Letter pubs.acs.org/JPCL
Photoswitching of Green mEos2 by Intense 561 nm Light Perturbs Efficient Green-to-Red Photoconversion in Localization Microscopy Daniel Thédié, Romain Berardozzi, Virgile Adam, and Dominique Bourgeois* Institut de Biologie Structurale, CNRS, Université Grenoble Alpes, CEA, IBS, 38044 Grenoble, France S Supporting Information *
ABSTRACT: Green-to-red photoconvertible fluorescent proteins (PCFPs) such as mEos2 and its derivatives are widely used in PhotoActivated Localization Microscopy (PALM). However, the complex photophysics of these genetically encoded markers complicates the quantitative analysis of PALM data. Here, we show that intense 561 nm light (∼1 kW/cm2) typically used to localize single red molecules considerably affects the green-state photophysics of mEos2 by populating at least two reversible dark states. These dark states retard green-to-red photoconversion through a shelving effect, although one of them is rapidly depopulated by 405 nm light illumination. Multiple mEos2 switching and irreversible photobleaching is thus induced by yellow/green and violet photons before green-to-red photoconversion occurs, contributing to explain the apparent limited signaling efficiency of this PCFP. Our data reveals that the photophysics of PCFPs of anthozoan origin is substantially more complex than previously thought, and suggests that intense 561 nm laser light should be used with care, notably for quantitative or fast PALM approaches.
P
assess protein copy-numbers, oligomeric state or clustering tendency.17−20 The first obstacle is that, once photoconverted, PCFPs tend to blink, transiently entering into more or less long-lived dark states.16,21,22 Each PCFP may thus reappear several times along data acquisition, causing overcounting errors. Various blinking correction methods have been proposed to minimize such errors,22−28 but none of them has been found fully reliable. The second obstacle to quantitative molecular counting is the limited signaling efficiency of PCFPs.13 In an ensemble of PCFPs, not all of them fold nor maturate properly, and among those that do, only a fraction can be detected in the photoconverted state.13,27,29 Consequently, the number of detected molecules is lower than expected, limiting the effective labeling density and causing undercounting errors. Although reported numbers vary widely,13,30 a careful study recently suggested a signaling efficiency of ∼60% for mEos2, hinting at the possible role of thus far uncharacterized dark states.29 Here, we investigate further the interplay of such dark states with the photoconversion mechanism of mEos2. Green-to-red photoconversion of PCFPs has traditionally been investigated by absorption spectroscopy using 405 nm light as the sole source of actinic light, despite the fact that intense 561 nm illumination is required to localize single molecules in PALM. Although absorption by green mEos2 at this wavelength is expected to be minimal due to a very small extinction coefficient, the common observation of “readout activation”
hotoActivated Localization Microscopy (PALM) has become a widely used super-resolution technique due to its high performance combined with a relatively simple implementation and because it offers perspectives for quantitative molecular counting.1,2 In PALM, the serial recording of the fluorescence signal from single emitters allows their localization to nanometer accuracy and the subsequent reconstruction of a high-resolution image. The most popular genetically encoded markers suitable for PALM are green-tored photoconvertible fluorescent proteins (PCFPs) such as mEos2, Dendra2, mKikGR, mClavGR2, mMaple, and their various derivatives.3−5 These fluorescent proteins are all derived from anthozoan species such as reef corals or anemones and bear a histidine as the first amino acid of the chromophoric triad. They emit green light upon illumination with cyan (488 nm) light in their native state, which facilitates preliminary adjustments before PALM data acquisition such as cell selection and sample focusing. However, upon illumination with violet (405 nm) light, a β-elimination reaction occurs that results in a backbone chain break coupled to an elongation of the chromophore conjugated electron system, irreversibly shifting fluorescence emission to orange-red colors (for recent reviews, see refs 6 and 7). This red-shifted fluorescence may then be read out with yellow/green (561 nm) light. Stochastic photoconversion of individual PCFPs and recording of their red fluorescence until photobleaching is thus a central concept in PALM microscopy. Many efforts have been dedicated recently to the engineering of bright and monomeric PCFPs.8−16 However, two major limitations have been found to always remain, seriously compromising the quality of PALM data, notably for quantitative studies aiming at counting molecules, e.g., to © 2017 American Chemical Society
Received: July 3, 2017 Accepted: August 29, 2017 Published: August 29, 2017 4424
DOI: 10.1021/acs.jpclett.7b01701 J. Phys. Chem. Lett. 2017, 8, 4424−4430
Letter
The Journal of Physical Chemistry Letters
agreement with the notion that off-switching involves the mEos2 anionic state (Figure S6). The strong sensitivity to violet light of switched-off mEos2, on the contrary, is consistent with a protonated chromophore, possibly twisted or isomerized in a trans configuration.16 The fluorescence decays in Figure 1A,B display a biphasiclike behavior, suggesting a mixture of a reversible phase, due to switching, and of a nonreversible phase, due to either nonreversible bleaching or photoconversion. However, attempts to fit the decays with a corresponding model (Scheme S1A) result in large fractions of the fluorescent protein population ending up in a bleached or photoconverted state (>50%, Figure S7A, D), a finding inconsistent with the highlevel of recovery (>80%, Figure 1) observed upon subsequent 405 nm illumination. Likewise, imposing a rate for bleaching and/or photoconversion matching this level of recovery did not allow us to obtain a satisfactory fit of the experimental data (Figure S7B, E). We checked that the shape of the fluorescence decays was not significantly influenced by spurious effects such as residual diffusion or reduced tumbling of the mEos2 chromophores within the PVA matrix (see Supplementary Discussion and Figures S8, S9). In contrast, the data support the hypothesis that, in addition to nonreversible photobleaching/photoconversion, not only one but (at least) two reversible dark states are formed. In fact, two dark states have been previously identified in red mEos2: a short-lived one attributed to chromophore distortion and a long-lived one attributed to twisting or isomerization.16 It is reasonable to assume that similar dark states can be visited by green mEos2. The corresponding kinetic model depicted in Scheme 1 (see also Scheme S1B) provides satisfactory fitting (Figure S7C, F), with rate constants summarized in Table 1. Interestingly, while the recovery rate of the short-lived dark state Dshort appears essentially light independent, that of the long-lived dark state Dlong increases linearly with the 561 nm illumination power (Figure 2 and Figure S3). This finding is consistent with the notion that the distorted chromophore in Dshort could be sp3-hybridized and not absorb visible light, whereas the protonated chromophore in Dlong still absorbs visible light.32,33 Dshort could correspond to the redox-induced blinking mechanism characterized in green IrisFP, a single mutant of EosFP engineered to display efficient switching in both its green and red states.32,33 Using the extinction coefficient (ε = 79000 M−1cm−1) at 488 nm reported for mEos2 at pH 7.4,16 quantum yields of Φlong ≃ 2.6 × 10−5 and Φshort ≃ 3.0 × 10−5 for off-switching to the longlived and the short-lived dark-states at this wavelength can be deduced, respectively. This results in an apparent overall switching-off quantum yield for mEos2 at 488 nm of Φ = 5.6 × 10−5. This value is much smaller than the switching-off yield found for IrisFP (Φ = 0.014), but only ∼5 times smaller than the value found for Dronpa (Φ = 3.2 × 10−4),34 highlighting that EosFP derivatives intrinsically bear substantial switching capacity. Interestingly, the model predicts that, upon prolonged illumination (e.g., 250 s, as in Figure 1A and Figure 1B) the Dshort dark state rapidly depopulates (see Figure S7C, F), resulting in a limited number of molecules in this state (
