Institute of Anatomy and Cell Biology, Heidelberg University, Heidelberg, Germany
BioQuant Centre, Heidelberg University, Heidelberg, Germany
Institute of Physical & Theoretical Chemistry, Goethe-University Frankfurt, Frankfurt/Main, Germany
Correspondence to: Mike Heilemann; Institute of Physical & Theoretical Chemistry, Goethe-University Frankfurt, Max-von-Laue Str. 7, 60438 Frankfurt/Main, Germany. E-mail: email@example.com
The spatial resolution of optical microscopy is governed by the diffraction limit. The consequence is that objects which are closer than about 200 nm (in the imaging plane) cannot be discerned; this hampers the use of conventional optical microscopy in studies of small cellular structures. Microscopic techniques that bypass the diffraction limit are referred to as “super-resolution” microscopy. They build on different concepts, which typically include deterministic (Gustafsson, 2000; Hell and Wichmann, 1994) or stochastic manipulation of fluorophores (Heilemann, 2010; Kamiyama and Huang, 2012; Thompson et al., 2012). Among the different microscopic techniques that are capable of bypassing the diffraction limit of optical microscopy are single-molecule imaging approaches which make use of photoswitchable fluorescent probes (Betzig et al., 2006; Hess et al., 2006; Rust et al., 2006). Different from conventional microscopy, the fluorescence emission is read out from single fluorophores which are stochastically activated into a fluorescent state. The position of a single fluorophore is determined with high precision, and a final image is reconstructed from all single-molecule positions in a pointillistic fashion.
The key to single-molecule localization microscopy (SMLM) is photoswitchable fluorophores (Furstenberg and Heilemann, 2013). On the one hand, fluorescent proteins which can be photoactivated, photoconverted, or reversibly photoswitched were used since the first days of SMLM (Betzig et al., 2006; Hess et al., 2006). A key advantage of fluorescent proteins is that they can be tagged to a target protein and allow for stoichiometric labelling and the generation of stable cell lines, which enables quantitative SMLM (Lando et al., 2012). The demand for novel variants of fluorescent proteins with superior optical properties drives an own research field (Adam et al., 2011; Brakemann et al., 2011; Grotjohann et al., 2011). On the other hand, synthetic fluorophores are used (Heilemann et al., 2008; Rust et al., 2006) with varying underlying photochemical or photophysical mechanisms that allow for photoswitching (Furstenberg and Heilemann, 2013). The advantage of synthetic fluorophores is their superior brightness, which increases the number of photons detected and thus the precision with which a single emitter can be localized.
Single-molecule super-resolution imaging with photoswitchable or photoactivatable synthetic dyes is typically performed in the presence of particular imaging buffers (Heilemann et al., 2008; Rust et al., 2006). These contain reducing agents, such as thiols, which reduce synthetic fluorophores into long-lived non-fluorescent (“off”) states. Occasionaly, oxygen scavenging systems are added (van de Linde et al., 2011b). The underlying photochemical mechanisms are quenching of the triplet state through electron transfer, which generates a radical anion (van de Linde et al., 2011a) and eventually further reaction products (Dempsey et al., 2009; Kottke et al., 2010). However, it is well known that amino acid residues, such as tryptophan, as well as nucleobases, such as guanosine, can also quench fluorescence of synthetic dyes (Doose et al., 2005). The degree of this fluorescence quenching depends on the oxidation potential of the fluorophore: whereas carbocyanines are not quenched by either tryptophan or guanosine, the opposite is the case for oxazines and rhodamines. The association and dissociation of fluorophore and tryptophan was previously exploited in the study of protein folding dynamics, where short-range folding steps were detected by combining photoinduced electron transfer (PET) with fluorescence correlation spectroscopy (Neuweiler et al., 2009).
Imaging a biological specimen requires that fluorescent dyes are brought close to or attached to biomolecules, and it is thus intuitively easy to understand that photoswitching properties will be impaired by these natural fluorescence quenchers, if present in the vicinity. However, these phenomena can also be exploited in an advantageous fashion: picking fluorophores which have the right oxidation potential for a particular natural quencher present in a specific probe would allow super-resolution imaging with synthetic dyes without the need for imaging buffers.
Here, we demonstrate how the photoswitching properties of synthetic dyes are governed by the probe molecule they are attached to, a fact which is often neglected or underestimated. We focus on the actin-probe phalloidin, which is composed of seven amino acid residues including a tryptophan residue that is known for fluorescence quenching through electron transfer (Doose et al., 2005). We demonstrate how SMLM can be performed by using various phalloidin-fluorophore conjugates in parallel with photoswitchable fluorescent proteins, without the need of toxic imaging buffers.
MATERIALS AND METHODS
Primary Cultured Hippocampal Neurons
Primary cultures of rat hippocampal neurons were prepared from E19 rat embryos as described previously (Dresbach et al., 2003). Cells from the embryonic brains were dissociated in calcium- and magnesium-free Hank's balanced salts solution (HBSS) by 20 min incubation with 0.25% trypsin (Sigma). These were plated onto poly-l-lysine-coated coverslips at a density of 60,000 cells/cm2 in Dulbecco's Modified Eagle Medium (DMEM) (Life Technolgies) including 10% fetal calf serum (FCS) (Life Technologies). Antibiotics (100 U/mL penicillin, 100 µg/mL streptomycin) (Life Technologies), and glutamine (2 mM) (Life Technologies). Twenty-four hours after plating, the medium was exchanged for Neurobasal (Life Technologies) including 2% B27 (Life Technologies), antibiotics (Life Technologies), and 0.5 mM glutamine (Life Technologies). Cells were kept in a humidified 95% air, 5% CO2 incubator and were used for imaging experiments after 12 days in culture.
A cell culture stock of Hela cells was washed with PBS (1 mL) (Life Technologies). Trypsin (1.5 mL; Sigma) was added. After 2 min of incubation, trypsin was removed from the cells. Total of 5 mL of medium [DMEM, 10% FCS, 5% PenStrep (Life Technologies), 5% l-glutamine (Life Technologies] was added. The cells have been diluted to a final concentration of 37.5 cells per µL. Total of 400 µL of this suspension were added into each chamber of an 8-well LabTek (Nunc). Cells have been incubated over night at 37°C and 5% carbon dioxide until 70% confluency.
mEos2 (Addgene) was subcloned to replace eGFP from the pAM vector containing Synaptophysin-eGFP (Groh et al., 2008) to generate Synaptophysin-mEos2.
The transfection of 9-day-old neuronal cultures was done as per the protocol from Invitrogen for Lipofectamine 2000 Transfection Reagent. Briefly, four aliquots of lipofectamine reagent were diluted in Opti-MEM Medium (Life Technologies). DNA for synaptophysin-mEos2 was then diluted in the above mix and incubated for 15 min. The DNA–lipid-complex was subsequently added to cells and incubated for 3 days.
Fixation and Staining
Both Hela cells and hippocampal neuronal cultures were first washed in 1 x PBS (Crystal PBS Buffer; Bioline) and fixed in 4 % paraformaldehyde (PFA; Sigma) with 0.5% glutaraldehyde (Serva) for 15 min. Cells were then blocked in 0.1 % Triton (Merck) mixed in 5 % FCS (Life Technologies) for 15 min. The cells were then treated for 2 h at room temperature with 1:200 phalloidin-dye conjugate (phalloidin-ATTO 488; phalloidin-Tetramethylrhodamine B isothiocyanate (TRITC) (Sigma) with initial concentration 20 µM and phalloidin-BODIPY 650 (Invitrogen) with initial concentration 300 U/mL) in the blocking solution. After 5 min washing with PBS, the cells were imaged in PBS immediately. Samples stained with phalloidin-Alexa 647 (Life Technologies) were imaged in oxygen-free aqueous buffer containing 100 mM mercaptoethylamine (MEA; Sigma)
A multi-line argon-krypton laser (Innova 70C; Coherent, USA) was coupled into an inverted microscope (IX71; Olympus, Japan) equipped with a 60x oil immersion objective (PlanApo 60x; NA 1.45, Olympus) and operated in total internal reflection fluorescence (TIRF) mode. Excitation and fluorescent light were separated using a dichroic mirror. Here either FF560/659-Di01 (AHF, Germany) or FF410/504/588/669-Di01 (AHF, Germany) or HC576/661 (AHF, Germany) were used. The fluorescence signal was read out with an electron-multiplying CCD camera (EMCCD; Andor Ixon DU 897, Belfast, Ireland). Illumination intensities were adjusted to 1–5 kW/cm2 (488 nm and 647 nm) and 5–7.5 kW/cm2 (568 nm). Images were acquired with an integration time of 30 ms and EM gain of 200. Typically, between 10,000 and 100,000 images were recorded. Images were reconstructed with rapidSTORM (Wolter et al., 2010).
Super-resolution imaging was performed in PBS. Samples stained with phalloidin-ATTO 488 were irradiated at 488 nm, phalloidin-TRITC at 568 nm and phalloidin-BODIPY 650 at 647 nm. Appropriate emission filters were used (BrightLine HC 550/88, ET Bandpass 610/60 and ET Bandpass 700/75; AHF Analysentechnik, Germany). Alexa Fluor 647 was imaged both in PBS as well as in the presence of 100 mM MEA in oxygen-depleted PBS at pH 7.4 according to the dSTORM protocol (Heilemann et al., 2008).
Dual-color super-resolution imaging was performed sequentially. First, synaptophysin-mEos2 was imaged by illumination at 568 nm. Subsequently, phalloidin-ATTO 488 was imaged by illumination at 488 nm. Fiducial markers (dilution 1:1000; TetraSpeck, Life Technologies) were used to register the two spectral channels. The images were aligned by applying the MultiStackReg plugin of ImageJ.
Single Molecule Analysis
The rate constants kon and koff were determined on sparsely labeled samples. First, a list of all single-molecule localizations was generated with rapidSTORM. All localizations were clustered using a hierarchical cluster algorithm (linkage and cluster) (MATLAB; Mathworks Inc.). The cutoff threshold option was set to a distance using a value of 200 nm for both ATTO 488 and TRITC. Each cluster is searched for localizations in consecutive frames. “On”- and “off”-times were calculated based on the exposure time and the number of frames in which the molecule was consecutively in the on-state or the off-state, respectively. Fluorophores which appeared only once were discarded for the analysis. The data obtained were binned and then fitted to a mono-exponential function. Curve-fitting was performed using Curve Fitting Toolbox in MATLAB.
Quantum yields were determined using a UV-VIS spectrometer (Cary 4E; Varian, Germany) and a fluorescence spectrometer (Typhoon 9400;GE Healthcare, Germany). ATTO 488 and TRITC as well as the corresponding dye-phalloidin conjugates were diluted in either PBS or unfolding buffer (7M guanidinium chloride, 20 mM tris-HCl, 10 mM DTT, pH 7.5) down to concentrations of 2.2–2.9 µM (ATTO 488) and 0.5–0.9 µM (TRITC), respectively, and excited at 488 nm (ATTO488) or 561 nm (TRITC). At the given concentration range, the optical density (OD) of the dyes was kept below 0.05, to guarantee linearity in the absorption and emission processes. Note that the same excitation wavelengths were chosen as in single-molecule localization microscopy experiments.
The fluorescence lifetime of TRITC was measured using a lifetime spectrometer (Fluo-time 100 Compact; PicoQuant, Germany). The free dye and the phalloidin conjugate were diluted in PBS to a concentration of 1 µM. A pulsed (5 MHz) laser diode emitting at 575 nm was used for excitation. The fluorescence lifetimes were determined with the manufacturer's software (PicoQuant).
RESULTS AND DISCUSSION
We have investigated fluorescence quenching as a mechanism for photoswiching for various fluorophores conjugated to phalloidin, a well-known drug that targets polymerized actin structures and is widely used in fluorescence microscopy. Phalloidin is a heptapeptide and contains a tryptophan residue. Considering the electrochemical reduction potential, tryptophan is able to quench the fluorescence of various synthetic dyes, including oxazines and rhodamines (Doose et al., 2005; Vaiana et al., 2003).
We studied fluorescence-quenching-induced photoswitching by tryptophan in phalloidin-dye conjugates at the single-molecule level by choosing three different fluorophores. As a test sample, we labeled the actin cytoskeleton of Hela cells with the respective phalloidin-dye conjugate. We selected ATTO 488 and TRITC, two rhodamine derivatives, as well as Bodipy 650, a bora-diaza-indacene derivative. We first recorded widefield fluorescence images of fixed cells in PBS buffer to verify adequate staining of actin (Figs. 1a–1c). We then increased the laser intensity to induce photoswitching, and recorded single-molecule photoswitching movies in the same PBS buffer. We analyzed the single-molecule movies with a SMLM software (Wolter et al., 2010), and reconstructed super-resolution images (Figs. 1d–1i) for all three synthetic fluorophores selected. We were able to demonstrate that all three dyes were efficiently quenched by tryptophan, which at appropriate laser intensities led to photoswitching and enabled the recording of SMLM images in aqueous PBS buffer, without the addition of reducing agents.
We investigated the type of fluorescence quenching by measuring the fluorescence quantum yield of the fluorophores ATTO 488 and TRITC. A decrease of the relative quantum yields down to 48% (ATTO 488) and 27% (TRITC) for the dye-phalloidin conjugates was found as compared to the free fluorophores. Addition of guanidinium chloride (7 M) restored the quantum yield to the original value. To determine whether quenching of the dye-phalloidin conjugate arises from static or dynamic quenching, the fluorescence lifetime was determined. For the TRITC fluorophore, we measured a fluorescence lifetime of 2.2 ns for the free fluorophore, 1.7 ns in the presence of 30 mM tryptophan, and 2.3 ns for the dye-phalloidin conjugate. These values suggest that fluorescence quenching in TRITC-phalloidin is largely due to static quenching. For ATTO 488, we measured a fluorescence lifetime of 4.1 ns for the free fluorophore. In the presence of 30 mM tryptophan, the fluorescence lifetime decayed biexponentially with 3.0 ns (82%) and 0.7 ns (18%). For ATTO 488-phalloidin, the fluorescence lifetime decayed biexponentially with 3.4 ns (78%) and 1.2 ns (22%). Hence, both static and dynamic quenching contribute in ATTO 488-phalloidin. The values for ATTO 488 are in good accordance to previously published results on fluorescence quenching of fluorophores with a similar chemical structure (Chen et al., 2010). As a result, we observe largely static quenching for TRITC and both static and dynamic quenching for ATTO 488. We assume that the formation of a dye–tryptophan-complex is the nature of the non-fluorescent off-state. In consequence, the photoswitching kinetics are governed by the association and dissociation dynamics of this complex.
The performance of a photoswitch in SMLM depends on the on- and off-switching kinetics of the fluorophore. In order to obtain a high-quality image and in particular for dense structures, it is necessary that the fluorophore has a long-lived off-state, compared to a short on-state. In other words, ratio of the photoswitching rates, kon/koff, should be sufficiently high (van de Linde et al., 2010). We determined the ratio of rate constants from sparsely labeled samples (see Materials and methods) by extracting on- and off-times for TRITC and ATTO488 (see Supporting Information). We found a ratio for kon/koff of 883 for phalloidin-ATTO 488 (Supporting Information, Figs. S1a and S1b) and 1023 for phalloidin-TRITC (Supporting Information, Figs. S2a and S2b). These ratios are sufficiently large to allow imaging even dense structures, such as the cellular actin network. Furthermore, we determined the photon counts from single-molecule localization events. On average, we detected 822 (ATTO 488, determined from >130.000 single-molecule localizations) and 939 photons (TRITC, determined from >280.000 single-molecule localizations), respectively. These values correspond well to published data for these fluorophores under similar imaging conditions (Dempsey et al., 2011). From these photon numbers, we estimated a theoretical localization accuracy of better than 10 nm (Smith et al., 2010).
In a control experiment, we investigated the phalloidin conjugate with the carbocyanine fluorophore Alexa Fluor 647, which is intensely used in super-resolution microscopy (Heilemann et al., 2008; Rust et al., 2006; van de Linde et al., 2011b). The redox potentials of carbocyanine dyes do not favor reduction by tryptophan (and thus photoswitching through tryptophan) (van de Linde et al., 2011a). To demonstrate this, we stained Hela cells with phalloidin-Alexa Fluor 647, recorded a widefield image (Fig. 2a) and increased the illumination intensity in order to adjust the density of single emitters suitable for SMLM. We recorded an image stack until all fluorophores photobleached, reconstructed the image data and generated a reconstructed image (Fig. 2b). Clearly, efficient photoswitching cannot be observed for Alexa Fluor 647 in PBS buffer, and a spotty SMLM image is obtained. By contrast, if MEA (at a concentration of 100 mM) is added to PBS (matching typical imaging buffers for Alexa Fluor 647), single molecule photoswitching occurred, and an SMLM image could be reconstructed (Figs. 2c and 2d). Contrarily to carbocyanine dyes, oxazine dyes (such as ATTO655) are very efficiently quenched by tryptophane; the fluorescence quantum yield is significantly reduced and stable dye–tryptophan complexes are formed, such that phalloidin–oxazine conjugates are less suitable for fluorescence microscopy (Doose et al., 2005).
Operating conventional synthetic dyes as photoswitches without the need of adding a specific switching buffer with high concentrations of reducing agents offers new opportunities for SMLM. For example, it now renders synthetic dyes fully compatible with photoswitchable fluorescent proteins. This was so far limited to very few fluorescent proteins that tolerate high concentrations of reducing agents, such as mEos variants; many other fluorescent proteins are significantly quenched in such imaging buffers (Endesfelder et al., 2011). To prove this concept, we transfected primary cultured hippocampal neurons with synaptophysin-mEos2, a fusion protein targeting mEos2 to synaptic vesicles, fixed the cells and post-labeled with phalloidin-ATTO 488. We then performed dual-color super-resolution imaging in PBS, the optimal buffer for photoconversion of mEos2 and other photoswitchable fluorescent proteins (Endesfelder et al., 2011), and at the same time well suited to photoswitch ATTO 488 in the vicinity to tryptophan.
This approach allowed us to image the spatial relationship of actin and synaptic vesicles at a higher resolution than previously possible. Synaptic vesicles are clustered at presynaptic active zones, yet, the molecular mechanisms ensuring vesicle clustering are still under debate (Sudhof, 2012). Actin, a cytoskeletal protein existing as monomeric G-actin and filamentous F-actin, is one of the candidate molecules thought to contribute to the functional organization of synaptic vesicles (Cingolani and Goda, 2008). Axonal actin, unlike dendritic actin, is organized not as continuous filaments but as rings which are spaced about 190 nm in distance (Xu et al., 2013). Therefore, axonal actin appears discontinuous (Fig. 3a). Furthermore, some of the F-actin signal may arise of glial cells. A large synaptic vesicle cluster labeled with synaptophysin-mEos2 is shown in the center of Figure 3b. Magnified views of the super-resolution images reveal a dense network of F-actin strands (Fig. 3c) and a vesicle cluster containing different densities of synaptic vesicles (Fig. 3d). The overlay of both frames highlights that actin overlaps with synaptic vesicles mostly within dense vesicle areas (Fig. 3e). In summary, these results both demonstrate the utility of direct quenching-induced photoswitching for super-resolution imaging under simple buffer conditions and address the colocalization of actin and synaptic vesicles in presynaptic terminals at a higher resolution than previously achieved.
In summary, we have demonstrated that photoswitching of synthetic dyes is heavily governed by the nanoenvironment of the fluorophore. By choosing phalloidin as a well-defined show-case, we demonstrate that this can be beneficial for the use of several rhodamine and bodipy dyes, as no specific imaging buffer is required. This concept could potentially be generalized toward intrinsically photoswitchable probes by conjugating tryptophan (or other natural quenchers of fluorescence) together with a suitable fluorescent dye to a target biomolecule.
We thank Michaela Kaiser for generating the synaptophysin-mEos2 construct. We are grateful to Joerg Langowski and Katalin Tóth (DKFZ, Division Biophysics of Macromolecules, INF 580, 69120 Heidelberg, Germany) for providing access to absorption and fluorescence spectrophotometers and for help with experiments. We also thank Dirk-Peter Herten and Arina Rybina (BioQuant, INF 267, 69120 Heidelberg) who provided access to a fluorescence lifetime spectrometer. This work was supported by contract research “Methoden für die Lebenswissenschaften” of the Baden-Württemberg Stiftung (grant nr. P-LS-SPII/11) (to M.H. and B.F.) and the CellNetworks Excellence Cluster EXC81 (to T.K.).