Single-Molecule Super-Resolution Imaging
The multiple imaging modes afforded by widefield, confocal, and multiphoton fluorescence microscopies permit noninvasive, temporally resolved imaging of fixed and living cells and tissues with a high level of biochemical specificity. Despite the advantages of traditional fluorescence microscopies, the spatial resolution of such techniques is limited by the diffraction of light (diffraction barrier). This restricts the amount of information that may be captured with standard microscope objectives, generally limiting light microscopy to characterizing micro-scale rather than nano-scale structures. For a detailed discussion of the diffraction barrier in light microscopy, please refer to our article The Diffraction Barrier in Optical Microscopy. Super-resolution microscopies apply novel approaches to circumvent the diffraction barrier, allowing users to acquire nano-scale information with optical systems.
Stochastic Optical Reconstruction Microscopy, STORM, is one of a family of Nobel Prize winning super-resolution Single Molecule Localization Microscopies (SMLM) for the visualization of biological systems with an optical resolution measured in the tens of nanometers (nm) in the x, y, and z directions. Pioneered in the laboratory of Xiaowei Zhuang at Harvard University, this technology is available via Nikon’s N-STORM system. Similar SMLM techniques include Photoactivated Localization Microscopy (PALM) and Ground State Depletion Individual Molecule Return (GSDIM), among many others.
STORM and other SMLMs are conceptually similar techniques: the photochemical properties of the fluorophore are exploited to induce a weakly emissive or non-emissive “dark” state. From the dark state, very small populations of (ideally) fluorophore are returned to an emissive state, excited, and detected. However, in order to be identified, emission profile must exhibit minimal overlap in each image. The centroid position of each identified molecule is statistically fitted, most often to a Gaussian function, and with a level of precision scaling with the number of detected photons. By imaging and fitting single emitters to a sub-diffraction limited area over many thousands of single images, eventually the user will have enough data to create a composite reconstruction of all identified emitters. Figure 1 above illustrates this imaging process. The majority of SMLMs operate using this common generalized concept, requiring virtually identical hardware and data analysis steps, but differing in the exact fluorophore chemistry exploited to realize such “on-off” switching behavior. A comparison of widefield epifluorescence and STORM of the same sample is given by Figure 2 below.
As with all imaging techniques STORM is not without certain critical steps, specifically the user is forced to follow much more stringent sample preparation protocols than dictated by more conventional techniques, such as confocal or widefield fluorescence. Perhaps most important is the proper choice of fluorophore, which is thoroughly detailed herein. However, while fluorophore selection for STORM is considerably more constrained than typical, it should not prevent one from pursuing the many benefits of super-resolved imaging where applicable. Provided here is an in-depth discussion of STORM, and more generally SMLM, with attention to multichannel, three-dimensional, and live cell imaging approaches, and a special focus on the critical steps of fluorophore selection, labeling strategies, and imaging buffer formulation. First, the properties of fluorophores for STORM will be explored. Fluorophores for STORM and SMLM include synthetic dyes, fluorescent proteins (FPs), and even Quantum Dots (QDots).
Properties of STORM Probes
Photon Output
The concept of localizing individual molecules with high precision was first described by the German physicist Werner Heisenberg in the 1930s and was more formally addressed with a strong mathematical foundation during the 1980s – 1990s by several research groups. To summarize their findings, although the image of a single fluorescent emitter is manifested in a finite-sized spot having dimensions limited by diffraction when captured on camera, the precision with which the exact position of the molecule can be determined is much higher (in the range of a few nanometers), provided enough photons are captured. This is due to the fact that each captured photon provides an independent measurement of emitter position, proportional to the square root of the total number of measurements. Methods for determining molecular localization coordinates are generally based on statistical curve-fitting algorithms for fitting the measured photon distribution to a Gaussian function. The goal of this exercise is to determine both the mean value of the photon distribution (μ) and the standard error in the fitted position (the uncertainty, σ), according to the equation:
Where s is the standard deviation of a Gaussian function approximating the true point-spread function of the emitter, N is the number of photons captured from the fluorescent molecule, a is the pixel size of the imaging detector, and b is the standard deviation of the background (including background fluorescence emission combined with detector noise). The first term under the square root symbol on the right-hand side of Equation (1) takes into account photon shot noise, while the second term includes the effect of finite pixel size of the detector. The last term factors the effects of background noise into the equation. By applying these simple techniques, a localization accuracy of approximately 10 nanometers can be achieved when the measured photon distribution of a fluorophore is approximately 1,000 photons and the background noise is negligible compared to the molecular signal. In the case of negligible background, Equation (1) may be approximated as:
As described by Equation (1) and Equation (2), the most critical element in obtaining the high-precision results necessary for precise molecular localization in STORM imaging is to minimize the background noise while simultaneously maximizing photon output from the fluorescent probe.
An important point to remember is that the precision with which the position of a single emitter can be localized is not equivalent to the optical resolution of the STORM reconstruction, which is highly dependent on a number of other variables, including drift, labeling density, etc. Furthermore, the optical resolution of STORM data can be difficult to quantitatively determine using standard techniques. However, in practice one can expect a lateral (x, y) optical resolution of about 20-50 nm from quality STORM data. Axial (z) resolution also scales proportionally with photon output, but is additionally dependent on the 3D detection modality, as will be addressed in greater detail. The effects of photon count, and other factors discussed here, on the quality of STORM reconstructions are explored in Figure 3 below.
Duty Cycle
Fundamental to reconstructing the position of single emitters is spatial and temporal separation of fluorescence emission profiles. Separation enables the sequential identification and fitting of emitters to sub-diffraction limited locations over many thousands of imaging frames. Duty Cycle is the ratio of time a fluorophore resides in the ‘on’ state compared to the ‘off’ state. A probe with a high duty cycle spends a lot of time in the fluorescent ‘on’ state, and conversely ‘low duty cycle’ probes spend more time in a non- or weakly fluorescent dark state. Probes with a high duty cycle are difficult to localize because their overlapping emission profiles preclude accurate identification of single emitters.
It should be noted that a variety of multi-emitter fitting algorithms are available for identifying single molecules with partially overlapping fluorescence emission profiles, albeit with modestly reduced localization precision and greater computational cost. Multi-emitter fitting is available on Nikon’s N-STORM systems as the “fit overlapping peaks” analysis option. This approach is often applied intentionally for live cell STORM imaging, where temporal resolution is of greater priority than spatial resolution, so more (and overlapping) molecules are detected in each frame so as to mitigate motion-induced blur. More recently, deconvolution-based analysis has been introduced for data with heavily overlapping emitters, providing a high performance compromise solution. The development of algorithms for STORM analysis has evolved into a complex field with a number of different options.
Switching Cycles
‘Switching cycles’ refers to the average number of times a single molecule can be cycled between the dark and fluorescent states before permanently photobleaching. This property varies widely between different probes and experimental conditions; many probes may only undergo a single switching cycle while others may be able to switch for many hundreds or thousands of cycles. The buffer chemistry plays a significant role in determining the average number of switching cycles, so its best to start with a trusted recipe and experimentally determine the optimal formulation.
The desired number of switching cycles depends on the type of information one wishes to extract from the system. A probe that switches only once may be superior for quantitative imaging applications, specifically experiments where a 1:1 stoichiometry between the fluorophore and the molecule of interest is needed. Conversely, a large number of switching cycles can be advantageous when trying to resolve fine ultrastructural features, such as nano-scale cytoskeletal assemblies (e.g. the microtubules in Figure 2) and the organization of biological membranes. Furthermore, dyes that undergo a single switching cycle may not be localized for a number of reasons, including photobleaching, overlap with other emitters, etc. This will complicate quantitative analyses where a 1:1 labeling and switching stoichiometry is assumed, and is especially problematic at the beginning of the experiment, when the entire ensemble needs to be driven to a dark state before single molecules become identifiable. Generally, synthetic dyes are able to undergo more switching cycles than FPs, as most FP variants undergo a single switch due to an irreversible alteration of the chromophore structure.
Survival Fraction
The survival fraction is the ratio of fluorophores in an ensemble still capable of switching to those that are permanently photobleached following a given period of illumination (400 seconds has previously been used as a standard). This number may correlate with the average number of potential switching cycles of the fluorophore in question, but not always. For example, under the high intensity illumination typically used for STORM, fluorophores may permanently photobleach without providing a usable emission event.
Contrast Ratio
In STORM imaging, background arises from natural or transfection reagent-induced autofluorescence, as well as from residual fluorescence from surrounding probes in a weakly fluorescent dark state. For example, molecules in a triplet state are still weak emitters, contributing to background but not providing usable emission profiles. Therefore, the fluorescent probes chosen for STORM imaging should display a high contrast ratio, which is defined as the ratio of fluorescence emission before and after photoactivation or photoconversion to a strongly emissive state.
Sampling Rate and Labeling Density
Realizing the theoretical spatial resolution of a conventional optical microscope requires digitally sampling the analog image with at least twice the frequency of the highest spatial frequency present in the diffraction-limited image, a requirement defined by Nyquist-Shannon sampling theory, also referred to as the Nyquist criterion. For a discussion of spatial resolution and sampling, including the Nyquist criterion, please refer to our tutorial on Spatial Resolution.
Applying the Nyquist criterion, an image with a lateral (x, y) optical resolution of 200 nm must be minimally sampled in 100 nm increments in the lateral directions. Generally, the analog signal is sampled 2.3 – 4x the highest spatial frequency in the data in order to account for real-world imperfections. It has been common practice to directly adopt this application of the Nyquist criterion for single molecule localization datasets. Applying this definition, STORM data with a mean localization precision of 25 nm would need to be minimally sampled every 12.5 nm. However, this is a misleading interpretation as a STORM reconstruction is not an intensity-based image, each point in a STORM image is a probability distribution of the centroid position of the identified emitters.
Though the camera pixel size must be small enough to sample the position of individual emission profiles in the raw data according to the Nyquist criterion, the fidelity with which the underlying molecular distribution can be resolved is independent of this metric as typically applied to microscopy data. To put it in perspective, the molecule of interest may be localized with high precision (less than 30 nm), but sparsely distributed in the sample (located more than 30 nm apart). In such a scenario one could conceivably localize every molecule of interest with high precision, but not satisfy the Nyquist criterion.
So how does one determine the quality of sampling in STORM-type data? Figure 4 illustrates that the molecular sampling rate – the proportion of the total population of the molecule of interest that is successfully localized – is more important than any arbitrary spatial sampling standard for accurately reconstructing a STORM image. However, there does not currently exist an accurate method for estimating the molecular sampling rate, which depends on a number of factors, many of which are difficult-to-impossible to accurately quantify. Furthermore, labeling of the specimen is an imperfect process, not every molecule of interest will be successfully labeled and equally represented in the data. Thus even the act of labeling the sample is a source of noise. The researcher must be aware that localization precision and optical resolution are not the same and that classic sampling guidelines do not apply to STORM data.
The necessarily high sampling rate required for building an accurate STORM reconstruction directly translates into the need to label the sample as densely as possible; ideally every molecule of interest should be labeled in order to maximize the molecular sampling rate. An appropriate stoichiometry between the molecule of interest and the chosen fluorescent label must also be determined. Bear in mind that it is possible to label the sample too densely – to such a point that quenching effects, aggregation, and other such impositions limit the practical labeling density. Optimal conditions should, as always, be experimentally determined.
Similarly, one should be cognizant of how the target of interest may be labeled. Many STORM probes, such as Alexa Fluor 647, are charged and thus do not effectively cross plasma membranes, making their application in living systems difficult. Additionally, many labeling systems (e.g. antibodies) are too large to pass through plasma membranes and/or their access limited by steric hindrance within the cellular environment. Though Alexa Fluor 647 can still be used for live cell imaging, it remains limited and requires extra time, labor, and disruptive techniques such as microinjection or electroporation to introduce into the cell. However, fixed cells can be permeabilized using standard immunocytochemical techniques and reagents. Some probes are organelle-specific or even bio-orthogonal (reacting near-exclusively with a specialized functional group present on the molecule of interest and with minimal perturbation to the native biochemistry of the system) while others will accumulate non-specifically within live cells.
Other Considerations for Probe Selection
An often-overlooked aspect of probe selection is imaging laser power. Off switching rates for most synthetic dyes are roughly proportional to the imaging laser power: higher power laser illumination switches emitters to a dark state faster (and can also result in accelerated rates of photobleaching). Ensuring that most of the fluorophore ensemble is in the dark state at any given time requires high laser power, especially for densely labeled samples. Furthermore, some dyes simply require a higher power density to achieve off switching than others, even if the dye has a low duty cycle.
Most high quality (and even ‘high power’) lasers for biological imaging do not provide sufficient power density for inducing photoswitching of most fluorophores. One must also take care not to use excessive power. Generally, the imaging laser should be rated at least 200 mW at the fiber, though some systems, such as the N-STORM, may provide additional beam magnification optics, helping one reach the necessary power density for inducing switching behavior in more difficult probes. Remember that switching kinetics vary vastly even with similar fluorophores, and high laser power may not always be necessary. Fluorescent proteins, in particular, switch effectively at lower power densities.
Unbound and excess dye will contribute (often significantly) to the background, degrading the signal from single emitters and resulting in reduced localization precision. In many cases background signal will preclude STORM imaging entirely (this is especially true for thicker specimens). For live-cell STORM it helps for the probe to be fluorogenic: where probe fluorescence increases significantly upon binding to the structure of interest, thus mitigating detection of unbound dye.
No single probe performs ideally in each of the categories explored here, the real challenge is determining which of these factors offers room for compromise in the context of the experimental demands.
Multi-Channel STORM Imaging
Among the primary advantages of fluorescence microscopy is the capacity to image specimens labeled with more multiple fluorophore types to generate images featuring two or more channels to help unravel the relative organization and interactions between the biological structures or molecules of interest. In cases where two or more fluorescent probes occupy a similar volume, they can be subjected to co-localization analysis to determine their relative overlap, yielding information about potential in vivo molecular interactions. Colocalization analysis of single molecule data is inherently more accurate than that performed on diffraction-limited data.
STORM was originally developed using a pair of orange and red-emitting carbocyanine dyes, Cy3 and Cy5, combined to form what is referred to as an activator-reporter pair, or more colloquially as a “dye pair”. In effect, Cy5 can be switched between the fluorescent and dark states in a controlled and reversible manner with the assistance of Cy3 (Figure 5). Excitation with red laser light (~647 nm) stimulates fluorescence emission from Cy5, and can also switch the dye to a dark state. Excitation of Cy3 results in a non-radiative transfer of energy to Cy5, stimulating its recovery from a dark state, a process referred to as activation. In many cases, activation of the reporter may be accomplished with a high-energy laser and without the aid of a proximal activator dye. The Cy3-Cy5 activator-reporter pair possesses an excellent fatigue rate, as Cy5 can be switched on and off hundreds of times before permanently photobleaching. This dye pair was employed to label short double-stranded DNA molecules, each pair separated by a well-defined number of base pairs, and STORM used to resolve the distance between adjacent dye pairs with a high degree of accuracy.
At the foundation of fluorescence multicolor imaging is the requirement to identify several optically distinguishable probes that can be simultaneously applied to a specimen. In STORM, a number of activator-reporter pairs are possible, each composed of a reporter probe, whose fluorescence is read out, and an activator probe to facilitate the recovery of the reporter to a fluorescent state (e.g. Cy3 in the example above), as illustrated by Figure 5. For example, up to nine distinguishable probes can be formed by a combinatorial pairing of three reporter fluorophores having different emission wavelengths and three activator fluorophores with different absorption spectra (e.g. Alexa Fluor 405, Cy2, and Cy3).
This strategy was utilized in STORM for the first multicolor imaging demonstrations using DNA molecules immobilized on a surface and two-color imaging of immunofluorescence specimens labeled for microtubules and clathrin-coated pits. The primary advantage of using a single reporter dye is that the separate channels are aligned because the localizations are derived from fluorophores with identical optical characteristics. However, this method often suffers from high levels of inter-channel crosstalk. Also dye-pair labeled antibodies are not commercially available and must be generated “in-house”. An example image acquired using the dye pair approach is provided in Figure 6.
STORM is widely performed using spectrally distinct reporter dyes without activator dyes, a technique originally described as direct STORM (dSTORM). The use of spectrally distinct probes allows the user to simply apply commercially available fluorophores and labeling reagents. The downside of this approach is that different reporter fluorophores generally have different buffer requirements. Also, chromatic aberration will cause light of different wavelengths to be focused at different points. This is generally not quite as critical in the diffraction-limited regime, but at the nanoscale the shift can be quite significant. However, this can be corrected for by acquiring calibration images of multicolor sub-resolution fluorescent beads for subsequent warp correction operations. Another approach is to use high performing but spectrally similar dyes excited with a single laser line, with emission split onto two different detectors (or two parts of a single camera chip) using a dichroic mirror. Crosstalk can then be reassigned to the correct channel using ratiometric approaches.
Multicolor super-resolution imaging can also be realized by combining spectrally separated synthetic dyes and FPs as labels. Multicolor imaging using FPs alone has proven to be more difficult because the emission spectrum of the pre-activated state of one FP often overlaps with that of the post-activated state of another due to the very broad absorption and emission spectral profiles of FPs in general. The first demonstration of two-color imaging with FPs coupled the dark-to-green FP Dronpa with the green-to-red photoconvertable FP tdEos for labeling actin and focal adhesions, respectively, in fixed cells. The two fluorophores were imaged sequentially, with tdEos being imaged first followed by Dronpa after all of the tdEos molecules were converted/photobleached. Unfortunately, the need to deplete one probe before imaging the other does not allow simultaneous multicolor imaging and thus presents an obstacle for time-resolved multicolor imaging, especially with respect to correcting for sample drift using autocorrelation methods.
Three-Dimensional STORM Imaging
The expansion of STORM imaging to the third dimension was an important early development for the technique. Unfortunately, accurate information on the axial position of a fluorophore is difficult to obtain when focused in a single 2D plane – necessary given that the speed of on-off switching in STORM precludes the acquisition of a z stack during any single emission event, as is normally performed for traditional 3D optical microscopies. In response, a number of strategies have evolved for 3D STORM imaging, including imaging in two or more different focal planes simultaneously, interferometry, tilted mirrors, and convolution of the point-spread function Fourier transform with a more complex mathematical model (such as the double-helix point-spread function) using a spatial light modulator.
One of the simplest and most straightforward techniques involves determining the axial position of a fluorophore on the basis of astigmatism present in the image of single emitters as a function of z depth. Originally developed as an auxiliary system for conducting three-dimensional imaging with STORM, this approach requires the addition of a weak cylindrical lens positioned in the emission path. The ellipticity of the image of each fluorophore becomes a highly sensitive measure of its distance from the focal plane, while the centroid of the image serves to localize the lateral (x,y) position.
Nikon’s N-STORM systems utilize the astigmatic approach as applied by the laboratory of Xiaowei Zhuang for 3D STORM imaging, yielding approximately 50-90 nm axial resolution over a range of about 1 micrometer (μm) for the N-STORM, and 4+ μm for the N-STORM 4.0 by using a proprietary focus offset mechanism to perform z stepping. The principle of 3D STORM imaging using the astigmatic approach is explored in Figure 7.
An alternative approach to STORM-type 3D imaging takes advantage of simultaneous multi-focal plane imaging. By simultaneously imaging two (or more) focal planes in the specimen, the images of activated fluorophores (one over-focused and one under-focused) can be analyzed to fit a three-dimensional point-spread function. The ratio of spot sizes is a monotonic function of the axial position and can thus be quantitatively assessed. Termed bi-plane axial localization microscopy, this technique is capable of imaging through approximately 800 nanometers in the z direction without scanning and several micrometers with the aid of axial scanning techniques. One of the primary benefits of bi-plane imaging is that the lateral resolution of the image is independent of the axial position. A more recent multifocal approach uses a specialized diffraction grating to image 9 different z planes simultaneously onto the same camera chip. This approach allows users to capture the full point-spread function of an emitter in a single camera exposure, but with slightly reduced signal.
3D STORM can also be realized using a widefield microscope featuring a spatial light modulator to generate a double helix (or otherwise structured) point-spread function. The instrument is configured to include what is termed a 4f image processing section in the detection path that is designed to convolve the standard microscope point-spread function (arising from single-molecule emitters) with the double-helix point-spread function generated by a spatial light modulator. This convolution is performed in a conjugate aperture plane by multiplying the Fourier transform of the standard microscope image with that of the specialized point-spread function (a phase-only function). The resulting point-spread function for a single emitter contains two dominant lobes whose angular orientation correlates with the axial position of the emitter. By creating a calibration curve, the axial position of a fluorophore (above or below the focal plane) can be determined with a precision of approximately 20 nanometers (standard deviation) and over a range approaching 2 μm. The most complex aspect of the double-helix microscope is configuration of the spatial light modulator, which requires several matched auxiliary achromatic lenses.
To date, the highest (axial) resolution demonstration of 3D single molecule localization imaging was performed using interferometry. Termed iPALM, the method is capable of achieving resolutions of approximately 10 nanometers in the axial and 20 nanometers in the lateral dimensions. Interferometry is used to measure the differences in two position-dependent pathways (two diametrically opposed imaging objectives) taken by a single photon emitted by the specimen after being recombined with a beamsplitter so that the photon can self-interfere. Among the most critical aspects of iPALM are the requirements of coherence, strict instrument calibration, and tolerance of the fluctuating nature of fluorescence, conditions that are met using a specialized multi-phase beamsplitter. In addition, the imaging (z) depth is limited to approximately 200 nanometers. iPALM has been used to resolve the axial dimension in membranes, focal adhesions, microtubules, and the endoplasmic reticulum. Unfortunately, there are no commercial implementations for iPALM currently and the complexity of the instrumentation (along with the rigorous calibration procedure) may pose limits to its widespread application. Other (non-interferometric) dual objective approaches have been validated as well.
Live-Cell STORM Imaging
The ability to conduct time-resolved image sequence captures of living cells and tissues is one of the hallmarks of fluorescence microscopy, and can be implemented with virtually any contrast or optical sectioning mode. Even though single-molecule approaches to time-resolved imaging are limited, several developments have occurred that show promise for future temporal investigations using these methodologies.
STORM-type super-resolution techniques have generally been thought to be too slow to be valuable for live-cell imaging, as single reconstructions typically require thousands of individual frames of image data. However several investigations have successfully demonstrated time-resolved STORM imaging of multiple targets with minimal motion-induced blur. Single-particle tracking in live cells with nanometer spatial precision has become fairly common, as epitomized by single particle tracking PALM (sptPALM), which has been demonstrated on numerous occasions using optical highlighter FPs to track the movement of up to thousands of protein molecules in the same cell simultaneously, yielding high-resolution information on the underlying cellular dynamics from particle tracks in both one and two-color experiments.
One of the first demonstrations of SMLM in live cells was for resolving focal adhesion complex dynamics, using the photoconvertable fluorescent protein tdEos. Videos with a lateral resolution of approximately 60 nanometers were obtained at frame rates between 25-60 seconds/image, successfully capturing the retrograde transport and elongation of focal adhesions. This data is presented in Figure 8 above.
Live-cell STORM has also been demonstrated using synthetic dyes instead of FPs. This technique relies on the presence of the naturally occurring thiol-containing agent glutathione, which is present in animal cells at millimolar concentrations. An example of this approach is the application of target-specific small molecule dyes (e.g. MitoTracker Red, LysoTracker Red) for live cell STORM. Structure-specific small molecule dyes used for STORM are included in the Table 1 for the reader’s reference. However, such dyes are generally lipophilic and thus do not target any particular molecular species (e.g. a protein of interest). Advanced hybrid labeling technologies, such as SNAP and HALO-tag systems, have been successfully demonstrated for live cell STORM imaging. When imaging synthetic dyes (and to a lesser extent FPs) cell health is a central concern due to phototoxic effects stemming from the application of high power laser illumination.
The application of high-quantum efficiency and low pixel-dependent noise scientific complimentary metal oxide semiconductor (sCMOS) detectors has enabled acquisition of STORM data at much higher rates than previously possible using slower-readout electron multiplying charge-coupled devices (EM-CCDs). Whereas earlier EM-CCD based STORM/SMLM systems were limited to about 30 frames per second (fps), modern sCMOS cameras allow acquisition of STORM frames at over 1000 fps, depending on model and the size of the region of interest (smaller areas are read out faster). The main limitation of this method is that the on-time of most fluorophores is in the tens of milliseconds (ms), so acquisition in the single ms regime partitions the signal from single molecules across many frames, effectively lowering the potential localization precision for each emitter, and with diminishing returns in terms of the number of unique molecules that can be localized in a given time frame. Still, sCMOS cameras have enabled video rate STORM (30 reconstructions per second) in combination with multi-emitter fitting algorithms (allowing the user to maximize the number of molecules in each frame) and FPs fused to a number of different targets.
Probes for STORM Imaging
Synthetic dyes suitable for STORM are found across several structural families, primarily cyanine and rhodamine derivatives, but also oxazines and others. A general guideline is that more red-shifted STORM dyes tend to perform better than their blue-shifted counterparts. In discussing synthetic dye molecules for STORM, we will address relevant aspects of fluorescent dye photochemistry, STORM buffer systems and their mechanisms of action, and finally the utility of specific dyes for STORM, organized by spectral class.
Fluorescent Dye Photochemistry
Many synthetic dyes for modern fluorescence microscopies are xanthene derivatives (the chemical structure of xanthene is illustrated in Figure 9). The basic xanthene core structure is found in all fluorescein, eosin, and rhodamine dyes, differing only in the attached functional groups, and thus sharing many common photochemical properties. The other major family discussed here, the cyanine dyes, distinguishable by a long polymethine bridge (-CH groups with alternating single and double bonds, Figure 9) connecting nitrogen-containing functional groups on one or both sides of the bridge.
Extended pi-conjugated systems allow energy from absorbed photons to be delocalized – stabilizing the electronic transition and thus allowing for the absorption of longer wavelength (lower energy) light. Different functional groups and other structural differences convey each dye with their own unique photochemical properties, including in many cases the ability to switch the dye into a non-fluorescent dark state. Figure 9 illustrates the chemical structures of several common dyes, with the fluorescent pi-conjugated system of each highlighted by its representative emission wavelength color.
A simplified Jablonski diagram is provided by Figure 10, illustrating the electronic energy states of interest of a hypothetical fluorophore for STORM. The singlet electronic ground state, S0, is the lowest available energy level. Note that each electronic energy level has multiple closely spaced vibrational energy levels. A fluorophore can be excited or relax to many different vibrational energy levels of a given electronic state, resulting in the relatively broad absorption (excitation) and emission spectra of most fluorophores. The first excited electronic state S1 is reached from S0 upon absorption of energy equal to the difference between S0 and S1. Absorption of a photon’s energy is generally followed by some degree of vibrational relaxation and internal conversion processes, with subsequent release of the remaining energy in the form of a fluorescence photon, allowing the fluorophore to relax to the ground state S0. Fluorescence mostly occurs as a transition from S1 to S0, electronic energy states higher than S1 relax almost exclusively by vibrational relaxation and internal conversion. Importantly for our purposes, from the S1 state a fluorophore can undergo intersystem crossing to the first excited triplet state T1, a metastable and weakly fluorescent state.
It is from the triplet state that many fluorophores can be reduced to a dark and stable radical state well suited for STORM imaging. Note that certain fluorophores can be instead oxidized to form a dark and stable cationic species (this is the proposed blinking mechanism of the rosamine dye MitoTracker Red). Dark state lifetimes can vary from seconds to hours depending on the fluorophore and buffer composition, with longer lifetimes being desirable for attaining a low duty cycle with STORM. Most rhodamine and oxazine dyes for STORM are reduced to form dark radical species, stabilized by delocalization of the extra electron in the pi-conjugated system of the fluorophore.
The reducing agent is usually a primary thiol, most commonly β-mercaptoethanol (BME) and/or mercaptoethylamine (MEA), though many alternatives have been explored. Reduced rhodamine and oxazine dyes are relatively unreactive, but absorb strongly around 400 nm, a property that can be exploited to drive the fluorophore back to the S0 ground state. A 405 nm (or similar wavelength) activation laser is usually integrated in STORM imaging platforms (including the N-STORM) in order to provide greater control over the degree of blinking.
Oxidizing species, such as triplet oxygen, quench the triplet and reduced dye states, generating harmful reactive oxygen species (ROS) in the process. Long-lived triplet and radical state lifetimes are essential for STORM imaging, and quenching results in higher duty cycle and increased rates of photobleaching due to irreversible oxidative damage of important functional groups. For this reason, an oxygen-scavenging system is generally included in STORM buffer recipes, lowering the effective duty cycle and reducing photobleaching. The enzyme glucose oxidase is the most commonly used oxygen scavenger for STORM, though several similar systems with greater long-term pH stability have been introduced and will be discussed.
The blinking mechanisms of Cy5, Alexa Fluor 647, and Cy7 using standard STORM buffers has been demonstrated to be at least partially due to the photo-induced formation of a non-fluorescent adduct between the dye and the primary thiol. Binding of the thiol interrupts the pi-conjugated system of the dye, thus inducing a dark state, as explored in Figure 11(a) above. The thiol adduct absorbs strongly in the UV-Violet range, a property that can be utilized to break the bond, restoring the dye’s pi-conjugated system (analogous to UV-Violet activation of rhodamine derivatives). The reducing agent tris(2-carboxyethyl)phosphine (TCEP) similarly forms a non-fluorescent adduct with Cy5/Alexa Fluor 647 that can be broken by UV-Violet light (Figure 11(b)).
Atto 655 and Atto 680 are both oxazine dyes validated for STORM imaging. Photoswitching of Atto 655 is similar to rhodamine dyes, but importantly the radical anionic state of the dye can be further reduced to form a colorless leuco-dye, as illustrated by the Jablonski diagram in Figure 10. The leuco-dye is very stable, but as with other dark states the fully reduced leuco form can be quenched by molecular oxygen. The duty cycle of Atto 655 is exceptionally low, roughly matching that of Alexa Fluor 647 under similar conditions.
More recently a series of dyes have been introduced by the laboratory of Stefan Hell (and commercialized by Abberior GmbH) that can be induced to switch without specialized buffer components, specifically without reducing and oxidizing systems or oxygen scavengers. These dyes are caged rhodamine spiroamides (RSAs) and Figure 12 illustrates their general photoswitching mechanism. RSAs are initially in a non-fluorescent caged state (corresponding to the closed ring structure) and uncaged/activated to a fluorescent state (ring opened) by UV-Violet activation light (e.g. 405 nm) or by high intensity imaging light. The uncaged fluorescent form thermally relaxes to the caged non-fluorescent form in milliseconds. These dyes do not require exogenous chemical additives but do perform best in polar solvents, such as physiological buffer solutions or culture medium, and at pH > 5.5.
Buffer Systems
Fluorophore photochemistry informs choice of STORM buffer system. The past ten years have seen incremental improvements of STORM buffer chemistries and several new alternative formulations. The ‘base’ of a STORM buffer system is generally the researcher’s physiological buffer of choice. Most published systems use Tris, but this is due to its more general effectiveness as a physiological buffer. Buffering is important so as to help maintain pH stability, most fluorophores are very pH sensitive and their performance can easily be compromised at non-optimal pH. Tris is popular because it has a pKa close to 8.0 at room temperature and it effectively buffers pH between 7.5-9.0. It has also been shown to slightly increase membrane permeability. Generally, a slightly basic pH (~8.0) is preferable for STORM imaging due to the pH drop associated with the action of glucose oxidase.
Most STORM buffer formulations include an oxygen-scavenging system. As discussed, molecular oxygen acts as a triplet and radical state quencher, shortening dark state lifetimes and producing ROS, which is implicated in common photobleaching pathways. For these reasons an enzymatic oxygen scavenging system is applied, most commonly the ‘GLOX’ system. GLOX consists of the enzyme glucose oxidase: an oxido-reductase extracted from the fungus Aspergillus niger that catalyzes the oxidation of glucose by molecular oxygen, yielding gluconolactone and hydrogen peroxide, a shown by equation 3. This is the mechanism of oxygen scavenging, note that it requires the addition of excess glucose to the buffer to drive the reaction.
The generation of hydrogen peroxide by glucose oxidase is problematic because it is also a harmful ROS. To address this problem, the enzyme catalase is included in the GLOX formulation, catalyzing the decomposition of hydrogen peroxide into water and oxygen as described by equation 4 above. Though this reaction produces oxygen, on net oxygen is being removed from the system, as shown by equations 3 and 4.
Though the GLOX system continues to serve as a convenient standard for STORM imaging, there are some problems associated with its use, especially for longer imaging experiments and when used with rhodamine derivatives, which seem to blink better with a higher steady state concentration of oxygen than remains when GLOX is applied. Over time the pH of a GLOX-based system will become more acidic as gluconolactone is hydrolyzed into gluconic acid, which has a pKa of 3.86. This can present a significant difficulty as the molecular transitions of fluorophores are highly dependent on pH and blinking is generally optimal at slightly basic pH. However, alternative, viable, and pH-stable oxygen scavenging systems have been thoroughly explored.
Pyranose oxidase is similar to glucose oxidase, but catalyzes the oxidation of glucose into the ketone 2-keto-D-glucose (a product that shouldn’t significantly affect pH) as described by equation 5 below. Like glucose oxidase, pyranose oxidase is also used in combination with catalase (equation 6) and glucose, forming the ‘POC’ system. Catalase is included because the ROS hydrogen peroxide is still formed in the reaction. The POC system is a viable and pH-stable alternative to GLOX for STORM, but also more expensive.
Another proven oxygen scavenging system is the combination of protocatechuate-3,4-dioxygenase (PCD) and its substrate protocatechuic acid (PCA). PCD catalyzes the formation of 3,3-carboxy-cis,cis-muconic acid from PCA and molecular oxygen. Despite forming a carboxylic acid as a reaction product, the PCD/PCA system is still more pH stable than GLOX (but less than POC). Interestingly, the PCD/PCA system maintains an approximate 5x lower steady-state concentration of molecular oxygen compared to GLOX.
More recently, “OxEA” buffer was introduced for multi-color STORM, and demonstrated to effectively switch rhodamine derivatives into a dark state at lower laser power than typical. The OxEA buffer utilizes OxyFluor™, an oxygen-consuming membrane fraction of E. coli commercially available from Oxyrase Inc. The substrate for OxyFluor™ is DL-lactate. Performance of OxEA buffer is dramatically limited for cyanine dyes compared to GLOX and comparable buffers, presumably due to higher steady state oxygen concentrations.
The final ingredient in a typical STORM buffer formulation is the primary thiol, or other suitable reducing agent. In addition to MEA and BME, alternative reducing agents include the thiols dithiothreitol, glutathione, and others (though not as widely tested for STORM). Typically the thiol is applied at a concentration of 10-200 mM. The thiol acts as an electron donor, assisting the formation of long-lived anionic dark states as illustrated by the Jablonski diagram presented earlier (Figure 10). Remember, with cyanine derivatives, the thiol forms a non-fluorescent adduct with the dye, which can subsequently be broken with UV-Violet illumination (Figure 11). Thiols may also act as weak triplet state quenchers.
The reducing agent TCEP reversibly quenches fluorescence in Cy5, Cy7, and other related dyes by forming a non-fluorescent covalent adduct with the dye via 1,4-addition to the polymethine bridge (Figure 11), similar to the formation of an adduct with the primary thiols MEA and BME. In addition to TCEP, an enzymatic oxygen scavenging system (e.g. GLOX), ascorbic acid, and methyl viologen are included to minimize transient blinking and bleaching behaviors. The exact formulation of the TCEP buffer is provided in Table 1, along with many common STORM buffer recipes for a variety of fluorophores.
Table 1 - Comparison of Different Organic Dyes and Fluorescent Proteins for STORM Imaging
Listed buffers for each probe are described in detail in Table 2. For simplicity, not all recommended buffers are given exactly as formulated in provided reference for the probe, in which case the reference is not cited parenthetically next to buffer. For probes with multiple species, multiple excitation and emission maxima are listed: (B) = Blue species, (C) = Cyan species, (G) = Green species, and (R) = Red species.
Fluorophore | Excitation Maximum (nm) | Emission Maximum (nm) | Photons/Emission Event | Photoactivation | Buffer(s) | Reference | Comments |
---|---|---|---|---|---|---|---|
Synthetic Dyes | |||||||
488 nm Excitation | |||||||
Vybrant Dye Cycle Violet | 369 (B), ~488 (G) | 437 (B), ~510 (G) | 2409 (1) | UV-Violet (405 nm) | Glycerol/OS (1) | 1-3 | Binds dsDNA. Low cytotoxicity and cell permeable. Does not require photoactivation but can be applied. |
DAPI | 364 (B), ~488 (G) | 454 (B), ~510 (G) | — | UV-Violet (405 nm) | Glycerol/OS (4) | 4 | DNA minor groove binder. Cell impermeant. |
Hoescht 33342 | 350 (B), ~488 (G) | 461 (B), ~510 (G) | — | UV-Violet (405 nm) | Glycerol/OS (4) | 4 | DNA minor groove binder. Cell permeant. |
Hoescht 33258 | 355 (B), ~488 (G) | 465 (B), ~510 (G) | — | UV-Violet (405 nm) | Glycerol/OS (4) | 4 | DNA minor groove binder. Cell permeant. |
SYTO-13 | 488 | 509 | — | — | 50mM MEA/OS (5) | 5 | DNA minor groove binder. Cell permeant. Also binds RNA. |
YOYO-1 | 489 | 509 | — | — | 50mM MEA/OS (7) | 6, 7 | DNA minor groove binder. Cell impermeant. Also binds RNA. Dimeric. |
YO-PRO-1 | 491 | 509 | — | — | 50mM MEA/OS (7) | 7 | DNA minor groove binder. Cell impermeant. Also binds RNA. Monomeric. |
Alexa Fluor 488 | 495 | 519 | 1193 (8) | UV-Violet (405 nm) | 10mM MEA/OS (8) 100mM MEA/OS (9) | 8, 9, 44 | High performance 488nm-excitable dye for STORM |
Atto 488 | 501 | 523 | 1341 (8) | UV-Violet (405 nm) | 10mM MEA/OS (8) | 8, 9 | Highest performance 488nm-excitable dye for STORM |
Oregon Green | 501 | 526 | 900 (10, Live) | UV-Violet (405 nm) | Live-OS | 10 | Recommend for live cell only (cell permeable), better green dyes for fixed work. BODIPY derivative. |
Picogreen | 502 | 524 | (Live) | — | Live-OS+AA (11) | 11 | Binds dsDNA. Cell impermeant |
Atto 520 | 516 | 538 | 868 (8), 1000 (12) | UV-Violet (405 nm) | 143mM BME/OS (8) 100mM MEA/OS (12) | 8, 12 | Recommend Atto 488 or Alexa Fluor 488 instead, if possible |
532 nm Excitation | |||||||
Alexa Fluor 532 | 532 | 552 | — | — | 100mM MEA/OS | 13 | |
Atto 532 | 532 | 552 | — | — | 100mM MEA/OS | 9, 14 | |
561 nm Excitation | |||||||
CF 543 | 541 | 560 | — | UV-Violet (405 nm) | 10mM MEA/OS (15) | 15 | Needs further testing |
TAMRA/TMR (Tetramethyl Rhodamine) | 546 | 575 | 4884 (8) 1100 (10, Live) | — | 10mM MEA/OS (8) Live-OS (10) | 8, 10, 16 | TMR conjugates (TMR Star) often used for SNAP-tag labeling |
DiI | 551 | 569 | 720 | UV-Violet (405 nm) | Live-OS | 17 | Live cell stain for plasma membranes |
MitoTracker Orange | 554 | 576 | — | UV-Violet (405 nm) | Live-OS | 17 | |
Alexa Fluor 555 | 555 | 580 | 2500 (2) | UV-Violet (405 nm) | Glycerol/OS (2) | 1, 2, 44 | A good red dye for STORM, also works well in thiol+oxygen scavenger buffer well (unpublished). |
CF 555 | 555 | 565 | — | UV-Violet (405 nm) | 10mM MEA/OS (15) | 15 | Needs further testing |
DY-547 | 557 | 574 | — | UV-Violet (405 nm) | 10mM MEA/OS (15) | 15 | Needs further testing |
Cy3B | 559 | 570 | 1365 (8, MEA) 2057 (8, BME) | UV-Violet (405 nm) | 10mM MEA/OS (8) 143mM BME/OS (8) | 8, 15 | Highest performing red dye for STORM. |
CF 568 | 562 | 583 | — | UV-Violet (405 nm) | 10mM MEA/OS (15) | 15 | Very promising red alternative, but needs further testing. |
FLIP-565 | 565 | 580 | — | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 18, 19 | |
LysoTracker Red | 577 | 590 | 820 | UV-Violet (405 nm) | Live-OS | 17 | Live cell stain for lysosomes. BODIPY derivative. |
Alexa Fluor 568 | 578 | 603 | 2826 (8), 1700 (10, Live) | UV-Violet (405 nm) | 10mM MEA/OS (8) Live-OS (10) | 8, 10, 15 | Good red dye for STORM |
MitoTracker Red | 578 | 599 | 790 | UV-Violet (405 nm) | Live-OS | 17 | Specific for mitochondria, for live-cell imaging |
ER-Tracker Red | 587 | 615 | 820 | UV-Violet (405 nm) | Live-OS | 17 | Specific for endoplasmic reticulum, for live-cell imaging |
647 nm Excitation | |||||||
DY-634 | 635 | 658 | — | — | 100-200mM MEA/OS (20) | 20 | Needs further testing. |
MitoTracker Deep Red | 644 | 665 | — | UV-Violet (405 nm) | Live-OS | 17 | Specific for mitochondria, for live-cell imaging |
DiD | 644 | 665 | — | UV-Violet (405 nm) | Live-OS | 17 | Specific for plasma membranes, for live-cell imaging |
SiR | 645 | 661 | 630 | — | Physiological Buffer or Culture Medium | 21 | Silicon rhodamine, doesn't require reducing buffer. |
HMSiR | 645 | 661 | 2,600 | — | Physiological Buffer or Culture Medium | 22 | Higher performance silicon rhodamine derivative, doesn't require reducing buffer. |
Cy5 | 649 | 670 | 4254 (8, MEA), 5873 (8, BME) | UV-Violet (405 nm) | 10mM MEA/OS (8) 143mM BME/OS (8) | 8, 23, 24 | One of the highest performing dyes for STORM |
Alexa Fluor 647 | 650 | 665 | 3823 (8, MEA), 5202 (8, BME), 2400 (25), 18050 (26) | UV-Violet (405 nm) | 10mM MEA/OS (8) 143mM BME/OS (8) TCEP (25) Vectashield (26) | 8, 15, 24, 25, 26, 44 | Currently regarded as best dye for STORM |
CF 647 | 650 | 665 | — | — | 10mM MEA/OS (15) | 15 | |
DyLight 650 | 652 | 672 | — | — | 100-200mM MEA/OS (20) | 20 | |
Atto 655 | 663 | 684 | 1105 (8), 1200 (10, Live) | UV-Violet (405 nm) | 10mM MEA/OS (8) Live-OS (10) Culture Medium (26) | 8, 10, 27, 28, 29 | Oxazine dye, first demonstrated in live cells in conjunction with TMP tag |
CF 660C | 667 | 685 | — | — | 100-200mM MEA/OS (20) | 20 | |
DY-678 | 674 | 694 | — | — | 10mM MEA/OS (15) | 15 | |
Atto 680 | 680 | 700 | 1656 (8) | UV-Violet (405 nm) | 10mM MEA/OS (8) Culture Medium (26) | 8, 15, 27, 28, 29 | Oxazine dye, similar to Atto 655 |
CF 680 | 681 | 698 | — | — | 10mM MEA/OS (15) 100-200mM MEA/OS (20) | 15, 20 | |
Alexa Fluor 700 | 696 | 719 | — | — | 10mM MEA/OS (15) 100mM MEA/OS (29) | 15, 30 | |
Atto 700 | 700 | 716 | — | — | 10mM MEA/OS (15) | 15, 29 | |
750 nm Excitation | |||||||
Cy7 | 743 | 767 | 997 | UV-Violet (405 nm) | 143mM BME/OS (8) | 8 | |
DiR | 748 | 780 | — | UV-Violet (405 nm) | Live-OS | 17 | |
Alexa Fluor 750 | 749 | 775 | 703 (8), 2800 (25) | UV-Violet (405 nm) | 143mM BME/OS (8) TCEP (25) | 8, 25 | Does not perform well without TCEP buffer system |
DyLight 750 | 752 | 778 | 749 | UV-Violet (405 nm) | 143mM BME/OS (8) | 8 | |
Fluorescent Proteins | |||||||
PA-FPs | |||||||
PA-GFP | 504 | 517 | 313 (32, Live) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 31, 32 | Recommended only for certain live cell and multicolor experiments as performance for STORM is relatively low. |
PA-TagRFP | 562 | 595 | 906 (32, Live) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 32, 33 | Recommend higher performing green-red PC-FPs except for certain multicolor experiments. |
PA-mCherry1 | 570 | 596 | 706 (32, Live) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 32, 34 | Recommend higher performing green-red PC-FPs except for certain multicolor experiments. |
PAmKate | 586 | 628 | — | UV-Violet (405 nm) or Blue (445 nm) | Physiological Buffer or Culture Medium | 35 | Only viable far-red FP for STORM. |
PS-FPs | |||||||
Dronpa | 503 | 518 | 262 (32, Live) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 31, 32 | Reversibly switchable dark-green. We recommend PS-CFP2 instead for most SMLM applications. |
mGeosM | 503 | 514 | 248 (32, Live) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 32, 36 | Better than Dronpa but higher duty cycle than PS-CFP2 |
Dreiklang | 515 | 529 | 700 (Live) | UV (365 nm) to activate and Violet (405 nm) to deactivate | Physiological Buffer or Culture Medium | 37 | Unique FP where 515 nm light elicits fluorescence, but can be activated or deactivated with two different wavebands. |
mIrisFP | 486 (G), 516 (R) | 546 (G), 578 (R) | — | UV-Violet (405 nm) for dark-green and green-to-red. Blue (488 nm) for dark-red | Physiological Buffer or Culture Medium | 38 | Photoconvertable from green to red state, both green and red states reversibly photoswitchable to dark state. Demonstrated for combined SMLM+pulse chase experiments. |
NijiFP | 469 (G), 526 (R) | 507 (G), 569 (R) | — | UV-Violet (405 nm) for dark-green and green-to-red. Blue (488 nm) for dark-red | Physiological Buffer or Culture Medium | 39 | Photoconvertable from green to red state, both green and red states reversibly photoswitchable to dark state. |
PC-FPs | |||||||
PS-CFP2 | 400 (C), 490 (G) | 468 (C), 511 (G) | 223 (32, Live) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 32, 40 | PS-CFP2 is easily imaged without photo-activation and is a great candidate for multicolor imaging. |
tdEos | 506 (G), 569 (R) | 516 (G), 581 (R) | 1200 (10, Live) 774 (32) | UV-Violet (405 nm) | Live-OS (10) Physiological Buffer or Culture Medium | 10, 31, 32 | Tandem dimer so possible artifacts. Highest photon output in class. |
mEos2 | 506 (G), 573 (R) | 519 (G), 584 (R) | 1200 (10, Live) 745 (32) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 10, 32, 41 | Tends to dimerize, recommend mEos3.2 or mMaple3 instead. |
mEos3.2 | 507 (G), 572 (R) | 516 (G), 580 (R) | 809 (32, Live) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 32, 42 | Along with mMaple3, considered best of the green-red PC-FPs. |
mEos4b | 505 (G), 569 (R) | 516 (G), 581 (R) | 850 | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 43 | Osmium Tetroxide resistant, engineered for correlative EM imaging |
mMaple3 | 489 (G), 566 (R) | 505 (G), 583 (R) | 675 (32, Live) | UV-Violet (405 nm) | Physiological Buffer or Culture Medium | 32 | Most recently introduced PC-FP for SMLM, very high detectable FP: total FP ratio. |
Dendra2 | 490 (G), 553 (R) | 507 (G), 573 (R) | 686 (32, Live) | UV-Violet (405 nm) or Blue (488 nm) | Physiological Buffer or Culture Medium | 32 | Blue light activation and good performance makes Dendra2 good candidate for live cell SMLM. |
Table 2 - Buffers for STORM Imaging
Buffer formulations are cited in Table 1 for specific fluorophores. Note that this table does not account for pH. pH and buffering agent concentration should be experimentally optimized to find the best working solution
Imaging Buffers | Component 1 | Component 2 | Component 3 | Component 4 | Component 5 | Total Volume |
---|---|---|---|---|---|---|
Fixed Cell Imaging | ||||||
MEA/OS (OS = GLOX or POC) |
790 uL Buffer B |
200 uL MEA Stock (200 mM) |
10 uL GLOX Stock or POC Stock | - | - | 1 mL |
890 uL Buffer B | 100 uL MEA Stock (100mM) | |||||
940 uL Buffer B | 50 uL MEA Stock (50 mM) | |||||
980 uL Buffer B | 10 uL MEA Stock (10 mM) | |||||
MEA/OS (OS = PCD) |
790 uL Buffer C |
200 uL MEA Stock (200 mM) | 25 uL PCD Stock | - | - | 1 mL |
890 uL Buffer C | 100 uL MEA Stock (100mM) | |||||
940 uL Buffer C | 50 uL MEA Stock (50 ) | |||||
980 uL Buffer C | 10 uL MEA Stock (10 mM) | |||||
BME/OS (OS = GLOX or POC) |
970 uL Buffer B |
20 uL BME Stock (286 mM) |
10 uL GLOX Stock or POC Stock | - | - | 1 mL |
980 uL Buffer B | 10 uL BME Stock (143 mM) | |||||
985 uL Buffer B | 5 uL BME Stock (71.5 mM) | |||||
989 uL Buffer B | 1 uL BME Stock (14.3 mM) | |||||
BME/OS (OS = PCD) |
920 uL Buffer C |
20 uL BME Stock (286 mM) |
25 uL PCD Stock | - | - | 1 mL |
930 uL Buffer C | 10 uL BME Stock (143 mM) | |||||
935 uL Buffer C | 5 uL BME Stock (71.5 mM) | |||||
939 uL Buffer C | 1 uL BME Stock (14.3 mM) | |||||
TCEP | 920 uL Buffer D | 50 uL TCEP Stock (25 mM) | 10 uL GLOX Stock | 10 uL MV Stock (1 mM) | 10 uL AA Stock (1 mM) | 1 mL |
Glycerol/OS (OS = GLOX or POC) |
800 uL Buffer E | 190 uL PBS | 10 uL GLOX Stock | - | - | 1 mL |
OxEA | 720 uL PBS (pH = 8-8.5) | 50 uL MEA Stock | 30 uL OxyFluor™ | 200 uL Sodium DL-lactate | - | 1 mL |
Live Cell Imaging | ||||||
Live-OS (OS = GLOX or POC) Optional MEA |
990 uL Buffer F | 10 uL GLOX Stock or POC Stock | OPTIONAL: 6 uL MEA stock (6 mM) | - | - | 1 mL |
Live-OS (OS = GLOX or POC) Optional BME |
990 uL Buffer F | 10 uL GLOX Stock or POC Stock | OPTIONAL: 5 uL BME stock (71.5 mM) | - | - | 1 mL |
Live-OS+AA (OS = GLOX or POC) Optional BME |
980 uL Buffer F | 10 uL GLOX Stock or POC Stock | OPTIONAL: 5 uL BME stock (71.5 mM) | 10 uL AA Stock (1 mM) | - | 1 mL |
Stock Solutions | Composition | Storage | Notes |
---|---|---|---|
Dilution/Storage Buffers | |||
Buffer A | 10 mM Tris (pH 8) + 50 mM NaCl | Room Temperature, long term | |
Buffer B | 50 mM Tris (pH 8) +10 mM NaCl + 10% glucose | Room Temperature or 4degC, long term | Recommend testing 10-200 mM Tris, higher pH may help due to OS-induced pH drop. |
Buffer C | 50 mM Tris (pH 8) +10 mM NaCl | Room Temperature or 4degC, long term | |
Buffer D | 200 mM Tris (pH 9) + 5% glucose | Room Temperature or 4degC, long term | |
Buffer E | 120 mg/mL Glucose in Glycerol | 4degC | |
Buffer F | L-15 Medium + 2-10% glucose | 4degC, protected from light, ~1 month | For live imaging, can substitute for other growth medium + 75 mM HEPES. Use phenol red-free media. |
Buffer G | 100 mM Tris (pH 8) + 50mM KCl + 1mM EDTA + 50% glycerol | Room Temperature or 4degC, long term | For storage of PCD. |
Reducing and Oxidizing Reagent Stock Solutions | |||
MEA Stock (1 M Mercaptoethylamine) | 77 mg MEA + 1.0 mL 0.25N HCl | 4degC for ~2 weeks | 100x Stock Solution |
BME Stock (14.3 M Beta-mercaptoethanol) | Provided by supplier as ~14.3 M stock. | 4degC, long term | 100x Stock Solution |
TCEP Stock (0.5 M tris(2-carboxyethyl)phosphine) | Provided by suppliers as 0.5 M ampoules. | 4degC, use same day ampoule opened. | 20x Stock Solution |
MV Stock (0.1 M methyl viologen) | 25.7 mg MV + 1.0 mL ddH2O | 4degC for ~2 weeks | 100x Stock Solution |
AA Stock (0.1 M ascorbic acid) | 17.6 mg AA + 1.0 mL ddH2O | 4degC for ~1 month | 100x Stock Solution |
COT Stock (cyclooctatetraene) | 20.8 mg COT + 1.0 mL DMSO | -20degC | Buffer additive for improved Alexa Fluor 647 photon statistics. 100x stock solution. |
Oxygen-Scavenging System Stock Solutions | |||
Glucose Stock | 45% (w/v) solution | ||
Catalase Stock | 17 mg/mL Catalase in dH2O | 4degC for ~1 month | For use in GLOX and POC Stock solutions. |
GLOX Stock | 56 mg/mL Glucose Oxidase + 3.4 mg/mL Catalase Stock in Dilution Buffer A | 4degC for ~2 weeks | Thoroughly vortex contents and centrifuge at max speed, harvest supernatant for use. |
POC Stock | 112 mg/mL Pyranose Oxidase + 3.4 mg/mL catalase in Dilution Buffer A | 4degC for ~2 weeks | Thoroughly vortex contents and centrifuge at max speed, harvest supernatant for use. |
PCD Stock | 1.4 mg/mL Protocatechuate 3,4-Dioxygenase in Dilution Buffer G | -20degC | |
PCA Stock | 15.4 mg/mL Protocatechuic Acid in ddH2O (adjust pH to 9.0 with 10 N NaOH) | 4degC | |
Sodium DL-lactate | 60% (w/w) solution | 4degC | |
OxyFluor™ | Unknown | -20degC | Do not agitate vigorously or exceed warming temperature of 37degC |
Best Synthetic Dyes for STORM
The super-resolution landscape is constantly changing, we expect improvements in available probes, buffer systems, and our understanding of the underlying photochemistry to continue to result in improved probes and labeling options, opening new research avenues for STORM imaging. Thus this list should not be taken as an absolute, but rather as a practical snapshot of the best currently available fluorophores in each spectral class as of the time of this writing.
UV-Cyan Dyes
Probes emitting in the UV to Cyan spectral region are generally unsuitable for STORM imaging. Though the exact reason has not been thoroughly explored, it has been suggested that the high energy light required to excite probes in this region, combined with the high intensities generally required to induce switching, quickly result in greater degrees of photo-damage compared to excitation of red-shifted fluorophores with higher extinction coefficients with lower energy light.
However, it has been reported that the nuclear stains DAPI, Hoechst 33342, Hoechst 33258, and Vybrant DyeCycle Violet can all be applied towards STORM imaging of DNA in fixed cells. It is hypothesized that these dyes can be deprotonated by UV-Violet light, resulting in photo-conversion to a species with red-shifted excitation and emission spectra. Photoswitching of these nuclear dyes has been demonstrated using blue (488 or 491 nm type) laser readout and activation using a 405 nm laser line. Though the detected light does not technically fall within the UV-Cyan spectral range, this does represent a potentially powerful and simple method for STORM imaging of chromatin with common staining reagents. STORM imaging of these dyes does require a more specialized STORM buffer consisting mostly of glycerol. Fortunately, co-imaging with Alexa Fluor 647 in this buffer has been shown to work adequately.
Green-Yellow Dyes
Though green fluorescent dyes generally don’t perform as well as many popular red fluorescent dyes for STORM, they are nonetheless popular for multi-color imaging experiments in combination with leading far-red dyes (Alexa Fluor 647). Co-imaging of red dyes with far-red or green dyes can result in significant bleed-through (depending upon the microscope configuration), complicating analysis and compromising resolution and specificity if not properly accounted for. Multi-color STORM imaging is often performed sequentially using multi-band pass filters, so crosstalk can be significant.
Currently, the best performing green dyes for STORM imaging are ATTO 488 and Alexa Fluor 488, a pair of structurally similar rhodamine derivatives. It should be noted that STORM imaging, especially of blue-shifted rhodamine derivatives, generally requires powerful laser illumination. This is true for both ATTO 488 and Alexa Fluor 488, which are not switched to a dark state as readily as their red-shifted counterparts in standard buffers. 488 nm lasers rated below 100-200 mW are generally unsuitable as they provide insufficient intensity to drive enough of the ensemble to a dark state for reliable identification of single emitters.
If using a SNAP-tag or CLIP-tag labeling system for live-cell imaging, discussed later, one option is Oregon Green, available in appropriately functionalized form from New England Biolabs Inc. (NEB). Importantly, this dye benefits from the natural presence of the reducing agent glutathione in living cells at approximate concentrations of 5-10 uM, depending on cell type, compartment, etc. This precludes the need to add exogenous thiols or other reducing agents to the STORM Buffer, which may compromise the specimen’s natural physiology. An oxygen scavenging system can still be included for best results, but is not strictly required.
The photon statistics of many dyes may be improved by caging in a non-fluorescent state, effectively decreasing the duty cycle since caged dye does not need to be driven to a dark state prior to imaging. This approach helps reduce initial photobleaching, thus allowing for a higher effective labeling density, important for appropriately sampling the structure of interest. Note that this caging approach has also been applied to a variety of rhodamine and cyanine dyes for fixed cell STORM imaging, including ATTO 488, and in some instances increases the average photon output by several orders of magnitude. Note, however, that caged dyes generally undergo only a single switching cycle.
Orange-Red Dyes
The highest performing red dye for STORM is Cy3B. However, it is often more convenient to use Alexa Fluor 568 or Alexa Fluor 555 as they perform similarly to Cy3B and are easier to obtain in pre-conjugated forms from a variety of commercial sources. Cy3B is available as a functionalized NHS ester, maleimide, or as a free acid from GE Healthcare (General Electric Company). However, to our knowledge, it is not available from any commercial source pre-conjugated to common labeling molecules, such as antibodies.
For live cell imaging TMR Star (NEB) has been shown to be suitable for STORM using SNAP-tag labeling, and related systems (e.g. CLIP-tag, HALO-tag, etc.). As with 505-Star, the live cell performance of TMR Star is enhanced by caging the dye in a dark state using sodium borohydride prior to labeling. Several popular lipophilic membrane dyes for live-cell STORM imaging of different organelles fall in this spectral range, including MitoTracker Red, DiI, LysoTracker Red, ER-Tracker Red, and many others. Further development of fluorogenic and lipophilic stains is currently a topic of great research interest.
Abberior offers FLIP 565, an RSA that can undergo several switching cycles, activated by UV-Violet light and thermally relaxing to a non-fluorescent state, with fluorescence lifetimes in the single millisecond range. FLIP 565 has been demonstrated for STORM imaging using the Nikon N-STORM system.
Far-Red Dyes
Without question, Alexa Fluor 647 continues to be the highest performance probe for STORM/SMLM. Alexa Fluor 647 emits approximately 6000 photons/cycle in typical STORM buffer conditions, has a very low duty cycle, undergoes many switching cycles, is extremely well validated in a variety of imaging conditions, and is readily available from commercial sources conjugated to a variety of labeling reagents. Cy5 is structurally very similar to Alexa Fluor 647, but the latter is generally preferred, in part due to Cy5’s susceptibility to degradation by ozone. Many far-red dyes exist in addition to Alexa Fluor 647 and Cy5 for STORM imaging, but generally not matching their performance for STORM.
A new silicon rhodamine (SiR) far-red emitting dye is available as a benzyl guanine conjugate for SNAP-tag labeling (NEB) as SNAP-Cell 647-SiR. Unlike Alexa Fluor 647 and Cy5, SiR is cell permeable and usable for live-cell imaging via the SNAP-tag system. Though SiR does not require specialized STORM buffers, its performance for STORM is somewhat limited by its relatively low photon output and mediocre duty cycle characteristics. More recently, HMSiR was introduced with improved switching and photon output.
Live cell STORM can be performed with a variety of far-red lipophilic and/or organelle-specific dyes, including MitoTracker Deep Red and DiD. ATTO 655 is a popular far-red choice for certain applications as it is an oxazine dye that can be shelved in a stable and colorless leuco form for very long periods of time, resulting in a low duty cycle. It has been demonstrated for STORM in conjunction with trimethoprim (TMP) tags in live cells (similar to SNAP-tag type system). Unfortunately, ATTO 655 has low photon output (~700 photons/switch), limiting its utility for STORM.
Near-Infrared Dyes
Near-Infrared (NIR) dyes can be difficult to implement for STORM imaging, however using specialized STORM buffers and imaging schema their performance most closely approaches that of the best far-red dyes compared to fluorophores in other parts of the visible spectrum. Using the TCEP-based buffer system described earlier, Alexa Fluor 750 exhibits an average of about 2800 photons per emission event (compared to about 500 photons/switch without TCEP). Moreover, Alexa Fluor 647 still exhibits average photon counts of over 3000 per emitter in the TCEP buffer system, a decrease compared to GLOX/thiol based buffer systems, but of great utility for two color imaging with Alexa Fluor 750 with similar precision. It should be noted that with typical GLOX/thiol buffers, NIR dyes such as Alexa Fluor 700, Alexa Fluor 750, and Cy7 generally average less 1000 photons per emission event.
One of the main difficulties associated with imaging NIR dyes is the excitation laser. Quality high power 700+ nm lasers are expensive and more difficult to integrate into an imaging platform than visible wavelength lasers. Thus it may be some time before such NIR lasers become widely available on turnkey systems for SMLM and/or STORM. Note that 750 nm gas lasers are also known to induce molecular vibrations in the sample that effectively decrease the localization precision. It is very difficult to resolve certain structures, such as ‘hollow’ microtubules, which are often used as a proof of concept sample for STORM imaging, using near-infrared excitation.
Fluorescent Proteins
Fluorescent proteins (FPs) are structurally distinct from synthetic dyes, and thus behave very differently for STORM-type imaging. FP on-off switching comes in several varieties. ‘Photo-Activatable’ FPs, or PA-FPs, are stimulated by high-energy (usually UV-Violet) light to undergo an irreversible switch from a dark to fluorescent state, usually by some irreversible change to the structure of the chromophore. ‘Reversibly Switchable’ FPs, or RS-FPs, are like PA-FPs in that they are activated from a dark to fluorescent state by high frequency light, but unlike PA-FPs can be repeatedly switched between the two states, allowing multiple localizations of the same FP (usually explained by cis-trans isomerization of the chromophore). The final class, the ‘Photo-Switchable’ FPs, or PS-FPs, undergo an irreversible emission spectral shift in response to light of the appropriate frequency, most commonly violet light is used to convert the FP from a green emitting to a red-emitting species. PS-FPs are the most commonly used class of FPs for STORM-type imaging.
A significant advantage of FPs is that they are genetically encodable. However, this can also be seen as a detriment since (with most experimental approaches) the FP is fused to over-expressed and non-native protein whose behavior may deviate from the physiological norm. Furthermore, only a fraction of the protein of interest can be labeled, resulting in a low molecular sampling rate that makes it difficult to resolve cellular structures with sufficiently high density. More sophisticated (and intensive) methods such as RNA silencing can be used to knockout the native protein, additionally CRISPR/Cas9 can be used to edit the gene of interest to include your FP of choice. It should be noted that these same concerns and methods also apply to genetically expressed protein tags, such as SNAP tag, CLIP tag, HALO tag, etc.
Buffer Considerations for Fluorescent Proteins
FP's are generally much less sensitive to the imaging/switching buffer conditions than synthetic dyes, relying on fundamentally different switching mechanisms than their synthetic counterparts. In fact, the culture medium of most common physiological buffer systems is sufficient for imaging. FPs also switch at significantly lower laser powers than comparable synthetic dyes, making them an attractive choice for live cell STORM applications. However, the low photon yield of FPs sharply limits their utility for imaging fixed cells compared to synthetic dyes.
Phenol red should not be included in the live-cell imaging media, as it will contribute to background signal. Additionally, it has been shown that a low oxygen environment may be beneficial when imaging certain FP varieties, especially in fixed cells. One approach is to add 1.0% (w/v) polyvinyl alcohol to the buffer system or to use glycerol as the buffer base, creating a low oxygen permeability matrix. Enzymatic oxygen scavengers can be applied, but FPs are also sensitive to changes in pH.
A great concern is the identification of STORM buffer systems for synthetic dyes in which FPs still perform adequately, as multichannel imaging often entails combining FPs and dyes and experimentally identifying a workable ‘compromise buffer’ recipe. Most FPs under-perform if the buffer includes a primary thiol, such as MEA and BME. BME should almost never be used when imaging FPs as they are nearly always incompatible, however many FPs still perform adequately at low concentrations of MEA, generally not above 10 mM if possible. Specifically, mEos2 has been shown to work well in co-imaging experiments with Alexa Fluor 647 using 10mM MEA/OS buffer. Oxygen scavenging systems, such as the GLOX system described, generally aren’t detrimental to FP imaging, with the exception of the pH drop caused by long-term imaging using the GLOX system.
In addition to the fluorophore qualities for SMLM/STORM listed earlier, there are several FP-specific concerns that should be noted. Many FPs are prone to dimerization and higher-order oligomerization artifacts, including many varieties with the monomeric “m” descriptor in their name, such as mEos2, which has a proven tendency to dimerize. While this may or may not be problematic for conventional imaging applications, or with certain FP fusion constructs, in the context of STORM it can result in quenching and misleading localization artifacts. Another proposed measure of FP effectiveness for SMLM is ‘signaling efficiency’, defined by the Zhuang lab as the ratio of detectable FP fusion proteins in a cell to its expression level. This is conceptually similar to determining the molecular sampling rate, and accounts for properties such as FP fusion protein misfolding, providing a measure for estimating the theoretically possible labeling density.
Photoactivatable Fluorescent Proteins
There currently exist four PA-FPs that have been demonstrated for STORM. PA-GFP was the first photo-activatable FP discovered, differing from the original wild type GFP by only a single point mutation; unfortunately it usually emits less than 300 photons when irreversibly activated from the dark to green fluorescent state. Remember that PA-FPs can only be activated to the fluorescent state and imaged once. This makes it difficult to resolve fine structural details and, also considering the poor photon statistics, results in poor (low-precision and low-density) reconstructions.
PA-mCherry and PA-TagRFP perform slightly better than PA-GFP, usually emitting closer to 400-800 photons per localization, but still not matching higher performance and spectrally similar PS-FPs. More recently, a photoactivatable version of the far-red FP mKate was introduced, PAmKate. Though not necessarily outperforming other PA-FPs and still far behind the performance of red and far-red synthetic dyes, this is exciting because it appears to be the first usable far-red FP for STORM. PA-FPs are generally difficult to use due to the fact that the unconverted species can’t be visualized. This makes identification of PA-FP expressing cells difficult without either a secondary marker or by activating a small population of the FP for low light visualization prior to STORM imaging.
Reversibly Switchable Fluorescent Proteins
RS-FPs are principally used for RESOLFT-type super-resolution applications rather than single molecule localization methods such as STORM. RS-FPs can be reversibly converted from a dark to fluorescent state, often for several hundreds of cycles, depending on the FP. Generally, an RS-FP is activated by UV-Violet light and fluorescence excited using a more red-shifted wavelength, which also drives the FP back to the dark state. Popular RS-FPs include Dronpa and related variants (Dronpa-3, bsDronpa, Padron, more), which can be activated by a low power 405 nm laser line and excited to fluoresce by low to moderate power 488 nm illumination, which also induces off switching. The switching kinetics of most RS-FPs are very fast, especially with newer varieties, such as RS-EGFP2, which was specifically engineered for RESOLFT imaging and incompatible with STORM. RS-FPs tend to emit low numbers of photons during each cycle, for example Dronpa generally emits an average of 100-200 photons per switching cycle. This nearly precludes imaging without a high sensitivity EMCCD detector in a field moving more and more towards sCMOS detectors with high photon output reporter fluorophores.
Photoswitchable Fluorescent Proteins
PS-FPs are generally the best-suited variety for STORM. In particular, the Eos family of FPs has long been of great utility for STORM-type imaging, their utility extensively demonstrated in numerous publications. Eos FPs are natively green emitting until activated by UV-violet light, at which point they undergo an irreversible chromatic shift to a red-emitting species. The most advanced Eos variety to date for STORM is mEos3.2, which emits nearly 1000 photons on average upon switching. Additionally mEos3.2 is free from the dimerization artifacts that plague many popular FPs, including earlier Eos varieties such as wtEosFP, tdEos, and mEos2. More recently mEos4 was introduced as a high performance FP for correlative STORM and Electron Microscopy (EM), able to withstand the harsh fixation conditions typical of EM sample preparation while still performing well for STORM.
The principal difficulty with using PS-FPs for STORM is that they occupy a broad spectral range when considering both species, from the green through the red, which makes multicolor imaging very difficult. Furthermore, the vast majority of PS-FPs are green-to-red photo-switchers. The notable exception is PS-CFP2, a cyan-to-green PS-FP. Though PS-CFP2 does not exhibit the same photon output as green-to-red PS-FPs, it does have a low duty cycle and occupies a very usable spectral range, emitting almost entirely in the green (the cyan species is extremely dim).
A popular alternative green-to-red PS-FP is Dendra2, which provides similar performance to mEos2, but has noticeably greater monomeric character and can be activated to the red state using blue light (488 nm), which is less harmful to live cells than the typical UV-violet wavelengths used for activation of other probes, making it an attractive choice for live cell work. More recently, an improved variety of the green-to-red PS-FP mMaple was introduced – mMaple3 – demonstrating very high signaling efficiency, low oligomerization tendency, and photon output similar to mEos3.2.
Conventional Fluorescent Proteins
Many conventional (non-optical-highlighter) FPs have been demonstrated for STORM-type imaging. Published examples include the FPs mNeonGreen, EYFP, and Citrine. Imaging of conventional FPs generally follows an approach described as Stochastic Single Molecule Super-resolution (SSMS) imaging, a type of SMLM characterized by an initial ‘bleach’ phase where excessive laser intensities are used to quickly drive the FP ensemble to a dark state, followed by imaging the spontaneous recovery of single FP fluorescence at much lower laser intensity. In general this approach does not perform as well as more conventional methods, but can be of great utility when the researcher already has a well-validated FP fusion protein that they wish to use for STORM.
Quantum Dots
Quantum dots, also known as QDots, are inorganic semiconductor nanocrystals with a number of applications, including fluorescence imaging. Though extremely bright and stable, QDots are not often applied towards STORM-type imaging. QDots are well known for their blinking, a property that has been exploited for super-resolution using the technique Super-resolution Optical Fluctuation Imaging (SOFI), but not for STORM due to extremely short dark state lifetimes, short enough to preclude the spatiotemporal isolation of single emitters using standard approaches. QDot-based STORM imaging instead takes advantage of their stochastic blueing properties. Under constant illumination, the emission spectra of QDots stochastically undergo a blue shift due to photo-oxidation of the QDot core (this may vary widely with different QDot formulations). Imaging sparse populations of blue-shifted and blinking QDots allows for enough separation between emitters to apply STORM analyses. However, this technique has not been widely adopted and we do not recommend its use without further development and standardization.
Labeling Strategies
Though choice of fluorophore is of paramount importance for STORM imaging, equally important is proper labeling of the sample. There exist a variety of labeling strategies that can be tailored towards STORM. Most popular labeling strategies have been adapted for STORM, including immunofluorescence, genetically encoded chemical tags (e.g. FPs, SNAP, CLIP, etc.), bio-orthogonal click labeling, organelle-specific small molecule staining (e.g. MitoTracker labeling), and more. We will explore the benefits and pitfalls of each of these approaches.
Immunolabeling
Most STORM imaging is performed on samples that have been labeled using standard immunofluorescence techniques, more commonly indirect labeling. Though this approach is most easily adopted, the application of large antibody molecules prevents the researcher from realizing the full resolution potential of STORM. Figure 13 illustrates this limit, where it is visually apparent just how far removed the fluorophore can be from the molecule of interest when applying antibody-labeling techniques. Typical immunoglobulin G (IgG) antibodies, as illustrated in Figure 13, measure about 10-15 nm across, their physical size on the same scale as the localization precision of most STORM probes. Thus many researchers prefer to label the antigen with a dye-conjugated primary antibody, an approach known as direct labeling and illustrated by Figure 13(c). Generally, this requires the researcher to conjugate the primary antibody “in-house” with the appropriate dye, as non-labeled primary antibodies are much more widely available from commercial sources.
Indirect labeling has the advantage of being able to use many different secondary antibodies to detect any primary raised in a compatible host animal, making it a much more robust solution than dye-conjugated primary antibodies. Traditionally, indirect labeling was not problematic because, even if there was a separation of 20+ nm between the dye and antigen, it was still on a size scale an order of magnitude lower than defined by the diffraction limit of light (~200 nm with highest NA objectives). This size problem can be mitigated in part by using fluorescently labeled secondary antibody fragments instead of whole IgG. F(ab)2 fragments are generated from IgG antibodies by pepsin digestion, cleaving the Fc portion of the antibody and yielding a pair of F(ab) fragments tethered together. Importantly, the Fc region binds non-specifically to cellular Fc receptors, meaning that application of F(ab)2 fragments generally yields lower background labeling by mitigating nonspecific interactions. IgG can be digested with papain instead, yielding a pair of separate F(ab) fragments approximately half the size of a F(ab)2 fragment. F(ab) fragments are also a good choice because, like F(ab)2 fragments, they lack the Fc receptor, and can better reach antigens in scenarios where steric hindrance is more of a limiting factor.
More recently researchers have sought to address the size problem imposed by antibodies for STORM by using fluorescently labeled “nanobodies”. Nanobodies, also known as single-domain antibodies, or VHH (variable domain of heavy chain) antibodies, only measure about 1-2 nm in each dimension and weigh about 12-15 kDa (typical antibodies in the range of 150-160 kDa and FPs are 27 kDa). The most popular nanobodies are fragments derived from a variety of heavy chain antibody found in camelids, including llamas and alpacas.
Application of fluorescently labeled nanobodies significantly increases the quantitative accuracy of localization due to their small size. Moreover, nanobodies experience decreased steric hindrance within the crowded cellular environment, allowing one to tag epitopes normally hidden or inaccessible to larger antibodies such as IgG. Nanobodies also happen to be much more stable than conventional antibodies. So why haven’t nanobodies totally replaced traditional antibody labeling? In large part nanobodies are still a new technology, very few commercial sources exist and, with the exception of a few target antigens, generation of nanobodies against proteins of interest requires cost-intensive custom production. However, one strategy for STORM imaging is to use cost-effective and more readily available anti-GFP and/or anti-RFP nanobodies for labeling of FP-tagged proteins with high performance STORM dyes. This approach is robust in that it allows researchers to use FP fusions they already know and are comfortable with.
Fluorescent Protein Labeling
There exist a number of strategies for expression of FPs fused to a protein of interest by a short unstructured linker sequence. Perhaps most simple is transient transfection of cells with the appropriate expression vector. Introduction of the plasmid can easily be performed using either a liposomal delivery system or electroporation, to list a couple of the more popular methods for cultured cells. When working with primary cells or in vivo it may be advisable to use a viral transduction system, such as an adenoviral or lentiviral system. Importantly, viral vectors are generally better choices when attempting to create stably expressing cell lines.
One of the greatest limitations of these expression systems is that the researcher is not labeling native protein, as discussed. More recently, the CRISPR/Cas9 system has become very popular for genome editing applications. Specifically, this allows for the editing of native genes to include an FP, and either homo- or heterozygously. Another option is to knock down expression of the native gene(s). However, both of these methods are time-intensive and need to be extensively validated for any given FP fusion construct.
FPs are valuable labels for live-cell STORM-type imaging as they are genetically expressed, relatively small and not far removed from their fusion partner. Another advantage for FPs in live cells that is frequently overlooked is that they are turned over and replaced relatively quickly, great for imaging the same cell at multiple time points. Perhaps the weakest point of fluorescent proteins for STORM is their low photon output.
SNAP-tag Labeling
The SNAP-tag is a mutant version of the human DNA repair protein O6-alkylguanine-DNA alkyltransferase (hAGT). This protein reacts quickly and bio-orthogonally with benzyl guanine (BG) and BG-functionalized molecules. In simplest terms, a SNAP-tag fusion protein is expressed in the cell with the same techniques used to express FP-fusion proteins. The BG-functionalized STORM dye of choice is added to the cellular environment, reacting with and forming a highly stable covalent link between the dye and the expressed SNAP-tag fusion protein. Many similar labeling systems exist, and can even be used in parallel with SNAP-tag labeling; these include CLIP-tag, HALO-tag, TMP-tag, FLAG-tag, and more. However, the SNAP-tag system has been most thoroughly validated for STORM imaging. Perhaps the greatest strength of such labeling systems is that protein-specific STORM imaging can be performed using synthetic dyes in living systems.
Generally SNAP-tags are most popular for live cell imaging. Cell permeable dye-BG derivatives enable live imaging, though the performance of available cell permeable dye conjugates does not equal that of higher performing cell-impermeable dyes, namely Alexa Fluor 647. This makes it difficult to perform live imaging with the necessary spatiotemporal resolution. A common complaint about SNAP-tags and related systems is that they are natively non-fluorescent, requiring the addition of exogenous (and expensive) functionalized dye for visualization. This makes SNAP-tag expression more difficult and time intensive to screen for compared to FP expression. Additionally, it is recommended that before plating cells in the imaging vessel it should be coated in 2M glycine (dissolved in water) for about 1 hour to prevent nonspecific adsorption of the dye-BG conjugate to the coverglass, which results in prohibitively high background fluorescence, and is exacerbated by imaging in a TIRF configuration.
Bio-Specific Small Molecule Dye Labeling
The Zhuang lab was first to demonstrate the utility of many common lipophilic and membrane stains for live-cell STORM imaging. Such dyes are generally small, charged, cell permeable, and have high affinity for the structure of interest. Importantly, theses stains tend to accumulate at high density, allowing ultra-structural and topological imaging with a very high sampling rate. However, note that the majority of these labels are organelle-specific rather than molecule-specific.
The plasma membrane can be imaged using the lipophilic cyanine dyes DiI (red), DiD (far-red), and DiR (near-infrared). Live-cell STORM imaging of mitochondria can be performed with MitoTracker Orange and Red, both are cationic rosamine dyes. The cationic far-red cyanine dye MitoTracker Deep Red can also be used, and is structurally similar to Cy5 and Alexa Fluor 647. The BODIPY derivatives ER-Tracker Red and LysoTracker Red are applicable towards live-cell STORM of the endoplasmic reticulum and lysosomes, respectively. A number of other lipophilic stains are undoubtedly of utility for STORM, but require validation.
The imaging buffer for these live-cell membrane stains is the culture medium including oxygen scavenging system (e.g. GLOX) and 2% glucose. As with live-cell SNAP-tag imaging this method relies upon the presence of the endogenous thiol glutathione at millimolar concentrations to act as the reducing agent in place of exogenous thiols or other reducing agents. However, live-cell STORM with the addition of small amounts of MEA and other reducing agents has been demonstrated.
There are several bio-specific small molecule dyes that don’t rely upon lipophilic or charge-dependent accumulation. Perhaps most popular are the phalloidin conjugates (e.g. Phalloidin-Alexa Fluor 647). Phallotoxins such as phalloidin and phallacidin are derived from the Amanita phalloides mushroom, and bind exclusively to filamentous actin (F-actin). Note that labeling and imaging with phalloidin conjugates can only be performed in fixed cells not exposed to methanol during staining. Phalloidin-Alexa Fluor 647 specifically has been shown to unbind from actin with high power illumination. STORM has also been demonstrated using labeled wheat germ agglutinin (WGA), a lectin that targets certain glycosylated proteins and lipids found chiefly in the trans-Golgi network.
Purified protein labeled with appropriate dyes can be introduced into the cellular environment for imaging. For example, transferrin and epidermal growth factor (EGF) labeled with Alexa Fluor 647 have been shown to be endocytosed by live cells and imaged using STORM.
Several nuclear stains have been shown to be STORM compatible, including DAPI, several Hoechst dyes, and Vybrant DyeCycle Violet, as discussed. Additionally, YOYO-1 and SYTO-13 are both cyanine dyes specific for DNA, and have been validated for STORM imaging. More fluorescent stains can be used for STORM imaging than many believe, simply requiring experimental validation and optimization. A number of probes for DNA, in particular, have been demonstrated to be of value.
Bio-Orthogonal Click Labeling
A powerful method for labeling modified cellular components is Click chemistry. Specifically, the copper-catalyzed azide-alkyne cycloaddition reaction (CuAAC) is a completely bio-orthogonal method for labeling alkyne functionalized cellular components with dye derivatives containing an azide functional group. Perhaps the best example of this is the labeling of DNA with the alkyne-functionalized thymidine analog EdU. For example, pulsing with EdU and subsequent fixation and labeling with an azide-functionalized dye (e.g. Alexa Fluor 647 azide) allows for super-resolution STORM imaging of newly synthesized DNA. Though such studies have largely been limited to fixed cells due to copper cytotoxicity and probe permeability issues, we expect that newly available copper-free click chemistries will lead to live-cell imaging solutions. Furthermore, click chemistry can be used to label nascent RNA and proteins using commercially available reagents.
PAINT Methods
Point Accumulation for Imaging in Nanoscale Topography (PAINT) microscopy is conceptually similar to STORM, single emission events are isolated in time and space and localized to a sub-diffraction limited area, with enough data points the result is a composite super-resolution reconstruction with drastically increased spatial resolution. However, instead of permanently fixing the fluorophore to the structure of interest for repeated rounds of imaging via typical on-off switching mechanisms, PAINT relies upon the transient binding of probes to the structure of interest. Transient binding occurs on timescales short enough that blinking is not required to isolate single emitters, rather a fluorescence ‘spike’ is observed upon binding due to immobilization of the probe, following which the probe quickly dissociates and/or photobleaches, allowing a fresh probe to take its place. PAINT was originally demonstrated in 2006, using the dye Nile Red to image large unilamellar vesicles. Importantly, a theoretically infinite number of emission events can be used to build a super-resolution reconstruction with PAINT, limited only by the time and reagents available to run the experiment. This is highly significant as PAINT is intrinsically optimized to maximize the molecular sampling rate.
As with STORM, PAINT is limited by the availability of compatible probes. Binding needs to occur on suitable timescales for single molecule localization – with short binding times not enough signal can be integrated for a high quality localization measurement, with long binding times spatiotemporal separation of emitters will become problematic. Another limitation of PAINT is high background fluorescence, even when using strongly fluorogenic probes. This limits the technique to being performed in conjunction with a strong optical sectioning approach such as TIRF or Spinning Disk. PAINT has mostly been demonstrated with lipophilic probes localizing preferentially to certain membrane-bound cellular structures, including lysosomes, the Golgi network, endoplasmic reticulum, plasma membrane, mitochondria, and others, but not associating with specific molecular species. Thus this technique was originally introduced as a ‘topographic’ approach.
Progress is being made with probe specificity by using antibodies that are labeled with short DNA oligomers (oligos), a technique known as DNA-PAINT. These oligos are engineered to have a relatively low affinity to complimentary dye-labeled oligos, which are introduced into the system and bind transiently, allowing for the fluorescently labeled oligo to be continuously replenished for a theoretically unlimited number of rounds of detection and localization and with molecular specificity for the given antigen (Figure 14). Multichannel detection is possible by using primary antibodies against different antigens, with different antibodies conjugated to oligos with different sequences, allowing one to sequentially introduce different oligos with the same high performance dye (known as Exchange-PAINT) or simultaneously using oligos carrying spectrally distinct dyes. This approach yields exceptionally high resolution, little-to-no crosstalk, and no registration errors between different channels due to chromatic aberration (assuming Exchange-PAINT). Exchange-PAINT has been demonstrated for 10-color super-resolution imaging, with sub-10 nm average localization precision. The recent commercialization of DNA-PAINT by Ultivue Inc. is promising for standardization of this technique.
Multi-labeling Strategies
With so many different available labeling strategies, it is very common to mix and match approaches in the same experiment. The primary limitation in such cases is not the ability to perform different types of labeling of the same specimen (e.g. mixing antibody-based detection with SNAP-tag labeling), but identifying ‘compromise buffers’ that work for fluorophores with different on-off chemistries. While this generally isn’t as problematic when using the same type of fluorophore (e.g. all synthetic dyes), it can be prohibitive when mixing synthetic dye and FP labels. FPs have a tendency to perform poorly in reducing and oxidizing buffer systems – as required by such high performance dyes as Alexa Fluor 647. However some varieties, such as mEos2, are more resistant to such environmental conditions than others. In any case, if one needs to image an FP in a reducing buffer system, keep the concentration of the reducing agent as low as possible (less than or equal to 10mM). One promising as yet unexplored avenue is to mix FP and PAINT labeling methods, as both are performed in physiological buffer.
Conclusions
STORM imaging provides unprecedented resolution, breaking what were once believed to be fundamental limits to light microscopy that stood for hundreds of years. Despite its great utility, many STORM imaging experiments fail not because of imaging technique, but due to sample preparation. Perhaps more so than any other fluorescence microscopy, STORM requires stringent attention to choice of fluorophore, labeling method, and buffer choice for successful imaging - in this work we’ve taken special care to both review the state of STORM and address these concerns. It should be noted that advances in the field of single molecule localization are occurring rapidly, so investigators interested in the application of the technique should be aware of emerging solutions, such as PAINT methods for ultra-high sampling rates. Though extensive information is provided here, it should not be treated as comprehensive, but rather as a starting point for STORM experimental design.
Contributing Authors
John R. Allen, Joel S. Silfies and Stanley A. Schwartz - Nikon Instruments, Inc., 1300 Walt Whitman Road, Melville, New York, 11747.
Michael W. Davidson - National High Magnetic Field Laboratory, 1800 East Paul Dirac Dr., The Florida State University, Tallahassee, Florida, 32310.
References
- D.Żurek-Biesiada et al.,Localization microscopy of DNA in situ using Vybrant(®) DyeCycle™ Violetfluorescent probe: A new approach to study nuclear nanostructure at singlemolecule resolution. Exp Cell Res 343, 97-106 (2016).
- K. Prakash et al., Superresolution imagingreveals structurally distinct periodic patterns of chromatin along pachytenechromosomes. Proc Natl Acad Sci U S A 112, 14635-14640 (2015).
- D.Żurek-Biesiada et al.,Quantitative super-resolution localization microscopy of DNA in situ usingVybrant® DyeCycle™ Violet fluorescent probe. DataBrief 7, 157-171 (2016).
- A. T.Szczurek et al., Singlemolecule localization microscopy of the distribution of chromatin usingHoechst and DAPI fluorescent probes. Nucleus 5, 331-340 (2014).
- C. Flors,DNA and chromatin imaging with super-resolution fluorescence microscopy basedon single-molecule localization. Biopolymers 95, 290-297 (2011).
- C. Flors,C. N. Ravarani, D. T. Dryden, Super-resolution imaging of DNA labelled withintercalating dyes. Chemphyschem 10, 2201-2204 (2009).
- C. Flors,Photoswitching of monomeric and dimeric DNA-intercalating cyanine dyes forsuper-resolution microscopy applications. PhotochemPhotobiol Sci 9, 643-648 (2010).
- G. T.Dempsey, J. C. Vaughan, K. H. Chen, M. Bates, X. Zhuang, Evaluation offluorophores for optimal performance in localization-based super-resolutionimaging. Nat Methods 8,1027-1036 (2011).
- S. van deLinde et al.,Photoinduced formation of reversible dye radicals and their impact onsuper-resolution imaging. Photochem Photobiol Sci 10, 499-506 (2011).
- S. A. Jones, S. H. Shim, J.He, X. Zhuang, Fast, three-dimensional super-resolution imaging of livecells. Nat Methods 8,499-508 (2011).
- A. Benke, S. Manley,Live-cell dSTORM of cellular DNA based on direct DNA labeling. Chembiochem 13, 298-301 (2012).
- A. Löschberger et al., Super-resolution imagingvisualizes the eightfold symmetry of gp210 proteins around the nuclear porecomplex and resolves the central channel with nanometer resolution. J Cell Sci 125, 570-575 (2012).
- M. Heilemann, S. van deLinde, A. Mukherjee, M. Sauer, Super-resolution imaging with small organicfluorophores. Angew Chem Int Ed Engl 48, 6903-6908 (2009).
- J. Fölling et al., Fluorescence nanoscopy byground-state depletion and single-molecule return. NatMethods 5, 943-945 (2008).
- M. Lehmann, G. Lichtner, H.Klenz, J. Schmoranzer, Novel organic dyes for multicolor localization-basedsuper-resolution microscopy. J Biophotonics 9, 161-170 (2016).
- T. Klein et al., Live-cell dSTORM withSNAP-tag fusion proteins. Nat Methods 8, 7-9 (2011).
- S. H. Shim et al., Super-resolutionfluorescence imaging of organelles in live cells with photoswitchablemembrane probes. Proc Natl Acad Sci U S A 109, 13978-13983 (2012).
- V. N. Belov, M. L. Bossi, J.Fölling, V. P. Boyarskiy, S. W. Hell, Rhodamine spiroamides for multicolorsingle-molecule switching fluorescent nanoscopy. Chemistry 15, 10762-10776 (2009).
- R. Galland et al., 3D high- andsuper-resolution imaging using single-objective SPIM. Nat Methods 12, 641-644 (2015).
- Z. Zhang, S. J. Kenny, M.Hauser, W. Li, K. Xu, Ultrahigh-throughput single-molecule spectroscopy andspectrally resolved super-resolution microscopy. NatMethods 12, 935-938 (2015).
- G. Lukinavičius et al., A near-infrared fluorophorefor live-cell super-resolution microscopy of cellular proteins. Nat Chem 5, 132-139 (2013).
- S. N. Uno et al., A spontaneously blinkingfluorophore based on intramolecular spirocyclization for live-cellsuper-resolution imaging. Nat Chem 6, 681-689 (2014).
- M. J. Rust, M. Bates, X.Zhuang, Sub-diffraction-limit imaging by stochastic optical reconstructionmicroscopy (STORM). Nat Methods 3, 793-795 (2006).
- M. Bates, B. Huang, G. T.Dempsey, X. Zhuang, Multicolor super-resolution imaging with photo-switchablefluorescent probes. Science 317, 1749-1753 (2007).
- J. C. Vaughan, G. T. Dempsey,E. Sun, X. Zhuang, Phosphine quenching of cyanine dyes as a versatile toolfor fluorescence microscopy. J Am Chem Soc 135, 1197-1200 (2013).
- N. Olivier, D. Keller, V. S.Rajan, P. Gönczy, S. Manley, Simple buffers for 3D STORM microscopy. Biomed Opt Express 4, 885-899 (2013).
- R. Wombacher et al., Live-cell super-resolutionimaging with trimethoprim conjugates. Nat Methods 7, 717-719 (2010).
- S. van de Linde, R. Kasper,M. Heilemann, M. Sauer. (Appl Phys B, 2008), vol. 93, pp. 725-731.
- J. Vogelsang, T. Cordes, C.Forthmann, C. Steinhauer, P. Tinnefeld, Controlling the fluorescence ofordinary oxazine dyes for single-molecule switching and superresolutionmicroscopy. Proc Natl Acad Sci U S A 106, 8107-8112 (2009).
- A. Lampe, V. Haucke, S. J.Sigrist, M. Heilemann, J. Schmoranzer, Multi-colour direct STORM with redemitting carbocyanines. Biol Cell 104, 229-237 (2012).
- E. Betzig et al., Imaging intracellularfluorescent proteins at nanometer resolution. Science 313, 1642-1645 (2006).
- S. Wang, J. R. Moffitt, G. T.Dempsey, X. S. Xie, X. Zhuang, Characterization and development ofphotoactivatable fluorescent proteins for single-molecule-basedsuperresolution imaging. Proc Natl Acad Sci U S A 111, 8452-8457 (2014).
- F. V. Subach, G. H.Patterson, M. Renz, J. Lippincott-Schwartz, V. V. Verkhusha, Bright monomericphotoactivatable red fluorescent protein for two-color super-resolutionsptPALM of live cells. J Am Chem Soc 132, 6481-6491 (2010).
- F. V. Subach et al., Photoactivatable mCherryfor high-resolution two-color fluorescence microscopy. Nat Methods 6, 153-159 (2009).
- M. S. Gunewardene et al., Superresolution imaging ofmultiple fluorescent proteins with highly overlapping emission spectra inliving cells. Biophys J101, 1522-1528 (2011).
- H. Chang et al., A unique series ofreversibly switchable fluorescent proteins with beneficial properties forvarious applications. Proc Natl Acad Sci U S A 109, 4455-4460 (2012).
- T. Brakemann et al., A reversiblyphotoswitchable GFP-like protein with fluorescence excitation decoupled fromswitching. Nat Biotechnol29, 942-947 (2011).
- J. Fuchs et al., A photoactivatable markerprotein for pulse-chase imaging with superresolution. Nat Methods 7, 627-630 (2010).
- V. Adam et al., Rational design ofphotoconvertible and biphotochromic fluorescent proteins for advancedmicroscopy applications. Chem Biol 18, 1241-1251 (2011).
- H. Shroff et al., Dual-color superresolutionimaging of genetically expressed probes within individual adhesion complexes.Proc Natl Acad Sci U S A104, 20308-20313 (2007).
- S. A. McKinney, C. S. Murphy,K. L. Hazelwood, M. W. Davidson, L. L. Looger, A bright and photostablephotoconvertible fluorescent protein. Nat Methods 6, 131-133 (2009).
- M. Zhang et al., Rational design of truemonomeric and bright photoactivatable fluorescent proteins. Nat Methods 9, 727-729 (2012).
- M. G. Paez-Segala et al., Fixation-resistantphotoactivatable fluorescent proteins for CLEM. NatMethods 12, 215-218, 214 p following 218 (2015).
- L. Nahidiazar et al., Optimizing Imaging Conditions for Demanding Multi-Color SuperResolution Localization Microscopy. PLoS ONE 11: e0158884.doi:10.1371/journal.pone.0158884 (2016).