Results and Discussion

Basics of Controlling Single-Molecule Fluorescence.

To switch ordinary dye molecules we intended to use generic radical ion states as controllable off-states. Therefore, single oxazine dyes were attached to biotinylated double-stranded DNA that itself was immobilized on BSA/BSA-biotin-streptavidin coated cover slides according to published procedures (26). Measurements were carried out in aqueous buffers, which allowed enzymatic oxygen removal and addition of ascorbic acid (AA) as reductant and methylviologen (MV) as oxidant.

Although it is comparatively simple to prepare a fluorophore in its radical ion state through photoinduced electron transfer with a reductant or oxidant in solution (23, 25), it is challenging to stabilize the radical anion and prevent thermal return to the ground state. This requires careful removal of all electron acceptors from solution. By using enzymatic oxygen scavenging, the potential oxidant oxygen is removed to only low micromolar concentrations (27). We therefore selected oxazine derivatives with high reduction potential, i.e., the reduced state has very low energy so that reoxidation by oxygen with E_red = −0.57 V becomes inefficient (28). To describe the photophysical properties of single fluorophores taking into account the interaction with reducing and oxidizing agents, we used a 5-level scheme adapted to the thermodynamic parameters of the oxazine derivative ATTO655 (Fig. 1A). The energy levels were obtained by cyclic voltammetry as well as from steady-state absorption and fluorescence spectra [see supporting information (SI) Text for details): The zero–zero energy E_0,0 = 1.86 eV, the first 1-electron reduction potential E_red = −0.42 V (vs. SCE), and the first 1-electron oxidation potential E_ox = 1.31 V (vs. SCE). The exact energy of the triplet state is not known, and it is assumed to be ≈0.3 eV lower than the singlet state (i.e., at ≈1.56 eV).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 1.

Photoinduced processes of the fluorophore ATT0655. (A) Scheme. After excitation to the first excited singlet state S₁, fluorescence is emitted at rate k_Fl (neglecting nonradiative processes). Intersystem crossing competes with fluorescence and leads to the infrequent formation of triplet states T₁ with the rate constant k_Isc. The triplet state is depopulated by intersystem crossing with rate constant k_T or by electron transfer. This may occur either through oxidation by methylviologen (MV) forming a radical cation F^•+ or through reduction by AA yielding a radical anion F^•−. The radical ions could be recovered to singlet ground-state fluorophores by the complementary redox reaction. Direct electron transfer from the singlet manifold (k_RedS, k_OxS) has to be taken into account at higher redox-agent concentration. (B and C) Fluorescence transients of ATTO655-labeled dsDNA immobilized in aqueous environment after oxygen removal, addition of 100 μM AA (B) and additionally 100 μM MV (C), respectively. The transients are binned in 10 ms. Samples were excited at 640 nm with an average excitation intensity of 1.5 kW/cm².

After placing a molecule in the laser focus, adding AA (100 μM) as reductant and removing oxygen, single-molecule fluorescence transients such as in Fig. 1B are obtained. A short fluorescent spike is observed terminated by rapid apparent photobleaching. In the presence of AA and MV (both 100 μM), however, single ATTO655 molecules show pronounced blinking with typical on- and off-times in the millisecond range (Fig. 1C). The fact that the fluorescence of ATTO655 resumes in the presence of the oxidant supports the idea that the fluorophore in Fig. 1B did not photobleach but instead entered a dark and—under these conditions stable—reduced state. The appearance of blinking for the investigated oxazines is attributed to inefficient recovery from the reduced state by MV although the presence of both reductant and oxidant have successfully removed blinking for some other fluorophores (23). Accordingly, the oxidant MV, which exhibits a reduction potential E_red = −0.69 V, is not able to efficiently depopulate the reduced state because of a calculated free energy for reoxidation of ΔG = 0.27 eV. By using oxygen rather than MV, the off-times are considerably longer with τ_off = 1,800 ± 1,000 ms.

Switching Single Oxazine Molecules.

To test the hypothesis that ATTO655 in the presence of only AA is not photobleached but indeed switched off to a stable reduced state, we scanned surfaces of immobilized dsDNA doubly labeled with Cy3B and ATTO655 using alternating laser excitation (Fig. 2) (29). Half of the time per pixel, the sample is excited with 533-nm light for excitation of the green dye Cy3B, and half of the time the red dye ATTO655 is excited by 640 nm. The first scan (Fig. 2A) was carried out in PBS containing ambient concentrations of oxygen, in which ATTO655 shows stable emission (see Fig. S1), to locate the doubly labeled DNAs. Molecules containing only Cy3B (Fig. 2, green), only ATTO655 (red), or both green and red dyes (yellow) were observed.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 2.

Confocal fluorescence images (6 × 6 μm) of ATTO655-Cy3B-labeled dsDNA. The colors green and red encode for the overall fluorescence intensity after green and red excitation, respectively. dsDNA bearing the 2 dyes are hence visible as yellow spots. (A) The first image was acquired in PBS. (B, D, F) The second scan shows the same part of the cover slide after the buffer was deaerated and 100 μM AA was added. (C, E, G) Rinsing of the surface and refilling the chamber with standard PBS switches the molecules back into their fluorescent form. This cycle of reversible redox-switching of single ATTO655-molecules was repeated several times, 3 cycles are shown in A–G.

Before taking the second scan of the same area, oxygen was removed, and 100 μM AA was added. Under these conditions, the scanned area shows only green spots because ATTO655 rapidly enters a long-lived dark state (Fig. 2B). The green spots exhibit some dark pixels because of the transient formation of radical anions (23). Because the sample can slightly move upon buffer exchange, the green dye merely served to identify the same molecule positions as in the previous scan. After all red dyes are switched off, the buffer is exchanged again by washing the surface 3 times with PBS, which contains oxygen. In the subsequent scan (Fig. 2C), all ATTO655 molecules reappear. This procedure can be repeated several times and 3 such cycles with perfect switching of ATTO655 (see circled molecules) are shown in Fig. 2 A–G. Alternatively, MV can be used to switch ATTO655 on again, evidencing that it is the oxidation that turns ATTO655 back on. However, the efficient removal of MV from the chamber is more difficult and requires many more washing steps. The described switching procedure was also conducted with other oxazine dyes such as described in Fig. S2.

The data of Fig. 2 indicate that oxazines act as molecular switches where, in photodynamic equilibrium, the fluorescence is switched off under reducing conditions and switched on by using MV or oxygen. Important properties of molecular switches are (i) their bistability (i.e., the stability of the 2 states), (ii) the reliability and response time of switching, and (iii) their fatigue resistance (repeatability). Obviously, one critical issue is the stability of the reduced state. The observation of very long off-times of the reduced state in the absence of oxidants supports the idea that the radical-anion state is nonreactive and therefore is not considered a photobleaching pathway. To determine its thermal stability with respect to return to the ground state, single ATTO655 molecules were placed in the laser focus to record fluorescence transients in the presence of AA and after removal of oxygen, i.e., such as in Fig. 1B. After once being switched off, ≈50% of the molecules (50 investigated) remained in the off-state for the entire recording time of 120 s. The remaining ≈50% showed 1 or more small photon bursts on the time scale of minutes indicating transient recovery to the ground state. Because enzymatic oxygen scavenging reduces the oxygen concentration to ≈15 μM and does not completely deplete oxygen (27), the reduced state itself is likely thermally stable and requires a reaction partner such as oxygen to return to the ground state. The variability of thermal stability of the reduced state might then be related to the different accessibility of oxygen to the fluorophores because of local heterogeneity (9).

The repeatability of the switching procedure was tested in the presence and in the absence of oxygen. The buffer contained the reductant AA as well as the oxidant MV at concentrations of 100 μM, 250 μM, and 500 μM each. As shown in Fig. 1C, the molecules exhibited frequent blinking because of continuous switching. Fluorescence transients were recorded for ≈50 molecules until irreversible photodestruction occurred. During this period, the molecules went through 436 ± 32 up to 3,016 ± 669 switching cycles per molecule (see Fig. S3). This number of switching cycles is significantly higher than for other single-molecule switches (7–9, 11–13, 30). A decreasing concentration of oxidant and reductant resulted in a lower number of switching cycles (Fig. S3) because of a smaller cycling rate but comparable photobleaching probability. Measurements without oxygen removal showed an average increase in switching cycles of ≈60% because of the faster switching with the additional oxidant oxygen in the solution and, again, comparable photobleaching probability.

Quantification of Photoinduced Processes.

The presence of reductant and oxidant at the same time offers the possibility of a “continuous switching mode,” whose response time is influenced by the encounter rate with the redox agents (i.e., their concentrations) and the reaction rate constants. In the following, we studied the influence of different concentrations of the redox agents on the switching properties and elaborated the details of the photophysical scheme shown in Fig. 1A. First, we determined the intersystem crossing rate k_isc and the rate from the triplet state into the ground state k_T for ATTO655 in the presence of oxygen using FCS to k_isc = (1.2 ± 0.2) × 10⁵ s⁻¹ and k_T = (6.6 ± 2.2) × 10⁴ s⁻¹ (experimental results from FCS and further details are in SI Text, and see Fig. S4) (31). To determine the remaining rate constants, fluorescence transients of ATTO655 using different concentrations of AA and MV were recorded. Beginning with transients recorded at constant MV concentration (100 μM) and varying AA concentration, 2 components are observed in the autocorrelation of transients recorded at 10 μM AA (see Fig. S5A and details on data analysis in SI Text). We ascribe the long component to the formation of radical anions and the short component to the triplet state or the radical cation. Increasing the AA concentration, the short component vanishes, because triplets (and possibly oxidized states) are depopulated by AA more quickly and the on-times are reduced (Fig. 3 A and D). At the same time, the off-times remain constant (Fig. 3 A and D). When varying the concentration of MV and keeping the AA concentration constant at 100 μM, the on-times remain constant, and the off-times become shorter with increasing MV concentration (Fig. 3 B and D). If the formation of radical cations via k_OxS or k_OxT played a role, the on-times should become longer with increasing MV concentration. This is because the pathway through a radical cationic state would, at 100 μM AA, be very fast, i.e., without noticeable long off-state. Accordingly, the number of long off-states assigned to the reduced state should decrease with increasing MV concentration if the pathways through a reduced state and through an oxidized state competed. Thermodynamic considerations using the Rehm–Weller equation ΔG_cs = e[E_ox − E_red] − E_0,0 + C (32), where e is the unit electrical charge and C the negligible Coulombic attraction energy, support the idea that oxidation from the triplet state (ΔG = 0.44 eV, i.e., endergonic) and from the singlet state (ΔG = −0.14 eV, i.e., slightly endergonic) cannot be detected because radical cation states of ATTO655 are not significantly populated, and the rates k_OxS and k_OxT are negligibly small. This part of the scheme is therefore shaded dark gray, and the corresponding red reaction arrows are dashed (Fig. 1A). Accordingly, all observed long dark states are reduced states that are depopulated by MV with rate k_Ox′·[MV] and the short component in the autocorrelation of Fig. S5A can clearly be assigned to the triplet state with k_T in the absence of oxygen being k_T = (5.5 ± 2.5) × 10³ s⁻¹ (see SI Text for analysis). The independence of the action of AA and MV is a remarkable result because it implies that 2 variables are available to independently manipulate on- and off-state lifetimes over a broad range. AA controls the duration of the on-states, whereas MV influences the off-states. When both redox reagents are used simultaneously at equal concentrations, their effects are additive, further proving their independent applicability (Fig. 3 C and F). It is noteworthy that these dependencies hold in the absence (Fig. 3 A–C) and in the presence of oxygen (Fig. 3 D–F). Some changes in the absolute values are observed that are ascribed to the influence of oxygen on some of the rates such as k_T as well as on the effective AA activity.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 3.

Dependence of the on- and off-state lifetime τ_on and τ_off on the concentration of the reductant AA and the oxidant MV in the absence and presence of oxygen. (A and D) AA dependence of switching at constant 100 μM MV concentration. (B and E) MV dependence of switching at constant 100 μM AA concentration. (C and F) The on- and off-times with simultaneous change of MV and AA. The first row shows data measured without oxygen, whereas all data shown in the second row were measured in the presence of oxygen. The error bars indicate the standard deviation from the mean. The data were fitted by using an allometric function (f(x) = a·x^b) to guide the eye.

Further experiments with terminally labeled ATTO655 showed that less reducing and oxidizing agents have to be used to obtain the same variations of on- and off-times compared with internally labeled ATTO655 DNA (see Fig. S6). This is explained by the better accessibility of the terminally labeled ATTO655 compared with the internally labeled ATTO655, which shows the tendency to be firmly bound to the DNA (see anisotropy and fluorescence lifetime measurements in SI Text). Similar results as for ATTO655 were obtained for the longer-wavelength oxazines ATTO680 and ATTO700 (see Figs. S2 and S7), emphasizing the general applicability.

After clarifying the independent role of AA and MV and the absence of oxidized states, we focused on the rate of reduction of the excited singlet state k_RedS and the rate for reduction of the excited triplet state k_RedT as well as the reoxidation rate k_Ox′, which are the remaining rates in Fig. 1A. Because of the long lifetime of the triplet state compared with the lifetime of the first excited singlet state, it is expected that k_RedT is responsible for the formation of radical anions at low reductant concentrations, whereas k_RedS becomes important in the millimolar range. The following discussion assumes that the number of photons, emitted during 1 on-time (on-counts N_on), is independent of excitation conditions. The on-counts can be described as the ratio of the sum of the rates depopulating the S₁ and T₁ states toward the ground state to the sum of rates that depopulate the S₁ and T₁ states yielding the radical anion: The term for N_on is additionally weighted with the detection efficiency η_det and the fluorescence quantum yield Φ_fl. With the fluorescence lifetime without additional quenching τ and the relation τ⁻¹ = k_Fl + k_nr + k_Isc, Eq.1 can be rewritten as: Fig. 4 shows a plot of the on-counts N_on versus the AA concentration obtained from data such as in Fig. 3A.

• Download figure
• Open in new tab
• Download powerpoint

Fig. 4.

Plot of the detected photons during an on-cycle, on-counts N_on, obtained from the fluorescence transients, versus the AA concentration. The error bars indicate the standard deviation from the mean. A fit using Eq. 2 is shown as black solid line.

The data could be well fitted by using Eq. 2 (Fig. 4, black line) with the ATTO655 fluorescence lifetime τ = 3.69 ns, the intersystem crossing rate k_isc = (1.2 ± 0.2) × 10⁵ s⁻¹ and the triplet rate constant k_T = (5.5 ± 2.5) × 10³ s⁻¹, yielding rate constants of k_RedS = (9.3 ± 5.2) × 10⁷ s⁻¹M⁻¹, k_RedT = (4.6 ± 2.4) × 10⁷ s⁻¹M⁻¹ and the product η_det·φ_fl = 0.08 ± 0.03. The faster rate constant from the singlet compared with the triplet is in accordance with expectations from Marcus theory because the driving force for the reaction is larger from the singlet state (33). Qualitatively, the S-shape of the graph can be understood by the reduction of the triplet state at low AA concentration that saturates at ≈500 μM AA. At higher AA concentrations, quenching of the singlet state becomes important and bends the graph downward again. The overall small values of the rate constants for the direct reduction of the singlet and triplet states is in good agreement with the fact that the addition of AA does not reduce the fluorescence lifetime up to 100 mM AA. By using reduced states formed from both singlet and triplet states enables an even broader dynamic range of dark state formation than shown in Fig. 3 with the possibility to generate dark states after, for example, 1 emitted photon, which is important for some superresolution microscopy approaches (34).

Finally, we determined the reoxidation rate k_Ox′ by MV from a fit to the τ_off graph in Fig. 3B to k_ox′ = (1.1 ± 0.1) × 10⁵ s⁻¹M⁻¹ because τ_off = (k_Ox′·[MV])⁻¹. All rates determined are summarized in Table S1.

The presented results demonstrate an exquisite control of the oxazine fluorescence that even allows switching for a well-adapted fluorophore-environment system. This example shows that for construction of fluorescent switches, not only the molecule itself is important for the switching properties but that a fine balance with the experimental conditions can make molecules switchable that are not commonly considered switches. In principle, every fluorophore might possibly be switchable under the appropriate conditions. Vice versa, fluorophores might be constructed that show the required properties in a preset environment. Besides applications of such fluorophores as specific ion sensors or sensors of local redox conditions, e.g., inside living cells, the controllable fluorescence is of potential interests for superresolving fluorescence microscopy.

Superresolution Imaging with Single-Molecule Switches.

Superresolution microscopy that circumvents the diffraction limit of light by subsequently localizing single photoswitchable molecules has opened up a field merging single-molecule fluorescence spectroscopy with diffraction-limit-breaking microscopy (2, 35–37). The common principle of subdiffraction resolution microscopy by subsequent single-molecule localizations is that only 1 fluorophore is active for a diffraction-limited area at any given time, and this fluorophore is localized by imaging with a sensitive camera. Thus, it is required that at any moment, most molecules are prepared in a dark state. Recently, it has been demonstrated that simply blinking of molecules, e.g., because of triplet formation, can be used, which drastically expands the range of usable fluorescent labels (24, 25, 38). Thus far, however, this blinking required oxygen removal (24, 25, 39), and the on-counts and off-times were poorly defined, although they determine the obtainable temporal and spatial resolution (25).

To demonstrate the potential of the presented molecular switch for superresolution microscopy, we recorded subdiffraction resolution “blink microscopy” images of ATTO655-phalloidin-stained actin filaments immobilized on cover slides (Fig. 5 A and B) and actin bundles in fixed cells (Fig. 5 D and E) (25). For data acquisition, 8,000–16,000 frames were recorded with total internal reflection fluorescence (TIRF) microscopy using appropriate concentrations of redox-active agents (see SI Text for details on superresolution imaging and analysis) (35, 36). The resolution of a blink microscopy image is mainly determined by the localization precision and the number of molecules that can be localized in a diffraction-limited spot (25). The on-counts N_on mainly determine the localization precision (40). Because reductants are used to control N_on, the localization precision is controlled simultaneously. This influence is exemplarily shown for the localization of single ATTO655-labeled dsDNA at 100 μM AA and 25 μM AA, yielding standard deviations of 26.8 and 18.7 nm, respectively (see Fig. S8). To increase the number of molecules localizable within a diffraction-limited area, the ratio of τ_off/(τ_off + τ_on) should be as high as possible. Therefore, we reduced the on-times by using higher excitation intensities for imaging (16–50 kW/cm²) compared with confocal measurements displayed in Fig. 3 and kept the off-times long by using low concentrations of oxidant. Thus, the adjustability of the on-counts and off-times allows choosing between spatial and temporal resolution. Fig. 5A shows a TIRF microscopy image of actin filaments immobilized on glass cover slides obtained from the first image frames when most molecules are still in their on-state. The reconstructed superresolution image in Fig. 5B clearly resolves structures that are not discernable in the standard TIRF image, as, for example, 3 actin fibers whose cross section profile is displayed in Fig. 5C. Using Gaussian fits, we determined the distance between 2 adjacent fibers to 88 nm (Fig. 5C). In cells, actin filaments form bundles that are observed in the TIRF image of NIH/3T3 fibroblasts labeled with ATTO655-phalloidin. These bundles are broader than the individual filaments shown in Fig. 5 A and B, but the blink microscopy image in Fig. 5E still resolves, for example, 3 actin bundles that are not resolved in the TIRF image (see Fig. 5 D–F).

• Download figure
• Open in new tab
• Download powerpoint

Fig. 5.

Total internal reflection fluorescence microscopy of ATTO655-phalloidin-labeled single actin filaments (A and B) and bundled actin filaments in fixed NIH/3T3 cells (D and E). (A and D) TIRF microscopy images and (B and E) Blinkmicroscopy images with subdiffraction resolution. (C and F) Histograms over the regions marked with a white rectangle in B and E and the corresponding peak to peak distances derived from Gaussian fits. The images were recorded with reductant and oxidant concentrations optimized for imaging speed, fluorophore density, and excitation intensity (see SI Text for details).

The adjustability and the possibility to work in oxygenated environments are important advantages of the switchable oxazine fluorophores for superresolution microscopy. Additionally, these fluorophores are much more photostable in the presence of oxygen than, for example, rhodamine or cyanine derivatives, which is also likely related to their redox properties: Because of their high oxidation potential they are not efficiently oxidized by molecular oxygen, which is considered to be a key step in photobleaching pathways (41). Recalling that living cells themselves provide redox active agents such as the reductants glutathione and ascorbic acid as well as oxidants such as quinone derivatives and oxygen, superresolution under in vivo conditions may become possible on the basis of the switchable molecules presented.