: multiple localizations can enhance sampling density for a moving probe, enabling a more continuous visualization of the imaged structures. However, multiple switching by a single probe complicates a quantitative evaluation of, for example, target copy number. In fact, spurious switching of some photoconvertible PTFPs in their converted red state has been noticed (33), which may induce clustering artifacts in PALM (34, 35). Moreover, substantial reversible switching in normally nonswitchable PTFPs can be induced by reducing agents such as MEA (36). Switching can turn into a nuisance and some protein engineering research aimed at avoiding it (37).
RSFPs can also be used advantageously in point-scanning nanoscopy schemes, in a technique called reversible saturable optical linear fluorescence transitions (RESOLFT) (38, 39, 23). The RESOLFT approach allows a considerable reduction of the light intensity needed to break the diffraction barrier as compared to STED. Similarly, RSFPs permit to achieve saturated structured illumination microscopy (SSIM) with low-light level, as recently demonstrated (40). Contrary to PALM-like approaches, RESOLFT and SSIM achieve resolution enhancement through multiple switching cycles of the marker and therefore cannot benefit from nonreversible PTFPs. However, to attain their best performance, these techniques require the development of fatigue-resistant RSFPs such as rsEGFP (23) or NijiFP (41).
Depending on the application foreseen, switching properties considered as “optimal” differ. In PALM microscopy, a relatively low on–off switching quantum yield favors accurate probe localization, as a large number of photons can be collected before switching (42). In contrast, SSIM, RESOLFT, OLID, or pcFRET microscopy benefit from a high switching quantum yield allowing a large number of switching events to be collected rapidly at low light level. In principle, the possibility to decouple fluorescence excitation from switching, as recently achieved in Dreiklang (25), constitutes an excellent solution for a single RSFP to fit varying requirements.
The p-HBI chromophore does not fluoresce in solution because, upon photon absorption, twisting around the methylene bridge linking the two cyclic moieties rapidly leads to deactivation by conical intersection and potential cis-trans isomerization (43). Thus, photoswitching by chromophore isomerization in RSFPs is a manifestation of intrinsic excited state chromophore dynamics. This is in contrast to photoactivation and photoconversion processes, which result from specific photochemical reactions of the chromophore promoted by a suitably arranged protein matrix. Chromophore isomerization in RSFPs has been found to result from a single-photon excitation process in both directions (44, 21). However, back-switching is also thermally driven and some FPs are able to change the isomeric state of their chromophore in the ground state following pH changes (45, 46). Upon intense illumination, isomerization has been invoked in some red-emitting FPs to account for band shifts (47, 48) but in this case the sequential absorption of two photons seemed to be involved and the occurrence of isomerization was deduced from IR or Raman vibrational spectroscopy and would need to be confirmed structurally.
Whenever cis-trans isomerization has been invoked to account for the switching mechanism of RSFPs, it has been found that the nonfluorescent state corresponds to the trans configuration of the chromophore. Nevertheless this isomeric state per se does not imply a lack of fluorescence, as other FPs exhibit strong fluorescence in the trans state (46, 49). Finding out the reasons for weak or strong fluorescence in one or the other isomeric state of the chromophore in RSFPs is therefore not obvious. Chromophore planarity, rigidity, protonation state and potential ESPT pathways all matter in a combined manner. Empirical proposals were made to quantify the ability of an anionic chromophore to fluoresce, depending on the tilt and twist angles between the cyclic moieties (20). Excited-state molecular dynamics can be a useful tool to monitor chromophore torsion upon photon-excitation, which directly relates to fluorescence quantum yield (15, 50, 21).
RSFPs can be classified as negative or positive (off-switching or on-switching results from illumination at wavelengths absorbed in the fluorescent on-state, respectively). The protein Dreiklang introduces a new category in which switching is neither negative nor positive but, as already mentioned, decoupled from fluorescence excitation.
Switching quantum yields in RSFPs typically cover the range 10−1 to 10−4. Interestingly, these yields tend to decrease with the transition energy required to engage the switch: In both positive and negative RSFPs, switching induced by cyan/green light display a moderate yield of the order of 10−3 to 10−4, whereas switching induced by violet/blue light is much more efficient with yields in the range 0.5 to 10−2. No rationale behind this observation has been given but this could be due to facilitated photoswitching in the protonated state of the chromophore, and/or to the larger amount of heat generated upon absorption of higher energy photons, which would promote destabilization of the chromophore environment and overcome steric hindrance to isomerization.
In the following, we describe in more details current knowledge about switching mechanisms of negative, positive and decoupled RSFPs.
Negative Switching in Anthozoan Proteins: Dronpa
The RSFP Dronpa (14) has been used successfully in several biologically relevant experiments because it exhibits bright fluorescence and is monomeric. Dronpa and various mutants including Dronpa-2, Dronpa-3 (51), rsFastLime (52), and bsDronpa (53) have been the subject of several mechanistic investigations at the single-molecule to the single-crystal level. Single-molecule (44) and fluorescence correlation spectroscopy (54) demonstrated that switching occurs on rapid timescales (ms to μs) and found that the light-induced off-state corresponds to a protonated state of the chromophore that differs from the pH-induced protonated state. ESPT to an intermediate state I was proposed to account for the high-yield back-switching reaction (44), which was confirmed by ultrafast absorption spectroscopy and kinetic isotope effect measurements using deuterated samples (55). After crystallographic structures of Dronpa in its on-state were published (56, 52), a structure of the off-state suggested that switching is due to cis-trans isomerization of the chromophore followed by protonation (16). Isomerization is accompanied by substantial conformational change of the chromophore pocket, and stability of the protein in both the on- and off-states is favored by a switch in the hydrogen bonding network around the chromophore (Fig. 1). The tightly H-bonded triad Glu144-His193-Glu211 in the cis configuration is replaced by the Glu144-Arg66-Glu211 triad in the trans configuration, with either His193 or Arg66 stabilizing the chromophore by π−stacking and π−cation interactions with the hydroxybenzylidene moiety, respectively. The same structural rearrangements are observed in mTFP0.7 (17) and IrisFP (18). In all three proteins, the trans state is also stabilized by the ability of Ser142, which maintains a strong H-bond with the hydroxybenzylidene moiety in the cis state, to find another H-bonding partner once the chromophore has been isomerized (18). Protonation of the chromophore in the trans state was then explained by the substantial change in electrostatic environment relative to the cis state (16). Whereas the poorly defined electron density around the chromophore argued for a strong disorder in the Dronpa trans state, in mTFP0.7 or IrisFP the protonated trans chromophore remains well-ordered, in a mostly hydrophobic or a polar environment, respectively.
Possible alterations of functional mechanisms in the crystalline state are always an issue (57). By employing NMR, Mizuno et al. (19) proposed, in contrast to the crystallographic results described above, that on–off switching in Dronpa follows from disordering of the chromophore, but not necessarily from cis-trans isomerization. The study pointed out the importance of the protein oligomerization state (found in the crystal) in possibly modifying the structural flexibility necessary for switching. Chromophore disordering would be initiated by light-induced protonation in the cis triplet state, after intersystem crossing. This hypothesis seems consistent with spectroscopic studies carried out at cryo-temperature which, by blocking the conformational rearrangements necessary for isomerization, strongly suggested that photo-induced protonation without isomerization are feasible processes (26).
On the basis of both X-ray and NMR studies, our view is that Dronpa in its off-state resides in a distorted and protonated trans configuration of the chromophore. Its loose attachment and/or severe distortion, weakening the π-conjugated electron system, would account for the absence of fluorescence in this state. ESPT in a twisted geometry may then be the driving force for off–on switching, as previously suggested (44, 55). This is also consistent with further theoretical calculations proposing that isomerization and deprotonation events during Dronpa off–on switching are concerted (58). The hypothesis of a photo-induced protonation gating on–off switching, in contrast, is surprising in view of the fact that the primary response of a free chromophore to photon absorption is to rotate around its methylene bridge. This observation, together with the clear structures of other RSFPs in the trans configuration (59, 18) supports the hypothesis that chromophore isomerization plays a leading role in Dronpa photoswitching. On–off photoswitching may follow a concerted isomerization/protonation mechanism in all negatively switchable RSFPs. Such coupling may arise from the fact that drastic changes in chromophore pKa occur during twisting in the excited state (22), in line with ab initio calculations (60).
Despite these comprehensive studies, no intermediate state along the switching pathways of Dronpa or other negative RSFPs could be structurally identified, and the precise isomerization pattern (one bond flip, or the more space conserving Hula twist) remains unknown to date.
Positive Switching in Anthozoan Proteins: asFP595 and Padron
The protein asFP595, the first protein found to exhibit reversible switching (61), is an obligate tetramer, which limits its practical use for imaging applications. Yet, this positive RSFP has also been subjected to a number of mechanistic investigations. First insights based on the spectroscopic analysis of asFP595 mutants hinted at a cis-trans isomerization of the chromophore (62). A remarkable study of the asFP595-A143S mutant (displaying a greatly enhanced fluorescence quantum yield) by combined crystallography, spectroscopy and molecular dynamics suggested that on-switching was linked to chromophore trans-cis isomerization via a Hula-Twist mechanism. The chain break immediately preceding the chromophore (63) was proposed to play an essential role in switching, by providing sufficient flexibility to the imidazolinone moiety (15). Contrary to negative RSFPs, there were only very limited rearrangements of the chromophore pocket between the trans and the cis configurations, consistent with the observation of kindling at 150 K, below the glass transition temperature (64). How the chromophore protonation state controls photoswitching in asFP595 was investigated by ab initio calculations and QC/MM molecular dynamics simulations (65, 66). It was suggested that trans-cis isomerization occurs in the neutral state of the chromophore, followed by a dark state equilibration to a zwitterionic fluorescent cis state (66). However, this study did not rule out the possibility that the chromophore may also photoisomerize in its anionic state. This last possibility was experimentally observed in Padron.
Padron is another positive RSFP, engineered from Dronpa (53). The essential M159Y mutation was sufficient to completely reverse the switching properties. After initial structural work revealing an isomerization mechanism similar to that of asFP595 (20), further investigations (21) allowed to cryo-trap two fluorescent intermediates along the Padron on-switching pathway (Fig. 1). The combined spectroscopic and crystallographic data suggested that, in Padron, trans-cis isomerization of the chromophore occurs entirely in the anionic state and precedes protonation. The results were intriguing in terms of protein dynamics: they showed fluorescence activation at 100 K by full trans-cis isomerization, a very substantial structural change at a temperature at which protein dynamical breathing is essentially stalled. Molecular dynamics simulations of Padron in the S1 excited trans state showed that, upon photon absorption, chromophore twisting occurred in ∼10 ps, followed by putative breaking of some hydrogen bonds to complete isomerization. As heat dissipation upon photon absorption also occurs on the picosecond timescale, a transiently “hot” chromophore might account for isomerization at cryo-temperature. Conversely, it was observed that protonation of the chromophore following isomerization only took place above the glass transition temperature (∼200 K), suggesting that protonation involves exchange with the surrounding solvent.
Overall, it appears that the mechanisms governing fluorescence switching in negative and positive RSFPs differ. In negative RSFPs, chromophore isomerization seems to be concomitant with protonation and couples to a large structural response of the protein matrix. In positive RSFPs, isomerization and protonation seem to be sequential and the protein matrix barely moves. Further studies will be needed to refine these assertions. Interestingly, the available data suggest that positive RSFPs can be photoswitched at cryo-temperature, opening interesting potential applications such as cryo-nanoscopy (21).
Decoupled Switching in Hydrozoan Proteins: Dreiklang
Positive and negative RSFPs are not optimal in that fluorescence excitation interferes with switching. In PALM, this complicates control of sparse stochastic activation in positive RSFPs and limits achievable photon counts in negative RSFPs. The protein Dreiklang, evolved from the GFP variant Citrine, nicely overcomes this problem and extends our vision of RSFPs (25). In Dreiklang, excitation at 514 nm does not induce significant off-switching, whereas near UV-light at 365 and 405 nm result in on and off switching, respectively. The existence of such decoupling had already been hinted in previous experiments on EYFP (13, 26). Although only a small fraction of the molecules responded, it was noted that UV-light at around 350 nm was able to partially reactivate bleached fluorescence whereas violet light at 405 mm rather tend to further deplete fluorescence. In Dreiklang, this behavior was deliberately exacerbated by random mutagenesis, resulting in an unprecedented switching mechanism. Combined crystallographic and mass spectroscopy analyzes provided evidence that a light-induced hydration-dehydration of the chromophore caused switching by disrupting the π-conjugated electron system (Fig. 1). Hydration of the C65 atom of the imidazolinone moiety appeared to be facilitated by the proper positioning of a water molecule held in place by hydrogen bonding between Y203 and E222. However, this water molecule is present at the very same position in Citrine and the exact roles of the 4 mutations (V61L, F64I, Y145H, and N146D) remain to be established. Interestingly, the off-state hydrated structure of Dreiklang closely resembles that of late intermediates in the chromophore maturation pathway of YFP variants (67). Although absorption bands responsible for on- and off-switching are relativity close to each other, suggesting the possibility of crosstalk, an excellent switching contrast (>50) could be achieved. Nevertheless, the use of UV light might be cytotoxic. A question yet to be answered is how the hydration/dehydration processes can be light-activated in the protonated state of the chromophore.
Blinking can be defined as a particular switching process in which the off-state is not a singlet ground state, but rather an excited (triplet) or a ground radical (doublet) state. Return to the on-state is thermally induced and, contrary to switching, on/off bi-stability cannot be easily achieved based on blinking. Blinking processes in organic dyes have been extensively studied and can be controlled to some extent by, for example, properly choosing the solvent composition (oxygen, triplet state quenchers, redox agents) (68). In this manner, it became possible to take advantage of blinking to develop dSTORM stochastic microscopy (69).
In FPs, blinking is still poorly understood and generally considered as a nuisance. The triplet state as well as chromophore protonation changes have been invoked (70). Recent QM/MM calculations suggested that protonation of the Cα atom of the chromophore methylene bridge, either in the triplet state or in a radical state, could cause blinking (71). Cα protonation results in a transient break of π−conjugation between the two cyclic moieties of the chromophore and is accompanied by a severe distortion of the latter. The distortion was visualized in IrisFP upon exposure to X-rays at cryo-temperature (72) (Fig. 2), using a combination of crystallography and in crystallo optical spectroscopy. It can be postulated that the strongly reducing electrons and oxidizing holes generated by X-ray-induced solvent radiolysis created a chromophore radical species favoring proton transfer from the nearby arginine residue Arg66 to the Cα atom. This work highlights the general susceptibility of FPs to electron transfer reactions, in line with the proposal that FPs may functionally serve as photoactivated electron donors (73). However, the generality of this blinking mechanism remains to be investigated as well as its possible interplay with cis-trans photoisomerization.
CONCLUSION: DESIGNING PHOTOSWITCHABLE FPS
Can every FP be engineered to become an efficient RSFP? Is it possible to completely shut off residual photoswitching that naturally takes place in many FPs and constitutes a nuisance (74)? Answering these questions is of course difficult. However, it is interesting to recapitulate general factors that favor photochromism in FPs. These are mainly the structural flexibility of the chromophore within the pocket, the proton distribution in the chromophore environment, the stabilization of both the on and off configurations, and, perhaps, the susceptibility of the chromophore to photoreduction or photooxidation. Remarkably, these factors can be drastically modified by single-point mutations in the chromophore itself or its close neighborhood (Fig. 3). The nature of the first amino acid of the chromophore triad can be sufficient to induce photochromism. For example, KikG is a FP that displays a chromophore environment nearly identical to that of Dronpa but does not photoswitch. This is because its first chromophoric residue is an aspartate instead of a cysteine, maintaining a much stronger interaction with the β-barrel (52). Amino acids close to the imidazolinone part of the chromophore can also critically influence photoswitching. Mutation of Q69 to a leucine (23) or methionine (11, 25) or of the fully conserved E222 to a glutamine (22) confers photochromism to yellow GFP derivatives, probably due to alterations of the hydrogen bonding network surrounding the chromophore and/or of electron transfer properties. In Anthozoan FPs, mutations of either one of the conserved residues F173, M159, or I157 close to the hydroxybenzylidene moiety of the chromophore can be sufficient to confer photoswitching properties. Mutating M159 and F173 towards smaller residues loosens the interactions between the chromophoric hydroxybenzilidene moiety and its microenvironment. The F173S mutation has been applied to the photoconvertible FPs EosFP and Dendra2, resulting in IrisFP (18) and NijiFP (41) which are thus amenable to multiple phototransformations (photoswitching in green state, green-to-red photoconversion, photoswitching in red state). Mutations of M159 significantly altered the photoswitching yield in variants of Dronpa, mEosFP and Dendra2 (51, 41), and replacing this residue with a bulky tyrosine completely reversed the photoswitching behavior in Padron (53). The residue at position 157 was shown to also control the photoisomerization kinetics in rsFastLime (52) or Dronpa-3 (51), although this residue is not sufficient to generate a photochromic property by its own (41).
The family of reversibly switchable FPs rapidly grows, as new members are constantly engineered or discovered. RSFPs even start to be found in other phyla than cnidarians (containing the Anthozoa and Hydrozoa classes), such as the positive RSFP called hbGFP1 (75), which possesses properties comparable to those of Padron. This predicts further expansion and improvements in the group of switchable genetically encoded highlighters. Future improved RSFPs will include redder, bluer, two-color, fatigue-resistant, and excitation decoupled variants still able to maintain other essential properties such as fluorescence brightness, maturation efficiency, pH stability or monomeric character. Such new variants will reinforce the already key importance of photoswitchable FPs in bio-microscopy and technology.
The authors thank Aline Regis-Faro for insightful discussions. Financial support by the ANR (ANR-07-BLAN-0107-01 and ANR-2011-BSV5-012-01) is acknowledged.