What is the resolution

Stimulated Emission Depletion (STED) microscopy is a cornerstone of superresolution imaging, enabling researchers to visualize structures at the nanoscale by overcoming the diffraction limit of light. Unlike conventional confocal microscopy, which is limited to a lateral resolution of about 200 nm, STED can achieve resolutions down to 20 nm – and even beyond – by exploiting the physics of stimulated emission to spatially confine fluorescence. While this requires some effort, e.g., using optimized dyes, a medium resolution increase in the range of 50-100 nm can readily be achieved with standard protocols.

of a STED microscope?

At the heart of STED microscopy lies the concept of temporarily switching-off fluorescence at the perimeter of the excitation spot using a second, donut-shaped laser beam (here is how to make a STED donut). This beam forces excited molecules back to the ground state without fluorescence, except at the very center of the donut where the intensity is zero. It is only there that the fluorophores are allowed to fluoresce. The result is a much smaller effective point spread function (PSF) (what’s this?), and thus, higher resolution. You can read more about how STED works here.

The resolution R in STED microscopy can be approximated by the formula

\(\textit{R}\approx \frac{\lambda}{2NA} \frac{1}{\sqrt{1+\frac{I}{I_{SAT}}}} \)

where

• λ is the wavelength of the depletion laser,
• NA is the numerical aperture of the objective,
• I is the STED laser intensity,
• I_SAT is the saturation intensity of the fluorophore, characteristic to each dye.

This equation highlights two critical factors: the STED laser intensity and the properties of the fluorescent molecules.

When the STED laser intensity is zero, the square root amounts to one and the resolution is just that of a diffraction-limited microscope ( \(\frac{\lambda}{2NA} \) ). When I is much higher than the saturation intensity I_SAT , the value of the square root is large and the resolution R can become infinitely small in theory.

So, among other things, resolution hinges on how much larger the STED intensity can become compared to the saturation intensity of the dye. Preferably, one uses dyes with a small I_SAT to maximize the resolution while keeping STED intensities low.

You see, the choice of fluorescent dye is pivotal. While STED-optimized dyes like abberior STAR RED offer exceptional performance, many commonly used dyes such as Alexa Fluor 488, Alexa Fluor 532, and even fluorescent proteins like GFP or YFP can also yield substantial resolution improvements. These dyes are widely used in standard immunofluorescence protocols and at the same time are compatible with STED systems equipped with standard depletion lasers (e.g., 592 nm or 775 nm).

For these general-purpose dyes, STED microscopy can still achieve lateral resolutions in the 50-100 nm range, making it a powerful upgrade to conventional imaging without requiring any changes to labeling protocols.

For any given dye, increasing the STED laser intensity enhances resolution by more effectively shutting of peripheral fluorescence. However, this comes with trade-offs:

• Photobleaching: Higher intensities can degrade some fluorophores faster.
• Phototoxicity: Live-cell imaging becomes more challenging.
• Re-excitation: At high STED intensities, especially with certain dyes, the depletion beam can inadvertently re-excite fluorophores, leading to background fluorescence and reduced image contrast.
• Signal-to-noise ratio: STED removes low-resolution photons in the focal plane, but not outside of it. If the sample has high background fluorescence – such as from out-of-focus regions or autofluorescence – this background can overwhelm the desired high-resolution signal, reducing image clarity.

Thus, optimizing laser power is a balancing act between resolution, signal, and sample preservation.

To push resolution below 50 nm, several advanced strategies can be employed:

Use of STED-optimized dyes

Dyes like abberior STAR RED are engineered for high photostability, low re-excitation, and efficient depletion (small I_SAT). These dyes tolerate higher STED intensities and enable resolutions down to 30 nm or better. Importantly, labeling procedures remain unchanged, allowing seamless integration into existing workflows.

Adaptive illumination techniques

Adaptive methods such as RESCUE or DYMIN dynamically modulate the STED beam based on local fluorophore density. This reduces unnecessary photobleaching, phototoxicity, and enhances signal and resolution, especially in sparse regions. These techniques are highly useful for live-cell imaging, when targeting delicate structures, or for pushing STED resolution to 20 nm and below. Read more about adaptive illumination here.

FLIM-STED

Fluorescence lifetime imaging or FLIM, such as abberior’s TIMEBOW, leverages differences in fluorescence lifetimes to distinguish between fluorophores or to enhance contrast. It can also improve STED resolution by refining the spatial localization of emission events. This is because the donut-shaped STED beam shortens the fluorophores’ lifetime in a spatially dependent way and this spatial modulation of lifetime encodes positional information, allowing FLIM-STED to extract finer details.

Array detection and background removal

Technologies like abberior’s MATRIX detector use spatially resolved array detection to distinguish in-focus signal from out-of-focus background. This allows for physical measurement and removal of background, significantly improving contrast and effective resolution, especially in thick or autofluorescent samples.

For best results it’s always preferable to provide perfect physical conditions, i.e., to optimize sample, dye, and microscope. Regardless of the imaging conditions, however, deconvolution remains a powerful post-acquisition tool to further enhance resolution. Algorithms like abberior’s TRUESHARP use knowledge of the system’s PSF to sharpen images, remove background, and reveal fine structural details that may be obscured in raw data. Deconvolution can be applied to both confocal and STED datasets, making it a universally beneficial step.

STED microscopy offers a flexible and scalable approach to superresolution imaging. While resolution is fundamentally governed by the interplay between dye properties and STED laser intensity, significant improvements can be achieved even with widely used dyes like Alexa Fluor 488 or GFP. For researchers seeking even higher resolution, options include switching to STED-optimized dyes, employing adaptive illumination, or integrating FLIM. In all cases, post-processing with deconvolution tools like TRUESHARP can further refine image quality. Together, these strategies make STED a versatile platform for exploring the nanoscale architecture of biological systems.

As you attempt to image deeper into a tissue specimen, different light phenomena dominate as the cause of aberrations.

Let’s start at the top. Table 1 gives you an overview of this discussion.

From a few to about 20 µm, confocal and multi-photon microscopy can produce clear images because they restrict the emission or the excitation light involved in image generation to just the focal plane (see further below). That is, they minimize contaminating signals from above or below the focal plane.

So, how do confocal and multi-photon microscopy work?

Confocal microscopy minimizes the amount of background fluorescence collected from the out-of-focus region of a sample by forcing emissions from fluorophores through a pinhole to filter out unfocused signals (more on optical sectioning here). This method gathers light only from sections very close to the focal plane. Then, through stacking of multiple optical sections, a three-dimensional rendition of the sample can be reconstructed.

Confocal microscopy works well with mildly scattering specimens of up to 200 µm thick. However, even with the targeted illumination of a confocal microscope, increasing imaging depth or sample density thwarts the pinhole effect: aberrations cause photons from the focal plane to miss the pinhole and photons from outside the focal plane to erroneously pass through. As a result, signal intensity decreases, and background noise increases as you image deeper.

Multi-photon microscopy was developed in response to the limited tissue penetration of confocal microscopy and performs best at millimeter depths in strongly scattering samples (Fig. 1). Like the pinhole of a confocal microscope, a multi-photon microscope sharpens the detected signal to 1/z² relative to the depth distance z. However, instead of using a pinhole to restrict emission light, a multi-photon microscope restricts fluorophore excitation to a point close to the focal plane, leaving the planes above and below unilluminated.

To do so, it excites a fluorophore with multiple low-energy photons that arrive simultaneously (within the span of a few femtoseconds) to bridge the same energy gap as a single higher-energy photon (Fig. 2). The wavelengths used in multi-photon microscopy have about half the energy of excitation beams used in confocal microscopy, tending to the red to near-infrared range, and one laser can be used to excite fluorophores with different emission wavelengths simultaneously. Optical sectioning is thus enabled by controlled excitation, and the emitted signal is projected directly and efficiently onto the system’s detector.

As we mentioned before, the first 20 µm of a specimen are well-served by confocal and superresolution microscopy. If sub-diffractive resolution is important, a STED microscope produces remarkably clear images. Given those two good options, investing in a multi-photon microscope is unnecessary. The high-energy, short-pulse lasers of a multi-photon setup are particularly expensive.

Deeper than 20 µm, aberrations caused by light refraction at the interface between media with different refractive indices become problematic and require correction. A correction collar on the objective of a confocal or multi-photon microscope can correct the aberrations by redirecting light. However, the collars can only be set to a fixed value that spans only a few microns on the z-axis, limiting corrections to just that depth range. Adapting the correction collar during a focus scan is not possible. Thus, this technology will never yield a z-slice image aberration-corrected from top to bottom.

With multi-photon microscopy, there is extensive photobleaching at these depths due to the high excitation doses needed to produce detectable fluorescence.¹ The bleaching outweighs any benefit, so a multi-photon microscope is not an efficient solution for imaging tissues at these depths.

However, once you move deeper than 200 µm, indeterministic light scattering dominates in causing image distortion. Here, correcting refracted light no longer improves images. This is where multi-photon microscopy excels. It offers exceptional optical sectioning in scattering media, particularly in lipid-rich specimens like brain tissue, creating clear images with reduced background noise. Also, using low-energy excitation reduces scattering in deep tissues. Thus, even with high photobleaching, multi-photon microscopy becomes the method of choice for specimens from 200 µm to a couple of millimeters thick.

Looking back at our stroll along the z-axis, we’re left with a noticeable gap in creating crisp, detailed images throughout a thicker specimen. Up to 20 µm, standard but robust confocal and STED microscopy get the job done. From about 200 µm to a couple of millimeters, multi-photon microscopy is the best option to handle pervasive light scattering. But in between? Are we left with the painstaking option of correcting a very limited span of 10 or maybe 20 µm in z with a correction collar objective while the rest of the image stays blurred? 

It would be great if adjustments were made by a dynamic system programmed to adapt to aberrations as you move along the z-axis and quickly redirect light accordingly.

Equipped with RAYSHAPE, a confocal or a STED microscope produces bright, high-resolution images as you move through sample depths of a few micrometers to about 200 µm. An array of 140 autonomous actuators adjusts a deformable mirror to optimize the focus quality of the excitation beam and compensates for aberrations in the emitted light as the focal plane of the microscope moves through a specimen. Because the actuators act upon the mirror’s surface in milliseconds and with great precision, correction is dynamic, and bright and crisp images are guaranteed even in the deeper layers of the sample. This innovative solution enhances image quality and minimizes stress on the sample and dye.

So, the bottom line is: if your specimen is 200 µm thick or less, RAYSHAPE-enhanced confocal or STED microscopy brings every depth into focus. STED, of course, also delivers superresolution for outstanding detail. For specimens thicker than 200 µm and up to a few millimeters, a multi-photon microscope reduces scatter to produce clearer results, albeit at the cost of increased photobleaching.

Get perfect focus and exceptional resolution for your images.

Find out more about dynamic aberration correction with RAYSHAPE.

¹ Singh, A. et al. 2015. Comparison of objective lenses for multi-photon microscopy in turbid samples. Biomedical Optics Express 6: 3113.

Remember that resolution is the power to distinguish two closely positioned points. In the case of ExM and superresolution methods like STED, PALM/STORM, and MINLFUX, those points are fluorescent labels that cluster at heterogeneously spaced structures in relevant areas of a sample. Each method applies a different principle to resolve this crowding. Table 1 provides an overview of how these three methods compare along different parameters.

Let’s delve a bit deeper into ExM to explore how it works and what it can and cannot do.

In ExM, the specimen is embedded in a hydrogel composed of a dense matrix of cross-linked polymers. The nanometer-scale polymers form a fine mesh far below the size scale of cells, wrapping around and traversing between biomolecules. Upon adding water, the hydrogel vastly expands, pulling apart molecules. However, the polymer mesh preserves the relative molecular organization. That is, the expansion is isotropic, at least under ideal circumstances and up to a certain limit (see below). As an analogy, think of drawing a picture on a lab glove and then inflating it (who hasn’t done this?). The ink particles move apart and the picture becomes larger (Fig. 1).

The expansion protocol involves more steps than “just add water”. To date, different forms of ExM have been developed to address different imaging needs, but all follow a basic workflow. The specimen is first fixed to eliminate variation and degradation from enzyme activity and tissue decay. Through a series of reagent treatments, cells in the sample are permeabilized and embedded in the hydrogel while biomolecules are anchored to the polymers. The polymer-embedded sample is then enzymatically or mechanically digested to break up macromolecular structures prior to expansion with water (Fig. 2). Fluorescent labels for visualization can be added before or after expansion.

Details of the protocol vary by application. In protein-retention expansion microscopy (ProExM), specimen proteins are anchored to the polymer, whereas expansion fluorescent in situ hybridization (ExFISH) anchors nucleic acid molecules. Iterative expansion microscopy (iExM) augments expansion by applying a second swellable hydrogel to the space opened by a first expansion. Each ExM variation achieves a different expansion factor and, thus, resolution.  An upper bound lies currently with iExM, where an effective resolution down to 15 nm is attainable using a 300 nm diffraction-limited objective lens.

New techniques for ExM continue to emerge. For example, click-ExM integrates click labeling with biotin and staining with fluorescently labeled streptavidin to allow visualizing lipids, glycans, and other small molecules in addition to proteins, DNA, and RNA. A development called Magnify eliminates the anchoring step in sample preparation with a gel that retains nucleic acids, proteins, and lipids universally.

The plethora of ExM protocol variations shows: it’s all somewhat complicated.

ExM promises to overcome many practical and scientific hurdles of superresolution techniques. To begin, ExM uses inexpensive and common reagents. ExM is also instrument agnostic. By increasing the space between biomolecules to the point where even a diffraction-limited microscope can discriminate them, ExM works on any microscope and shifts a range of commonly used imaging techniques into the realm of superresolution. Thus, ExM can be integrated into existing laboratory infrastructures with a minor financial investment.

Second, the expanded specimen is a more accessible imaging environment. Expanded with water, samples also become completely transparent, which eases and accelerates volumetric imaging. Finally, when staining samples after expansion, the augmented space accommodates more labeling molecules, thereby improving multiplexing capacity.

While a “superresolution technique” using common reagents and conventional microscopes sounds great, there is, of course, a catch – in fact, there is more than one.

The obvious one is ExM’s incompatibility with live-cell imaging. Cells and tissues must be fixed to withstand the permeabilization, polymerization, and expansion process, not to mention that dragging apart biomolecules means that they won’t interact anymore. The same expansion also leads to a diluted signal – often up to 50% – as labels added prior to expansion end up occupying a wider space or may be damaged during separation.

At first glance, sample preparation for ExM looks straightforward. However, it is a finicky multi-step process that can take up to 4 days, which substantially lowers sample throughput. Moreover, preparing samples for ExM requires a trained hand to carry out reliably. Successful expansion depends on an even, mesh-like distribution of the polymer chains between and around biomolecules and a uniform homogenization of the specimen, so that no structure resists expansion. That proficiency comes only from experience. In the hand of novices, expansion is rarely uniform in all four directions, which introduces distortions in a final image. Especially in iExM, a 5–10 nm error caused by anisotropic expansion makes very precise nanoscale imaging impossible. In fact, expansion may vary between organelles within a single cell or even between sub-structures of the same organelle.¹

To visualize the problem of anisotropic expansion, it helps to recall the above example of the picture drawn on a lab glove: the lab glove does not expand evenly when inflated, and those parts of the picture close to the fingers and the opening will be enlarged less than the parts in the middle of the glove.

Hence, validating isotropic expansion is mandatory in ExM and the possibility of distortions should always be kept in mind when analyzing ExM images.

Such problems are unknown to superresolution methods like for example STED. As STED relies on pure physics, the generated raw data is final and mistakes in the imaging process result in no image at all.

Handling ExM specimens is another issue. The transparent, unstable, and jelly-like sample is difficult to manipulate and must be protected against evaporation. That consistency and high water content also make it impossible to add anti-bleaching agents or use anything but water objectives. Then there is the sample size: ExM samples are large, which makes transfer from, for instance, a petri dish to the microscope slide challenging. Also, a larger sample depth means that, in order to image the same structure, you have to image deeper inside the sample than you would need to in a non-expanded specimen, and the difference may easily be 10-fold. Index mismatches have a stronger effect; deeper imaging equals more aberrations and in consequence lower signal and resolution, and the working distance of the objective lens becomes an issue, as well.

ExM is an imaging solution that is potentially accessible to any lab with conventional microscopes. It comes, however, with a lengthy onboarding as the operator learns the “art” of evenly and reliably expanding samples. Moving from confocal to expansion is much more challenging than moving from confocal to, for example, STED. Furthermore, expansion excludes the option of live-cell imaging or tracking the movement and interactions that define life.

Investing instead in a super-resolution technology like STED or MINFLUX is the optimal choice for routine work. Sample preparation with a low risk of distortion, live specimen compatibility, intuitive handling, and the temporal resolution to see life in action are just some of the advantages that make technologies like STED and MINFLUX attractive. And acquiring superresolution microscopy doesn’t have to break the bank. STEDYCON, a full confocal and STED microscope the size of a shoebox, fits on every microscope frame with a camera port and is the very definition of ease-of-use at an affordable price.

So, how will you expand into superresolution microscopy?

¹ Büttner M, Lagerholm CB, Waithe D, Galiani S, Schliebs W, Erdmann R, Eggeling C, Reglinski K. Challenges of Using Expansion Microscopy for Super-resolved Imaging of Cellular Organelles. Chembiochem. 2021 Feb 15;22(4):686-693. doi: 10.1002/cbic.202000571.

Conventional light microscopy has been around for hundreds of years, has been and still is of great service to science . It is capable of peeking into organisms and even cells and resolves details invisible to the naked eye. But there is something standing in its way and preventing it from resolving even smaller details: the diffraction limit. It confines the resolution of even the most powerful conventional light microscopes such as confocal systems to roughly half the wavelength of the light used, which corresponds to roughly 200 to 400 nm. Any two points closer together than this cannot be separated by conventional light microscopy because their point spread functions (PSF) overlap and they appear as a single light source.

If you seek a thorough explanation of what precisely the PSF is, what it has to do with resolution, and the physics determining the diffraction limit, check out this article in our knowledge base.

When Stefan Hell invented STED microscopy more than twenty years ago, he of course did not change the laws of physics. He merely outsmarted them. His idea: don’t allow all fluorophores in the excitation spot to emit light at the same time to be able to distinguish them. And this is how he did it: in STED, like in confocal microscopy, fluorophores in the sample are excited by a laser pulse, which generates a conventional, diffraction-limited focus. So far, so ordinary. The trick is using a second laser pulse. This one is of a longer wavelength than the first and transiently knocks back all fluorophores it reaches to their ground state, before they can emit any photons. This effect is called stimulated emission depletion, or STED for short. So here is a tool that allows the targeted on- and off-switching of fluorophores.

As a side note: the general concept of switching fluorophores on and off is something STED shares with other superresolution methods like PALM/STORM. What sets STED apart is that it reaches superresolution by applying pure physics and generates superresolved raw data. So there is no need for post-processing in STED (but it’s possible, of course). In contrast, PALM/STORM and also SIM require extensive processing of the acquired data to generate a meaningful image at all.

But back to the STED principle: As long as the STED beam has the same shape and covers the same area as the excitation beam, all excited fluorophores are pushed back to the ground state and nothing is won. The gain in resolution beyond the diffraction limit only comes with a manipulation of the STED beam. It’s modulated so that its laser intensity is zero in the center – effectively, it has a hole in its middle, making it look like a donut (read here how it’s done). That way, all the fluorophores in the outer regions of the focal spot are de-excited while the ones in the center, where there is no STED light, continue to normally fluoresce as if nothing had happened. The result is a photon-emitting focal spot much smaller than the diffraction limit.

You may now argue that the donut itself is still limited by diffraction – and you are right! However, a second effect comes into play here, namely that the STED effect saturates. There is no such thing as switching a molecule “more off” than it already is. This is the nonlinearity that fundamentally breaks the diffraction limit, confining the region from which fluorescence is allowed to the very center of the donut.

But what about detection? It is diffraction-limited, as well, isn’t it? Indeed it is. But with the donut confining fluorescence to a small region, we know that all detected photons must necessarily originate from there. And as the location of this spot is precisely known by the scanner position, we precisely know the location of the molecule, even if its image on the detector is again diffraction limited.

Generally, higher STED laser power means higher resolution. But the relationship is not linear: Even modest STED powers immediately result in a large increase in resolution. In theory, increasing STED power will eventually result in an infinitely small fluorescing area (again, read our article on resolution for the why), which would mean that it’s just a matter of laser power to achieve single-digit nanometer resolution. This logic obviously cannot stand up to practice, where increasing laser power will rather sooner than later lead to secondary unwanted effects, such as photobleaching. Routinely, STED microscopes reach a standard resolution of about 30 nm in biological samples.

This resolution can be further improved to less than 20 nm by sophisticated modules like abberior’s FLEXPOSURE and TIMEBOW: FLEXPOSURE adaptive illumination intelligently reduces the light dose on the sample by switching the lasers off where there is no structure to be detected. The saved photon budget can either be invested in resolution or more images of the same region in a time-lapse experiment. TIMEBOW provides fluorescence lifetime information. As the signal’s lifetime depends on the distance from the donut center (lifetime drops as de-excitation increases), it encodes spatial information that can be used to increase resolution.

Summed up, the secret of STED is essentially little more than adding a second, donut-shaped laser beam to your system that de-excites fluorophores and narrows down the region in which fluorescence is allowed to happen to a sub-diffraction limit size.

However, building a STED microscope that reaches maximum resolution while keeping the light burden low requires sophisticated engineering on both the hardware and software level – something we at abberior are quite proud of, to be honest. A central prerequisite is, for example, that the hole in the STED donut is a true intensity zero, and that the beam size can be flexibly adjusted to the back aperture of various objective lenses. This is something not every STED microscope is capable of.

Now you know how STED works on the conceptual level. But it does not end here. If we sparked your interest, you may continue to browse our knowledge base and learn about the set-up of a STED microscope and why operating a STED system is not any more difficult than using a confocal, or how the hole comes into the donut.

And then there obviously is the question most relevant for biologists: how does the resolution gain impact my research? What do I see with STED that I don’t see with other imaging methods?

Superresolution with stimulated emission depletion (STED) microscopy – also called STED nanoscopy – relies on transiently silencing fluorophores in the outer regions of the excitation spot, thereby shrinking the zone from which molecules are allowed to fluoresce to sub-diffraction dimensions (more on how STED works here). In this scheme, the STED beam acts as the light switch. To do its job, it has to have a specific shape: zero intensity in the center surrounded by high intensity light all around – yes, it’s a yummy donut!

Making a real sweet donut in the kitchen is simple (provided you have a good recipe), you simply need the right baking sheet whose cavities define the donut’s shape. But how do you tell a light beam to become a donut? Obviously, there is no shaping pan for a light beam. Or is there?

Before we answer this question, we need to take a step back and remind ourselves of the physical properties of light. As you know, light propagates as a wave, and like any wave its properties can be described by its wavelength or frequency, amplitude, direction, and phase. We will focus on the phase here. The phase describes when and where a wave reaches a certain state (e.g., peak or valley). The phase can be measured by fractions of the wavelength λ.

When two waves meet, the difference between the waves’ phases determines the amplitude of the resulting new wave. If for example two interfering waves are completely in phase – meaning that their peaks and valleys align perfectly – their amplitudes add up in constructive interference. If, in contrast, one of the waves is shifted by half a wavelength (λ/2), peaks and valleys cancel each other out, and interference is destructive. In the special case of two waves with precisely the same amplitude, destructive interference is complete: no light, zero intensity.

By now you probably have guessed where we are headed.

Creating a donut-shaped STED beam is all about cleverly manipulating the beam’s phase to achieve complete destructive interference when the light waves meet in the focus.

A convenient way to change a beam’s phase is by placing a phase plate in its path, a device that retards waves passing through. In its simplest form, this can be a flat piece of glass that takes longer for the light wave to pass than a slab of air with the same thickness.

So with a phase plate we have the means to change the phase of the STED beam. But does that turn our beam into a donut? Not quite. After all, we have to create a phase difference to produce destructive interference. Uniformly shifting the phase of a beam doesn’t do the job.

So we obviously need to change the phase not of the entire wavefront but only of parts of it. The easiest way to achieve this is by using a phase plate that only covers half of the beam. By choosing just the right glass thickness, we can thus introduce a phase difference of exactly half a wavelength between opposite sides of the beam. When the waves from both sides meet in the focal plane, this results in complete destructive interference on a line parallel to the edge of the phase plate. But wait, we need a donut, not a hot dog bun, don’t we…?

To move from line to donut, the best option is a vortex phase plate. Essentially, this is a rotationally symmetric version of the hotdog bun phase plate. It’s a miniature spiral staircase with helically increasing thickness that introduces a spiral phase shift between 0 and a full wavelength over the wavefront. Any two waves on opposite sides of the beam get a mutual offset of precisely half a wavelength and cancel each other out in the focus. And eventually we have a perfect STED donut…

… but only in 2D. Resolution along the optical axis still is diffraction-limited in this setting as the donut only extends laterally in the focal plane.

Luckily there are more options of how to construct a phase plate to manipulate a light beam. For 3D STED, the plate takes an annular shape, with the outer ring being thicker than the central disk, introducing a phase shift of precisely half a wave between them. Waves going through the two different parts of the annular phase plate thus focus to two different, but perfectly aligned point spread functions (PSF) in the focal plane (what’s a PSF?): a larger one resulting from the light passing the inner disc, and a smaller one generated by the light passing the outer ring. Due to the phase shift between the two, the smaller PSF essentially stamps a hole into the larger one.

And so, we finally get a three-dimensional STED donut – which rather is a hollow sphere of light with a central intensity zero – that shrinks the effective excitation spot to sub-diffraction extent in all three dimensions. 3D STED achieved! And all microscopists can live happily ever after – a fairy tale would end here.

But reality doesn’t. For one, obtaining isotropic resolution in all three dimensions with a phase plate setup remains a challenge. Then there are aberrations that give the 3D donut a hard time as they quickly fill the donut center with STED light, eating up fluorescence signal. And phase plates are not adjustable to the varying back apertures of different objective lenses.

You see, our story demands to be continued. Watch out for the sequel, in which you will learn about a different way to make a STED donut and how this solves several problems at once.

In spite of the revolutionary increase in resolution, the setup of a microscope capable of stimulated emission depletion – better known as STED – doesn’t overthrow the well-established design of a confocal system. In fact, the general principle of how a STED microscope elicits fluorescence in the sample, filters out-of-focus light, and generates an image based on the detected signal is identical. But state-of-the-art STED is more than topping off a confocal with some additional components – it also places special demands on design and optical quality.

What’s in a confocal? We have at least one laser (usually more) that generates a light beam which is focused on the specimen by means of an objective lens. A scan unit moves this focal spot over the sample, line by line. Fluorescence emitted by the sample is collected by the objective lens and sent to a detector. The detector is located behind a pinhole whose function is to provide optical sectioning. In between are dichroic mirrors separating incoming from outgoing light and guiding the way.

Note that this is the very basic setup. There are countless variations of a confocal microscope on the market which all deviate from what is described here in one way or another.

All the elements listed above can be found in a STED microscope, as well, with the very same function and essentially the same arrangement. But there is something on top, something that changes the outcome from diffraction-limited to superresolved. Most importantly, we have at least one additional laser source generating the STED beam. This beam passes a device that turns it into a donut with a central intensity-zero. Then it is coupled into the main light path by another dichroic mirror.

Let’s take a closer look at those components transforming a conventional confocal microscope into a superresolving STED machine.

The laser providing the STED beam is red-shifted relative to the excitation lasers and should be pulsed to reduce the light burden on the sample.

On its way, the shape of the STED beam has to be transformed into a light distribution with an intensity zero surrounded by light. In the most basic setup, a phase plate takes care of this. This device delays portions of the beam slightly (by half a wavelength) such that rays interfere destructively in the focus instead of constructively, as is the normal case. Destructive interference in the focus results in an intensity zero surrounded by intensity maxima: the well-known donut shape. 3D STED requires a different phase plate. And being able to mix both 2D and 3D STED at an arbitrary ratio results in a slightly more complex setup with a beam-splitter and two phase plates, one for 3D and one for 2D STED. In more sophisticated STED systems like abberior’s MIRAVA POLYSCLOPE, this arrangement is replaced by a sequential spatial light modulator (SLM). This avoids having to split up the beam, something that can cause serious misalignment.

In terms of essential components needed to move from confocal to STED, this is already it. But we haven’t put all the pieces fully together yet. For integrating the additional elements and functionalities into a microscope is accompanied by some challenges. After all, the goal is to acquire images with a resolution of a few ten nanometers – the demands on optical precision and mechanical stability (think drift and vibrations) are correspondingly higher. And this is where the wheat is separated from the chaff: only STED systems that meet these demands can fully exploit the potential gain in resolution and produce high-quality superresolution images.

Firstly, the STED beam has to be aligned with the excitation beam(s). Here, perfectionism is key: only when the STED and excitation beams are exactly on top of each other and remain there over time fluorophores can be efficiently de-excited. Any deviation results in a loss of resolution and signal.

abberior takes care of this with its Full Autoalignment, the first complete, fully automated alignment procedure for all beam paths that also includes the pinhole and STED beam shaping. The STEDYCON solves the issue its own way: it is fully aligned by design by coupling all lasers into a single optical fiber.

Then there is the fact that STED imaging gets rid of all the low-resolution photons. Consequently, detected signal levels are somewhat lower, while the background is not, because it is unaffected by the STED beam. abberior addresses this problem with a special detector. The MATRIX array detector measures in-focus and out-of-focus signal separately, facilitating a superior signal-to-background ratio and making the pinhole obsolete Dive deeper into the MATRIX here.

Another element demanding special attention in a STED setup is the scanner. abberior optimized its QUAD scanner for optical quality and the particular requirements of STED imaging. It works without a scan lens, reducing optical losses, and allows arbitrary scanfield rotation even for STED (which is something other scanners aren’t capable of).

Dedicated STED microscopes are not confocal microscopes with a retrofitted STED option. They were designed from scratch targeting the highest optical and optomechanical demands. It becomes clear that this also lifts confocal imaging to another level: an excellent STED system has so many excellent components that it is an excellent confocal system, too.

So, in summary, it takes essentially two things to turn a confocal microscope into a STED system: firstly, couple a donut-shaped laser beam into the beam path, and secondly, trim the individual components for best optical quality.

What remains is the question what all this means for you, the microscopist. Is operating a STED microscope any more complicated than using a confocal system?

Not at all! In fact, modern software – the STEDYCON smart control and the LiGHTBOX are exceptionally easy to use. Their internal logic reflects decades of STED experience in a condensed form to provide the user with suitable default settings for each scenario. With this, getting a good STED image is just three clicks. And it’s not all or nothing, either: you can seamlessly adjust resolution between confocal to STED simply by moving a single slider, giving you the best of all worlds, from pure confocal to high-end super resolution. You’re not missing out on the simple things, just because you can also do STED!

Ok, you may say, convinced: STED isn’t complicated, neither in theory nor in practice. But the sample preparation! Surely, I will need to establish a whole new protocol, right?

Again: no. There are some adjustments advisable, but generally the protocol hardly differs from established confocal sample prep.

So, hand to heart: is there any rational reason why you shouldn’t give STED a try?

In fact, there is not the one reason as to why a confocal microscope beats out classical wide-field; it’s a little bit of everything. The resolution of a confocal microscope is certainly higher, but most importantly, optical sectioning is significantly improved. Optical sectioning is the ability to generate a clear image of the focal plane – i.e. the plane in the sample the microscope’s objective lens is focused on – by suppressing signal that comes from out-of-focus areas.

The best way to think of optical sectioning is to imagine a simple sample structure, such as two cell membranes lying on top of each other. If one membrane lies exactly in the focal plane of the microscope, we should in principle be able to observe it brightly and sharply. But what about the second one, which is out of focus? Without optical sectioning, the blurred light coming from this defocused membrane would add to the focused light of the first membrane. We would see both at the same time, unable to distinguish them. With perfect optical sectioning, in contrast, the out-of-focus membrane would be completely dark, greatly reducing the background in our image.

A side-by-side comparison, as summarized in the table , highlights differences between the two methods that can impact their appropriateness for different applications. At abberior, however, we find this comparison to be moot. For us, both technologies have their rightful place in scientific inquiry. In fact, the two are inextricably connected. What makes out a STED system is the use of a donut-shaped light beam – the STED beam – in addition to the excitation beam. Take the STED beam away, and you’re left with a standard confocal microscope.

With focused illumination, high-intensity fluorescent signal, and pinhole filtering, a confocal microscope produces exceptional optical sectioning and 3D reconstructions of suitable samples (see Optical sectioning, or: tackling the background problem).

STED microscopy does the same, but with up to 10x higher resolution.

The commercialization of STED microscopy and other superresolution methods marks the advent of the era of nanoscopy, a time when not only the structure but also the dynamics of life can be examined in unprecedented detail by light microscopy. STED has revealed a sub-diffractive world of biological structures that confocal microscopes could never discriminate, and it can be applied to living systems. Researchers now visualize the striped pattern of ßIV-spectrin in axons, the interaction of SNAP25 with other proteins of the synaptic vesicle fusion machinery, the intricate cristae of mitochondria as they split and fuse, and the ultrastructure of the cytoskeleton, gut microvilli, and inner ear hair cells.

How does STED do this?

As mentioned above, a STED system is a standard confocal microscope using two superimposed light beams. The first excites fluorescent probes to emit photons, and like in a confocal microscope, its light diffracts over an area of about 200 nm, making structures smaller than this indistinguishable. In comes the donut-shaped second beam. Wherever the intensity of that beam is non-zero, it forces excited molecules back to their ground state, essentially ‘silencing’ them. Molecules in the central “donut hole”, where there is no STED light, are unaffected and emit detectable photons (read here how STED works in more detail). Thus, STED uses the donut-shaped beam to restrict emitting molecules to an area significantly smaller than 200 nm and you get sub-diffractive resolution.Tune thepower of the STED beam to zero, and you get a normal confocal microscope with diffraction-limited resolution.

The amount of STED action mediates a continuum between the two methods: confocal microscopy is diffraction-limited, STED microscopy is not. Everything in between is possible: from no STED at all, over using just a little bit of gentle low-power STED to gain a, say, 3-fold resolution increase with hardly any impact on the sample, and up to super-high superresolution (albeit at somewhat lower speeds and signal).

More compelling than pitting STED against confocal microscopy is the realm of opportunities that have emerged from the continued evolution of STED. While the gain in resolution by STED is an undeniable boon, the development and improvement of components and analytics for abberior STED systems has been a hotbed of innovations that excite not only our STED evangelists but also our confocal die-hards. Enhancements and breakthroughs to improve STED microscopy also solve problems shared with confocal microscopes. Here are some examples.

Biological specimens are often heterogeneous. We then pack that heterogeneity in an embedding medium, blast it with light through an immersion medium, and are disappointed when our confocal or STED microscope can’t produce the quality image we envisioned. The mismatched refractive indices of various media and the heterogeneity of the specimen itself cause light diversion. Our microscope’s ability to focus is thus compromised while light emitted by the fluorophores in the sample is also diverted. Together, these effects increasingly snuff out and smear signals the deeper you move into the sample. RAYSHAPE uses a deformable mirror that redirects diverted light back to where it belongs. With 140 digitally controlled actuators that precisely adjust the mirror surface within milliseconds, RAYSHAPE dynamically corrects aberrations at different depths as the focus plane moves through the specimen.

Whether confocal or STED, say bye-bye to aberrations with RAYSHAPE !

Fluorescence lifetime imaging uses differences in the decay rate of photon emission from fluorophores in a sample as an additional layer of information about a specimen. While lifetime imaging is an established method in standard fluorescence microscopy, its compatibility with STED microscopy is less well-known. And because we insist on going the extra mile, TIMEBOW — our module for seamless detection and analysis of lifetime data — works with our MATRIX detector (see below) to deliver accurate temporal and spatial information at once. This simultaneous recording of photon lifetime and its origin eliminates background and renders rich, high-contrast images. An upshot of that augmented information is a boost in resolution that allows you to reduce the intensity of the excitation beam of your confocal or STED microscope. So, you gain information and safeguard your specimen and probes.

Add the dimension of time to your imaging with TIMEBOW.

Both confocal and STED microscopes have some background when imaging thick samples. The pinhole of a confocal microscope cannot eliminate all out-of-focus signal, and STED excels at eliminating low-resolution signals in the focal plane but does little for the background. A point detector on either microscope records the in-focus signal and background together. Even sophisticated deconvolution can only partially subtract background and is prone to producing artifacts. However, an array of detectors that all “look” at the sample from slightly different angles can tease apart in-focus signal from background, just like your bifocal vision lets you distinguish objects in the foreground from those behind it. That’s MATRIX: multiple high-efficiency avalanche photodiodes (APDs) arranged in a hexagonal pattern on a single detector chip. MATRIX measures and subtracts the background of every pixel, so you get clear images with drastically reduced background levels for both confocal and STED microscopy.

See with 20 eyes instead of one with MATRIX.

A fundamental tradeoff in fluorescence microscopy – especially when working with live specimens – is the damage to fluorophores and tissues from prolonged exposure to high-intensity light. However, most samples are sparse, with only limited areas relevant for visualization. So, why illuminate blank spaces? FLEXPOSURE adaptive illumination minimizes photobleaching and phototoxicity by applying light only where necessary. FLEXPOSURE encompasses different adaptive illumination methods that access structural information point-by-point in a sample using low-intensity confocal and STED probing to determine where full illumination is required. Using FLEXPOSURE reduces light exposure on a sample, facilitating live-cell experiments, high-resolution imaging, and superior signal. And FLEXPOSURE applies to both STED and confocal imaging.

Take it easy on the light dose with FLEXPOSURE.

Of course, STED microscopy is at the core of what we do. The intention of every idea we transform into enhancements, power, and elegance for our instruments is to gain clear and precise insights into the nanoscopic world. Some developments are exclusive to our superresolution systems, like the single-beam design of Easy3D STED that eliminates signal and resolution loss due to beam misalignment. Yet, the varied improvements that extend over to confocal microscopy show that the establishment of paradigm-shifting technologies like STED prompts a surge in novelty that ripples widely. In this way, our STED microscopes are not only cutting-edge superresolution systems but also first-class confocal instruments. As a user, you may freely tune the resolution to what you need: anything between 200 and 20 nm is possible.

The shape of a donut seems mundane. Maybe that’s because we see it everywhere. It is, of course, that lovely pastry. It also moves cars along the road, keeps children safe in pools, and it can decorate an earlobe, a finger, or even a house door in December.

But the donut is quite remarkable for its simplicity.

Arguably, the donut is not the most common shape in nature. That distinction may belong to the sphere of rain drops or the hexagon of honeycomb and snowflakes. The donut or torus is, however, the shape of power. Think of Earth’s hazardous radiation belts of charged particles trapped from solar wind. They’re shaped into tori (that’s plural for torus) by the Earth’s bipolar magnetic field. A torus is also the chosen shape of a tokamak, that plasma-confining reactor that may someday produce energy by nuclear fusion. And in optical microscopy, the donut delivered a powerful blow to a stubborn paradigm curtailing the resolution of microscopes for decades.

It was the optical equivalent of a sonic boom riding on a ring of light.

Among the stimuli that evolution chose for us to perceive the world, light is like a self-absorbed celebrity. Light is vain. No one can move faster than her. Light is also capricious. She refuses to make up her mind about being a particle or a wave. That latter whim vexes optical microscopy. No matter how sophisticated the instrument, a point of light viewed through a microscope becomes a smeared blob because light waves passing through the aperture of an objective diffract. The result is that when two points of light are close enough to have overlapping smeared blobs, we can’t tell them apart. That is, there is an inherent limit to the resolution of a microscope dictated by the very nature of the signal it detects. At the microscale, this doesn’t matter. The diameter of the smear is in the hundred nanometer range. Zooming in, however, to look at the smallest of things – like molecules bouncing around in cells – is impossible. In 1873, physicist Ernst Karl Abbe told us that the best we can do is resolve objects that are about 200 nm apart. That’s 100 times larger than the size of a molecule. All molecules within 200 nm from one another appear as one blob under a microscope. This immutable fact, called the Abbe diffraction barrier, is literally etched in stone at the Friedrich Schiller University of Jena in Germany, and for decades microscopists sighed in resignation.

Of course, microscopes got better over the last century. Fluorescence confocal microscopy did wonders in revealing the inner workings of cells. Yet, resolution remained at Abbe’s limit. That is, until someone in 1990s tackled the problem from a different angle. Rather than trying to change the optics of a microscope, Stefan Hell exploited a loophole to outsmart the arrogance of light and blast through the resolution barrier.

Hell worked with fluorescence microscopes. Here, molecules of interest in a sample are tagged with tiny fluorescent probes or fluorophores. The fluorescence microscope targets light onto the probes, sending them into a state of excitation, and then detects the photons they emit upon returning to their ground state. Focusing on this difference in states was the crux of Hell’s loophole. He reasoned: the beam of excitation light projects onto the focal plane of a sample in a smear of approximately 200 nm. All fluorophores within that range become excited, emit photons, and thus, any structures smaller than 200 nm are indistinguishable. So, to improve resolution some of those fluorophores must be de-excited.

Donut. Enters. Scene.

Hell followed excitation with a second beam of light in the shape of a donut. That second beam was at a wavelength that knocked excited fluorophores back down to their ground state. Thus, all fluorophores in the donut-shaped area were silenced and the effective area of detectable photon emission was confined to the donut hole. Moving the donut over the sample yielded an image with sub-diffraction resolution.

Boom! The mundane donut bumped up the resolution power of microscopes ten-fold and beget the field of superresolution microscopy, also called nanoscopy.

Hell’s breakthrough technology, which he called stimulated emission depletion (STED), was the first to open the doors into the nanoscopic world. Later, other techniques also worked around Abbe’s diffraction limit with variations on the approach. Soon, superresolution was the leading edge of numerous developments in microscopy and its impact was recognized with a Nobel Prize in 2014.

Hell shared the Nobel with William Moerner and Eric Betzig, who in 2006 achieved sub-diffraction resolution with a method called single-molecule localization microscopy (SMLM). Like STED, this method exploits the different states of fluorescent probes. Where the technologies differ is in how they localize the probes. STED determines the location of fluorophores by restricting the area where they are allowed to emit photons – so location is defined prior to imaging. SMLM does no such thing. Instead, it randomly excites individual fluorophores and records their individual, smeared-out emission signal. That’s right. Just like light shining on a sample is smeared out, light coming back from the sample is smeared out by diffraction. The changes in pixel intensity of a single, non-overlapping smear are statistically predictable. So, to pinpoint the location of a fluorophore, SMLM fits a localization likelihood curve to the pixel intensities of the smear. That is, location is defined after imaging.

While both methods are available commercially and used widely, neither achieves a resolution much better than 20 nm. That leaves us short of our goal to tell apart individual molecules. What limits their performance is a dependence on high numbers of photons. Resolution could be improved if that dependency were inverted. So, in a world of technologies that scour for maxima in photon emission, how do you use fewer photons?

For that, the donut has an encore.

Hell is not a man to be satisfied with incompletes. He didn’t want to just break through the diffraction barrier. He wanted to push resolution as far as possible. Single-digit nanoscale resolution was his target. He understood the importance of photon numbers for resolution and was frustrated with the photon gluttony of STED and SMLM. So, he asked, is there a better way? The answer was to flip the job of the donut on its head.

Go back to the donut-shaped beam used in STED to silence fluorophores and consider what happens if you use it to excite them instead. A fluorophore in the ring will emit light, but at the very center of the donut – at the zero – you get nothing. Like the eye of a storm, the center is still. A fluorophore residing there is unexcited. Dark. No emission. Now move the donut such that the fluorophore comes closer to the ring of excitation light. The likelihood that the fluorophore is excited and emits a photon increases the further the donut moves. In fact, that likelihood takes the shape of a parabola. It’s high when the fluorophore is excited on one side of the donut, then dips down to zero as the fluorophore traverses the center hole, and then rises again on the other side of the donut.

Dredge up those faint memories of math and you’ll remember that a quadratic equation describes that parabola, and the minimum of that equation is at the zero of the donut. So, by moving the donut of excitation light and detecting just two emission signals, one on each arm of that parabola, you can pinpoint the fluorophore. The donut constrains the possible location to just the area of its hole and provides the information to determine the fluorophore’s exact position.

In flipping the job of the donut from depletion to excitation, Hell turns the principle of superresolution microscopy itself on its head. Instead of seeking the maximum of photon flux, his new technique determines the location of fluorophores by homing in on the minimum. And so, he named his new donut trick MINFLUX.

Of course, in practice more than two emissions are used for localization. However, each detected emission informs the movement of the donut to determine the fluorophore’s position, so you dramatically lower the number of photons needed (roughly by a factor of 10). Welcome to donut-based, photon-lean microscopic triangulation. And the best part? This shiny new donut knocks down resolution another 10-fold, bringing us into the unprecedented single-digit nanoscale. Molecules are now clearly visible.

How to achieve superresolution

At the end of the day, a lot of the distinction between photoactivated localization microscopy (PALM) and stochastic optical reconstruction microscopy (STORM) is historical. They were developed at around the same time and given two different names by their respective inventors. In practice, PALM and STORM are subsumed under terms like “blinking microscopy” or “single molecule localization microscopy” (SMLM). From a hardware perspective, there are zero differences anyway (but hardware obviously isn’t everything). Both techniques require a good microscope stand with strong widefield illumination, a high-end objective, and a fast, sensitive camera.

To find noticeable differences between PALM and STORM, we need to look at what they have in common first: both are types of superresolution fluorescence light microscopy. Other examples of such microscopes techniques are stimulated emission depletion (STED) and MINFLUX, which offers even more performance.

In a “normal” fluorescence microscope, the resolution is limited to about 200 nanometers by diffraction. Diffraction blurs the image of individual dye molecules and because of this, molecules that are close together cannot be seen separately. The image becomes fuzzy and details beyond the diffraction limit cannot be detected.

Without exception, all available superresolution techniques circumvent this problem by manipulating individual dye molecules such, that they do not light up at the same time. This in turn makes it possible to image them separately, despite them being blurred by diffraction.

There are roughly two approaches for this manipulation. Well, in fact, three.

Historically, the first one was STED microscopy, where a donut-shaped second “off-switching” light beam ensures that all emitters are dark – except for a small ensemble in the center, in the dark hole of the donut, where there is no off-switching light. Effectively, this silences most of the molecules that would otherwise emit at the same time (read more about how STED works). Thus, resolution is increased, with the advantage that this method is physically very direct and does not require any mathematical processing. STED is a point scanning method and can upgrade existing laser scanning microscopes.

Second, there are those methods that randomly keep fluorescent molecules in a dark state most of the time. Then, they allow a few of them to light up briefly, so that within a radius of 200 nanometers only one emitter is bright at a time. Although its image is still blurred, its center of mass can be located very precisely and thus the position of the molecule is also clear. PALM and STORM belong to this group. They take many images of isolated molecules on a camera and use software algorithms to determine their positions offline. All positions plotted in one image then result in the final high-resolution picture.

Before we get to the third approach, what is the difference between PALM and STORM now? In short, it’s not a big one. In fact, it is solely the type of fluorescent molecule. STORM uses organic dyes (e.g. abberior FLUX, Alexa Fluor 647 or Cy5) that stochastically flicker under certain chemical conditions. They are attached to their target structure for example by means of labeled antibodies and require a dedicated buffer. For PALM, on the other hand, photoactivatable or -convertible fluorescent proteins are used (mCherry, mEOS, etc.). To this end, the cell is genetically modified so that when it produces its own proteins, it attaches a fluorophore by itself. This makes the staining of living cells relatively easy. For capturing an image, a random subset of non-overlapping fluorophores is then stochastically activated and imaged before they bleach, and the next subset is imaged.

*with Adaptive Illumination

A comparison of PALM, STORM, STED and MINFLUX.

With both methods, the user must ensure that there are no two fluorophores in an emitting state within a diffraction limited region at the same time; otherwise, they would be counted as one molecule at a position exactly in between. For STORM, this means that the concentration of the buffer must be adjusted so that enough molecules are on, but not too many, to avoid the risk of overlap. With PALM, the user has to regulate the intensity of the activation laser for the same effect.

In general, there is obviously a certain effort involved to push the resolution beyond the diffraction limit. With PALM/STORM, most of the intricacy is in the sample preparation and the correct analysis of the raw camera images, for which the user is responsible. In contrast, STED requires more complex instrumentation, but the responsibility for success is more on the side of the instrument. The good thing is, commercial companies like abberior have nowadays taken care of this and using STED instruments is as straightforward as using a confocal microscope.

But why do we need a third option now? It becomes necessary when PALM/STORM or STED reach their limits. Although both can theoretically achieve unbounded resolution, in practice they do not. This is because the resolution depends on the number of photons, and no amount of photons can become arbitrarily large. The numbers may be high, but they are not unlimited and so is the resolution. But yet another technique, MINFLUX, gets around this by combining the advantages of blinking molecules with the advantages of using a donut-shaped light beam. MINFLUX is much more economical with photons than PALM/STORM and STED, and the savings can be invested in higher, even sub-nanometer resolution, and in much higher speeds. This is the reason why MINFLUX has a resolution even better than the size of a molecule and can measure their position several thousand times per second.

In 2001, Mats Gustafsson published a short communication in the Journal of Microscopy.¹ In it, he pointed out that resolution improvements of confocal fluorescence microscopy over wide-field microscopy were limited. He explained that “a confocal microscope detects extended resolution information only weakly and only if operated with a pinhole that is significantly smaller than the Airy disk. Such a small pinhole discards much of the desired in-focus emission light together with the unwanted out-of-focus light.” Gustafsson then proposed an alternative method: illuminating a sample with fine stripes of light would produce patterns containing high-resolution information that could be extracted computationally. He called the method structured illumination microscopy (SIM).

Since then, SIM has become a widespread tool for fluorescence imaging. SIM is fast because it requires no scanning, compatible with conventional fluorescent labels, and reliably generates decent images due to integrated data processing (but beware – not everything that looks good is good, as will be discussed below). Nevertheless, resolution with SIM is just as limited as with confocal, albeit at higher signal levels. Generally, commercial SIM systems can improve resolution over a wide-field microscope by a factor of 2 – about 100 nm laterally and roughly 350-400 nm axially. So, in terms of generating clear images at resolutions in double or even single-digit nanometric scales, superresolution systems like STED far outstrip the capabilities of SIM (see table further below for a comparison).

To understand that limitation, we need to take a quick jaunt into the world of frequencies.

Any visual stimulus – a signal, an image – can be represented as the sum of multiple sine waves of different frequencies and amplitudes. The high frequencies contributing to the sum generally correspond to fine detail, whereas low frequencies are the coarse aspects of the stimulus. A Fourier transform breaks down that sum into its constituent sine waves, essentially translating information from the spatial domain to the frequency domain.

The diffraction barrier that curbs the resolution of conventional optical systems in the spatial domain corresponds to a passband limit in the frequency domain that prevents high spatial frequencies from passing through the system.

SIM uses a striped, sinusoidal pattern of known frequency to excite fluorophores in a sample. Let’s say that this frequency is just at the edge of the passband of the microscope. The resulting distribution of excited fluorophores in the sample is essentially the product of this excitation frequency with the unknown frequencies that describe the sample. This means, some frequency mixing must have happened; we now also have the difference of the excitation and sample frequencies in addition to the original ones. This resulting difference frequency is small for all sample frequencies that are just a little higher (and lower) than the excitation frequency, just as the difference of two similar numbers is small.

But small frequencies can be detected, because they are lower than the original frequencies and can pass through the diffraction-limited detection passband. This means that we now can detect frequencies that are a bit higher than the excitation frequency that we put at the limit of the passband; frequencies that would normally be blocked. Thus, the emission signal includes extended resolution information.

However, that information is not directly accessible to produce an image, because the down-shifted frequencies mingle with the lower sampler frequencies that are still present. Also, striped illumination with a single excitation frequency obviously cannot excite fluorophores at all positions, and we’re also still only mixing along one spatial direction. Therefore, multiple readouts are made with the excitation illumination in different phases and angles, and these are then processed to computationally reconstruct the sample image.

In essence, confocal microscopy does the same thing as SIM. The key difference is simply that there is not just one frequency of excitation stripes as in SIM, but many of them. A confocal microscope’s illumination spot contains a continuous spectrum of low to high frequencies up to the passband, whose sum is exactly an Airy pattern.

Consequently, you can think of the emission signal as a sum of original and down-mixed sample frequencies. The key point is that some high frequencies are again shifted into the passband of the system, and the detected signal will include more detail than the equivalent in a wide-field microscope, where the frequency of excitation is exactly zero, meaning that no mixing happens. In a confocal microscope, excitation is localized in a tight spot instead of being spread out as stripes all over the sample like in SIM. For this reason, one can perfectly get away without computational processing. The raw data are the image, albeit just one small, though detailed, spot.

So, confocal microscopes scan the field of view to generate a complete image, one or a few spots at a time (depending on microscope type). What’s important to recognize about both SIM and confocal microscopy is that image formation is still diffraction-limited and the resolution is finite. An exception is nonlinear SIM, which uses high powers to saturate excitation, thereby generating a series of frequencies even higher above the passband, but nonlinear SIM has not seen widespread use due to excessive bleaching.

Modern SIM setups perform long imaging sessions (>1 h) and, being inherently parallelized, offer speeds of over 500 frames per second with a large field of view (> 200 µm). They can image samples up to a few cell layers thick and they standardly accommodate 4–6 different color labels. SIM images tend to look better than confocal images. For one, closing the pinhole of a confocal microscope to improve resolution also curtails the detection signal. And two, the mandatory post-processing step in SIM usually entails some kind of smoothing step.

However, these advantages of SIM are accompanied by some relevant limitations. The accuracy of SIM depends on knowing the shape of its excitation illumination. Optics and sample can distort those stripes, which generates artifacts in the reconstruction process, in particular with thicker samples or imperfect match of refractive indices. Although those artifacts have been characterized, they won’t always be apparent because the post-processing will still return an image based on warped data. Therefore, imaging fidelity requires careful alignment and calibration of the microscope, attention to differences in the refractive index of sample and immersion liquid, and knowledge about the impact of illumination and point spread function characteristics on the reconstructed image.

How does this compare to superresolution using stimulated emission depletion (STED)? While SIM and STED share some strengths, the disadvantages of SIM detailed above are unknown to STED. STED microscopy means true physical superresolution and the raw data is final, leaving no room for misinterpretation (read here how STED works). To top it off, today’s STED comes as a plug-and-play box that lets users generate superresolution images immediately and with minimal training.

The most important distinction between STED and SIM, however, is the fact that structured illumination imaging is still wavelength-dependent. That limitation cannot be overcome in the nanometric realm. Only systems that circumvent the diffraction limit will look deeper and more clearly at the molecular intricacies of the natural world.

So, maybe it’s time to move on.

Electron microscopes shoot a concentrated beam of electrons at a target object. An image is produced as the electrons pass through the specimen and are detected. Because the electron wavelength is several thousand times shorter than that of light, the resolving power of an electron microscope is a thousand times greater than that of a light microscope.

But is it the best resolution? Highest, yes. Best? Well, that depends on what you want to do with the resolution. In biological research, time and context often matter at least as much as size, which is why superresolved light microscopy techniques like STED and MINFLUX can play to their full strength here.

As you consider the advantages of resolution for your research, you should have a clear idea of what resolution means. It is distinct from magnification. In light microscopy, it is limited by an immutable property of light. And it is achieved and measured in different ways. Here are two articles to get you up to speed: “What is resolution?” and “How to measure resolution?“

Every scientific instrument comes with tradeoffs. Augmented power in one dimension, invariably curtails another. In the case of microscopes, boosting resolution complicates sample preparation and narrows the application spectrum. The best resolution is not always the highest resolution. When you’re looking for the best resolution, consider what you want to see. Most of the time, not only size but also time and context matter.

Biological specimens require a lot of sample preparation for electron microscopy. Why? Because you’re doing the equivalent of sending them into space and then blasting them with high energy. First, the specimen must be fixed. Otherwise, the energy of the electron beam will destroy it. The specimen must also be dehydrated to survive the intense vacuum inside the microscope. Then, many biological specimens are not conductive. As a result, electrons can’t pass through the sample, and you don’t get an image. Making biological specimens conductive involves coating them with a thin layer of metal.

Clearly, the hostile environment of an electron microscope precludes working with live or unfixed samples. So, if you’re interested in the movement, changes, and context that constitute life, a slight downgrade in resolution is likely your best bet. Enter the world of super-resolution microscopy.

In the realm of light microscopy, MINFLUX has repeatedly demonstrated single-digit nanometer resolution and below. The characterization of nuclear pore architecture and mitochondrial protein patterns are just two examples of the technology’s spatial resolving power. This capacity grants a completely new perspective on molecular structure in biological context, revealing the architecture of biomolecules and their interactions.

And yes, you can work with living cells.

In fact, the most distinctive feature of MINFLUX is its temporal resolution, which gives it the most advanced tracking capabilities of currently established microscopy technologies – by a long shot. That means that you can watch changes and movement happening inside living cells. The ability to differentiate events that are just a hundred microseconds apart expands the applications of MINFLUX from structural biology and slow processes like gene expression, to diffusion phenomena and even conformational changes of biomolecules. An example of this unprecedented power was the recent tracking of a kinesin-1 molecule walking along microtubules, including the corresponding configurational changes occurring at each step.The movement of kinesin-1 has never been tracked in a living cell before.

Research that does not require characterizing individual molecules but rather their spatial relation to others has a broader range of microscopy technologies at its disposal. Another step down on the resolution scales makes widefield, confocal, STED, and PALM/STORM microscopy all options (see figure). As super-resolution technologies, STED and PALM/STORM outperform the spatial resolution of diffraction-limited confocal and widefield microscopy by a factor of 10. Commonly discriminating objects at 20 nm, STED is also fast, which has enabled, for example, visualizing the fission and fusion of mitochondria with exceptional clarity.

Perhaps the most important differentiators of STED are its economic photon budget and easier sample preparation and data analysis compared to PALM/STORM. In fact, as a mature technology, STED microscopes like the MIRAVA POLYSCOPE are as easy to use as standard confocal microscopes. Furthermore, STEDYCON uniquely merges all three parameters – strong resolution, intuitive usability, and broad flexibility – into one exceptional instrument: a sleek box with a favorable price tag that transforms your existing widefield microscope into a confocal and a full-fledged STED instrument.

In fact, there is not the one reason as to why a confocal microscope beats out classical wide-field; it’s a little bit of everything. The resolution of a confocal microscope is certainly higher, but most importantly, optical sectioning is significantly improved. Optical sectioning is the ability to generate a clear image of the focal plane – i.e. the plane in the sample the microscope’s objective lens is focused on – by suppressing signal that comes from out-of-focus areas.

The best way to think of optical sectioning is to imagine a simple sample structure, such as two cell membranes lying on top of each other. If one membrane lies exactly in the focal plane of the microscope, we should in principle be able to observe it brightly and sharply. But what about the second one, which is out of focus? Without optical sectioning, the blurred light coming from this defocused membrane would add to the focused light of the first membrane. We would see both at the same time, unable to distinguish them. With perfect optical sectioning, in contrast, the out-of-focus membrane would be completely dark, greatly reducing the background in our image.

The optical sectioning ability of an epi-fluorescence microscope is exactly zero. This is because at all times the full thickness of the sample is both illuminated and detected. Fluorescent molecules that are not in focus will be blurred, but not dark, and therefore will always contribute in their entirety to the unwanted background signal. This effect is similar to taking photos with a camera: objects outside of the depth of focus are blurrier, but not darker.

In contrast, a confocal microscope restricts both illumination and detection to a small area. The sample is excited with a focused laser beam that has a high intensity almost only in the focal plane (we’ll get to the “almost” in a moment). Additionally, there is a pinhole in front of the detector that records the light coming back from the sample. This pinhole in turn restricts detection to almost only the focal plane. Fluorescent light that does not come from the focal plane will be defocused and not make it through the pinhole (Fig. 1). You may learn more about the differences between epi-fluorescence and confocal microscopy here.

Laterally restricting the field of view by illuminating and detecting only a small area obviously has a quite relevant disadvantage, similar to using a laser pointer to illuminate your desk instead of a lamp: only a very small spot is illuminated at a given time. If we want to see everything, we have to move this illumination – our microscope becomes a laser scanning microscope.

This is an acceptable trade-off, however, because optical sectioning is clearly better. In a confocal microscope, fluorescence quickly becomes darker with increasing distance from the focal plane, more precisely with the square of the distance: Doubling the distance of a molecule from the focal plane makes it four times dimmer. Fluorescence of a membrane that is far out of focus, for example, contributes little to the image signal and the background, which is important especially for thick samples.

The smaller the pinhole, the better this works, but at some point, it will work too good and also cut off lots of wanted light coming from the focal plane.

There are other ways to create an optical section. You could optimize illumination and try to restrict it to a thin plane. That’s exactly the idea of light sheet microscopy. The only thing is that, for physical reasons, you have to come from the side to do that. As a result, specimen storage and embedding becomes more difficult, and one has to switch to low-aperture objectives. So although light sheet microscopy has its uses, it doesn’t solve all the problems.

A similar approach is taken by two-photon microscopy (or, more generally, by multi-photon microscopy), which effectively restricts excitation to the focal plane. Here, a fluorophore is excited not by one but by two photons with twice the wavelength compared to single-photon excitation. These two photons have to reach the same fluorophore at the same time to elicit fluorescence, which requires high intensities present only right in the focus. Away from the focal spot, two-photon excitation is highly unlikely and as a result, there is hardly any background signal. Two-photon microscopy thus achieves very good optical sectioning and is well suited for deep tissue imaging. However, it requires expensive lasers, does not work with all fluorophores and suffers from a comparably weak signal, low axial resolution, and slow imaging speed. Therefore, it all depends on your sample and experimental setup whether two-photon microscopy is your method of choice.

Which leaves us with confocal microscopy and its pinhole as the presumably most universal technique for optical sectioning. However, it is far from perfect as the illumination laser shines almost only in the focal plane due to its focus, but only almost. After all, the laser has to get there first, and on this way (and also on the other side after the focus) there is a non-negligible probability that molecules are excited to fluoresce and show up as background in the image. This is not a problem in thin samples like stained membranes, but in thick samples such as deep tissue this dimming is not enough when there are easily tens of thousands of molecules in the background. And some of the background fluorescence will always pass the pinhole and reach the detector.

We can do better, can’t we? Yes, we can!

Array detectors like the MATRIX consist of more than 20 individual detector elements arranged side by side. They “look” at the sample from many point of views and therefore record in-focus and out-of-focus light separately. This way, the background contribution is measured for every pixel of the image and is removed from the total signal (Fig. 2). The result is greatly improved optical sectioning.

The step from confocal to MATRIX is as big as the one from wide-field to confocal. We remember that in a wide-field microscope, no sectioning is possible at all, so all molecules shine with the same brightness, no matter what distance they are from the focus. In a confocal microscope the signal decreases quadratically with the distance to the focal plane. With a MATRIX detector, it does so with the 4^th power. Structures that are 10 times further away from the focus are therefore not 100 times dimmer, but a whopping 10,000 times.

Optical sectioning becomes possible by laser beam and pinhole joining forces. But only array detection will bring it to perfection. 

Thus, being in the market for a high-quality, high-performance research microscope can feel like navigating a minefield. There are numerous types, models, and makes of microscopes to consider. The language of listed features often uses enigmatic brand names that obscure comparability, especially for buyers beginning to delve into available technologies. And then there’s your budget. It’s not helpful to size it up to the generic and broad price ranges accessible without a quote.

Prices for a decent but very basic confocal microscope start at around 100,000 US dollars, while list prices for high-end confocal systems sometimes reach a million dollars. Including superresolution features like MINFLUX, PALM/STORM, or STED typically adds another few 100,000 dollars to the price each, so that the required spending for the most sophisticated superresolution microscope can be well over a million. However, consider that an electron microscope with a similar resolution as MINFLUX can cost many times more. 

Be aware that taking these price ranges at face value is misleading. You can’t really compare an entry-level transmission electron microscope to, for example, a high-end MINFLUX microscope. Comparing price ranges may help you exclude some platform categories. A category may exceed your budget. The type of instrument may not match your microscopy needs (do you really need an electron microscope?). The size of objects you wish to visualize may require a resolution beyond the diffraction limit.

But while the price of a microscope is unquestionably a deal-breaker for its purchase, a more helpful criterion is its value. So, once you have narrowed down your options, think about the long-term nature of your research. Specifically, compare and contrast three aspects: resolution, flexibility, and ease of use.

The higher the resolution you need, the more limited your choice of microscopes. To visualize objects separated by less than 20 nm, your choices are MINFLUX (or a similar non-commercial platform) or electron microscopy. The price tag of these instruments reflects the complexity of their underlying technology (which often correlates with complex sample and instrument handling). Nevertheless, MINFLUX offers a comparative edge if you seek to visualize living cells and are looking for high temporal resolution.

Temporal resolution, you ask? Yup. If you are only interested in snapshots of cellular processes, you don’t need good temporal resolution. You’ll be taking a single image. If, however, you are interested in the dynamic changes that occur in a cell, investing in a fast microscope can generate savings in the long-run. An interactive graph in the article “Superresolution for biology: when size, time, and context matter” lets you examine the tradeoffs and coverage of spatial and temporal resolution for a range of platforms, from widefield to expansion and scanning electron microscopy.

Most microscopes today are built to offer modular functionality, allowing you to combine different microscopy methodologies into one instrument. Naturally, that modularity comes with careful engineering that ensures robust, seamless, and harmonized operation of each module regardless of when or how you use it. Add more modules to your microscope, expect the price to go up. The combination of modules is also finite and may entail compromises for other dimensions. Keep these tradeoffs in mind as you evaluate how you will meet research needs. Nevertheless, an investment in a scope that allows you to explore more than one microscopy methodology, although expensive in the short-term, will likely save money in the long-run.

Ease of use is often a “last thought” criterion in the purchase of advanced equipment. And yet, it is pivotal to the return on your investment. An advanced microscope in a lab is meant to produce data. That means that the instrument should be used as extensively and by as many people as possible, preferably without tying up your time. That simply won’t happen if handling is complicated and fastidious.

Some publications outline ways to upgrade an existing microscope to new performance levels or to build your own platform for a fraction of the price of a commercial instrument.^1,2 With the right expertise in laser and optics technology, such DIY platforms are an option. However, your time and the progress of your research are as relevant to your success as your budget. Consider the time you need to set up and troubleshoot such a solution. And will everyone in your lab be able to use it? How will you train operators? Furthermore, how will you handle instrument downtime when repairs or improvements are needed? All these factors influence your return on investment.

Flexibility, power, and ease of use are the key to a microscope that serves for many years and across many projects. Bearing all that in mind, a slightly higher price point may in fact be better value.

A conventional light microscope, also known as an optical microscope, is a versatile and widely available tool that uses light at 400-700 nm wavelength to illuminate a specimen. The interaction of visible light with the biological specimen leads to absorption, transmission, or reflection, and the resulting image is magnified by a combination of objective and eyepiece lenses. Optical microscopes vary in layout depending on the application, but the basic setup includes a condenser to focus light on a specimen, plus an objective lens and eyepieces for magnification (Fig. 1). Various types, such as dark-field and fluorescence microscopes, offer specialized capabilities for studying specific cellular structures.

MINFLUX is the spearhead of light microscopy today. With a resolving power down to a few nanometers, it penetrates into the realm previously reserved for electron microscopy. What’s more, in contrast to electron microscopy, MINFLUX may even be used to examine living samples and molecular dynamics (see below).

But let’s first take a look at how an electron microscope works. An electron microscope uses a beam of electrons to unveil specimen details. Operating in a vacuum, it employs magnetic lenses – a condenser, objective, and projection lens – to focus and magnify the electron beam and capture specimen information (Fig. 1). The image is generated by the transmission or scattering of electrons. In transmission electron microscopy (TEM), electrons pierce through specimens and leave details on fluorescent screens. Scanning electron microscopy (SEM) crafts 3D-like images by scanning a specimen’s surface with a focused electron beam.

Conventional optical microscopes magnify a specimen up to 1,500x. However, their resolving power is limited by the wavelength of visible light. The performance of optical lenses decays at shorter wavelengths (∼ 400 nm) and Abbe’s diffraction limit puts the resolution boundary at roughly half that wavelength. Thus, structures smaller than 200 nm laterally and 600-700 nm axially are blurred, leaving numerous subcellular structures inaccessible.

By comparison, electron microscopes offer far superior resolution, achieving a magnification of up to 1,000,000x and sub-nanometer resolution about 250 times that of a conventional light microscope. What’s the difference? The specimen is illuminated with electrons, which have a far shorter wavelength than photons. Operated at an accelerating voltage of 200 keV, an electron microscope’s illumination source has a wavelength of roughly 2.5 pm. That’s picometers. The details of viruses (250-30 nm), proteins (10 nm), and even glucose molecules (1 nm) come clearly into view. For electron microscopes, resolution is not limited by wavelength but by its electromagnetic lenses, which cannot be shaped as precisely as the optical lenses of light microscopes.

Compared to electron microscopy, light microscopy steals the show here. You can use a variety of specimen types – living or dead, fixed or unfixed, stained or unstained, with thickness in the micrometer range. You can peer into specimens, like living cells, as they are transparent to photons. And it doesn’t take much to prepare these specimens. Staining with colored dyes or fluorophores to enhance contrast and highlight specific cellular structures is straightforward, and various colors can be used simultaneously to reveal multiple targets at once. Along with mounting on a glass slide with a coverslip, preparation is done within minutes to hours.

Electron microscopy excludes live-specimen observation. Also, marking structures of interest with colored dyes and labels, like in light microscopy, is not an option. Specimens must be ultra-thin (usually 0.1 µm or below) and undergo a series of intricate preparation steps, from fixation to dehydration and coating with heavy metals to reflect electrons before mounting on a copper grid. Preparation is labor-intensive, requires advanced skills, takes several days to complete, and makes it impossible to keep a specimen alive.

Both conventional light microscopy and electron microscopy have their place in science. Each unveils unique aspects of life at different scales. On the one hand, simplicity and user-friendly operation make conventional light microscopes a fundamental choice for many scientific labs. They are compact, lightweight, suitable for field use, and come at an affordable price and low maintenance costs. On the other hand, electron microscopes are far superior in magnification and sub-nanometer resolution, clearly showing the molecular world. That high resolution, however, comes with a high price tag, requires dedicated rooms and trained operators, and the imaging conditions are incompatible with live specimens. So, is there a way of getting the best of both microscopes?

We’ve compared the electron microscope to conventional light microscopes, but today’s microscopy armory includes superresolution instruments that circumvent Abbe’s diffraction limit, revealing structures well into the nanometer range and below. MINFLUX is one instrument in particular that has pushed the boundaries of superresolution. Noteworthy is that MINFLUX steps in where the electron microscope comes to a halt: examining live specimens and investigating live molecule dynamics. Multicolored MINFLUX visualizes the two- and three-dimensional distribution of molecules within live cellular landscapes at a resolution of 1–3 nm. It has been used to image single molecules in peroxisomes, mitochondrial membranes, photoreceptors, neurons, and nuclear pores (Fig. 3). Furthermore, with a temporal resolution of less than 1 millisecond, MINFLUX also captures conformational changes of molecules, like the step-by-step movement of kinesin-1 along microtubules^1,2 and the activation of the mechanosensitive ion channel PIEZO1.³

With simplified sample preparation and affordability, this advanced imaging technology is accessible and powerful. Want to know more about how MINFLUX makes the invisible visible? Dive into the details in this article.

¹ Deguchi, T. et al. 2023. Science 379: 1010.
² Wolff, J. O. et al. 2023. Science 379: 1004.
³ Mulhall, E. M. et al. 2023. Nature 620: 1117.

Microscopy methods and microscope types abound for a broad range of applications. Among them, confocal microscopes have become a staple of biological imaging. First commercialized in the mid-1980s, confocal microscopes offered higher resolution and more sensitive detection than epifluorescence wide-field microscopes. They also generated clearer images. Instruments were subsequently outfitted with z-stacking, and today’s confocal microscopes reconstruct three-dimensional renditions of optically sectioned specimens. State-of-the-art stimulated emission depletion (STED) microscopes equipped with MATRIX array detection will even yield superresolved images of living specimen with exceptional signal-to-background ratio and clarity.

Optical sectioning is the process of imaging the x-y plane of a sample in the focal plane of a microscope’s objective and successively moving along the z-axis to generate a series of images spanning the depth of the sample. Overlay these images correctly, and you can build a 3D map of your sample, much like stacking the slices of an orange to reconstruct the complete fruit. The value of this process is realized only if each section includes information exclusively from molecules present in its specific focal plane. Any out-of-focus light from above or below the plane obstructs the clarity of the reconstruction and confounds any quantification of fluorescence intensity. This is precisely where an epifluorescence wide-field microscope falls short.

An epifluorescence wide-field microscope illuminates and detects the entire thickness of a sample. You inundate the specimen with excitation light, and fluorophores everywhere shine right back. Those fluorophores in the objective’s focal plane will appear as clear, strong signals. However, the camera will also capture light from molecules above and below the plane. These will appear as a haze that muddles the objects of interest. Generally, the haze is a minor nuisance when working with thin or sparse specimens at low magnification. Wide-field microscopy quickly captures very good images. As samples get thicker, however, out-of-focus light becomes increasingly problematic, especially at higher magnifications.

The laser of a confocal microscope illuminates only a small area at the focal plane in the sample. Fluorophores in and out of the focal plane become excited, and the emitted light of the latter is blurred, as in a wide-field microscope. The difference here is that a confocal microscope includes a pinhole positioned before the detector (Fig. 1).  Emitted light from molecules on and close to the focal plane of the objective passes through the pinhole to the detector. Much of the out-of-focus light is blocked. Because of the pinhole, the light emitted by fluorophores outside of the focal plane contributes significantly less to the image and background, creating clearer images. The smaller the pinhole, the better out-of-focus light is blocked, but the more wanted signal is lost.

So far, though, we have only recorded information from a small spot of the sample. You have two options to capture a complete image. One option is to use a laser scanning confocal microscope (LSCM), which creates a complete image by making a series of recordings along a raster that covers the field of view in a sample. The detector is often a single detector, such as a photomultiplier tube, an avalanche photodiode or similar, that quickly amplifies the low light captured from a spot. Because the location of detected light is encoded in the raster, a camera detector is unnecessary. An LSCM produces high-quality, in-focus 3D images of thick samples, but it takes time to build optical sections of a sample spot-by-spot and then stack these into a 3D rendition.

A spinning disk confocal microscope (SDCM) is a more time-efficient alternative. An SDCM replaces the single pinhole on the detection path of an LSCM with hundreds of pinholes arranged in spirals on a disk. With each disk rotation, the pinholes sweep across the field of view, allowing light from multiple unique spots on the sample through to a high-efficiency CCD. This accelerates signal acquisition compared to an LSCM.

The higher number of pinholes of an SDCM is also its limitation, however. In thicker samples, there is a point where out-of-focus light overlaps with adjacent pinholes on the disk. That light will reach the detector and muddle up your image.

So, is there a rule of thumb in choosing a suitable microscope? The following table can help.

Confocal microscopy isn’t the be-all and end-all of clear optical sectioning and 3D imaging. A significant improvement on the confocal strategy of “block out-of-focus light” is “record and subtract out-of-focus light.” That’s possible by changing the way that emission light is detected. The MATRIX detector is an array of photodiodes that record light emitted by fluorophores in a sample from different angles. Integrating these perspectives allows MATRIX to record the background separately from the in-focus signal. That information is then easily subtracted for a clear, in-focus image. Add the MATRIX detector to a STED microscope, and you get exceptional clarity, superresolution, and live-cell imaging in one. The what and why of confocal versus STED microscopy will be the topic of another article.

Confocal and superresolution microscopy rely on the specific labeling of proteins of interest with a fluorescent tag. The classic approach to this end is immunofluorescence, where antibodies carrying a fluorescent label recognize and selectively bind to target proteins (you may learn more about immunofluorescence here). Thanks to its versatility, this strategy is still the most popular. However, there are drawbacks, one of them being that antibody staining does not work in living cells. But the recent development of nanobodies opens new ways here and adds to the strategy’s potential in particular in superresolution microscopy (again, another article in our knowledge base will tell you more).

And there are other strategies to make your sample light up under the microscope which come along with distinct advantages, like co-expression of a fluorescent protein or fluorescent labeling via click-chemistry.

A comparably new option is the HaloTag^®.¹ The HaloTag^® is a small protein derived from the bacterial enzyme halokane dehalogenase.^2,3 It can be fused to a protein of interest by genetic engineering and binds synthetic ligands. These HaloTag^® ligands are key players in the labeling process. They consist of a linker and a functional group (we will come to what this is in a moment). Once the ligand comes into the vicinity of the HaloTag^® protein, a covalent bond – i.e. a very stable chemical connection – is formed between the protein and the linker: they stick together as if glued, permanently uniting the protein with the ligand’s functional group. And this functional group can be many things, depending on the experimental setting. One thing the functional group can be is a fluorophore.

The HaloTag^® thus allows to precisely target proteins of interest with fluorescent labels, opening the door to a wide range of applications in (superresolution) microscopy. In particular live-cell imaging benefits from the use of HaloTag® ligands carrying organic fluorophores as these can easily enter living cells and don’t harm them. But also in fixed cells and other in vitro systems the HaloTag^® demonstrates its power and versatility as the stability of the protein-ligand gluing guarantees good staining results even under challenging conditions.² Last but not least, the HaloTag^® system comes along with minimal background fluorescence and an excellent signal-to-noise ratio, both criteria decisive for microscopic imaging.

So with the HaloTag^® we have a super-stable protein-ligand complex, strong imaging contrast, and can do live-cell imaging. All good, end of story? Not quite… as every coin has a flip side.

For fluorophores, this is photobleaching: Fluorophores exposed to light will bleach at some point. When quasi-permanently bound to their target proteins and exposed to a lot of very powerful and focused light – as is usually the case in fluorescence microscopy – this point will be reached quite soon and certainly sooner than any microscopist could wish. And once this point is reached, there is no way to restore fluorescence – pretty much the same as a poster exposed to sunlight for a long time will lose its color and the only option to restore it to original brightness is to replace the poster with a fresh version. With covalently coupled fluorophores, this is not possible at any rate. Bleached stays bleached.

There are, however, ways of fluorophore coupling that are not permanent – more the kiss-and-run site than the glue-site, so to speak. One example is DNA-point accumulation for imaging in nanoscale topography (PAINT)⁴: It uses short stretches of DNA as ligands. The DNA stretches specifically but transiently bind to their targets – which are DNA sequence tags on the proteins of interest –, i.e. they bind and unbind at a specific rate. The advantage in terms of photobleaching is clear: As fluorophores at target proteins are constantly replaced by “fresh” ones from the virtually unlimited reservoir in the surrounding buffer, bleaching is no longer an issue.

Which brings us back to the HaloTag^®: Scientists were not satisfied (when are they ever?) with the classical glue-type HaloTag^®’s limitations and were looking for ways to expand its applicability.⁵ So they chemically engineered the HaloTag^® ligand, replacing the chloride atom in the linker with a sulfur-containing reactive group. This seemingly miniature change brought about a major change in the molecule’s chemical behavior: The manipulated linker still fits into the reactive pocket of the HaloTag^® protein but the chemical reaction forming the covalent bond can no longer take place: The HaloTag^® ligand binds and unbinds only transiently, just like the DNA in DNA-PAINT. It has become exchangeable and we now call it HaloX^®.

Note that only the ligand is changed but not the protein. This means that HaloX^® is easily applicable: Researchers previously using the HaloTag^® system can continue with their established HaloTag^®-coupled proteins: Just use the new HaloX® ligands and go ahead!

HaloX^® is an alternative to DNA-coupled labeling in PAINT and also expands opportunities for MINFLUX microscopy: The MINFLUX concept relies on fluorophores to switch between light and dark states to avoid signal overlap and to achieve its exceptional spatial resolution and high localization precision.

Generally, such stochastic light-dark-switching can be achieved by intrinsic characteristics of fluorophores such as conformational changes that either suppress or allow fluorescence to occur. With HaloX^®, like with other exchangeable dyes, it is no longer necessary to work with stochastically switching fluorophores. The light-dark-switch is achieved by the binding-unbinding of the fluorophore to its target, like in DNA-PAINT for MINFLUX⁶: While freely diffusing, the coupled fluorescent labels remain “dark”. Upon transient binding to a target protein, they light up, thus delivering photons from a defined position. This behavior is called fluorogenic.

Fluorogenicity is a property generally required for exchangeable ligands, not only in the context of MINFLUX: In classical staining strategies such as antibody labeling, residual and unbound fluorophores are washed away before imaging. Working with fluorophores coupled to exchangeable ligands, in contrast, requires imaging to take place in the presence of fluorophores freely diffusing in the buffer as a reservoir to constantly replace those bound to target proteins. If the fluorophores were not fluorogenic, they would be excitable also in their unbound state, resulting in extremely high background staining.

The potential of HaloX^® is not yet exhausted. Currently, it is restricted to one-color imaging as a second color requires the HaloTag^® protein to be modified. However, it has already been shown that suitable combinations of modified HaloTag® protein and linker are possible.⁵ This will facilitate two-color imaging using two different exchangeable HaloTag^® variants and will extend the system’s applicability even more.

HaloTag^® is registered trademark of Promega Corporation

HaloX^® is registered trademark of Spirochrome AG. This product is covered by one or more license from Spirochrome AG and, is intended for Research Use Only (RUO).

Let’s start with a rather obvious fact: Optical microscopy requires light. The specimen has to be illuminated in some way to reveal information on its shape and structure. In classical widefield microscopy, the demands on the light are quite simple: it has to be bright enough to reveal sufficient detail. Standard widefield microscopes usually work with mercury or xenon gas-arc lamps or LEDs and the light is evenly distributed over the specimen.

For confocal fluorescence microscopy, however, things are more complex, and in stimulated emission depletion (STED) microscopy even more so. Which is the reason why at abberior we devote particular attention to the laser topic.

But before we go into more detail, let’s first see what requirements confocal microscopy places on the light. For one, only a very small part of the sample is investigated at a time, and the light needs to be focused on precisely this spot. For another, fluorescence microscopy relies on fluorescent molecules – called fluorophores – in the studied sample, hence its name. When irradiated with light of a defined wavelength, these fluorophores emit light of a different (higher) wavelength. (The wavelength determines the light’s color: light with a wavelength of around 450 nm, for instance, is blue and light of approximately 700 nm is red; green and yellow lie in between the two). Every fluorophore has its own specific spectrum of excitation – where it takes up the light’s energy – and emission – where it emits light of its own. For this to work effectively, the excitation light has to be of exactly the right wavelength and highly focusable. Which is why you need a laser.

Laser stands for “light amplification by the stimulated emission of radiation”, which describes the way it generates light. Even kids – and not only those having watched Star Wars – will be able to tell you that lasers produce powerful light beams in a multitude of colors.

And this description is in fact pretty close to the actual physical definition, stating that a laser is a light source generating monochromatic, coherent, and unidirectional irradiation.

Ehm, come again?

All right, one by one: Laser light is

• monochromatic: it consists of only one wavelength or a narrow range of wavelengths, equaling a specific color.
• coherent: it has one wave form and frequency
• unidirectional: all light waves go in the same direction, creating a beam

These are all properties unavailable from classical light sources, which is why the development of lasers in the 1960s incited an avalanche of new technologies exploiting the new power of laser light.

But there is not “the one laser”. Many different types of lasers have been developed over time and they come with quite distinct properties, perfectly shaped for various applications all around us – from laser pointers to printers, scanners, welding lasers and lasers for medical or scientific applications, to name just a few. They are usually classified by their gain medium, which is the actual source of their light: There are gas lasers, liquid lasers, solid-state lasers, and semiconductor or diode lasers.

In microscopy, lasers are most commonly used to excite fluorophores. A typical excitation laser has a power output in the range of a few milliwatts. For a long time, one of the most commonly used excitation lasers in fluorescence microscopy was the Argon-ion laser. This type of laser produces light of a number of wavelengths conveniently matching the absorption spectrum of various fluorescent dyes. This allows for the excitation of multiple dyes simultaneously, facilitating multicolor imaging.

Another popular laser used in fluorescence microscopy is the Helium-Neon laser (HeNe). HeNe lasers produce a single fixed wavelength of light (usually red), which makes them a stable and easy-to-use option.

In recent years, compact and affordable diode lasers have replaced many other types. They offer several advantages, including low power consumption, high efficiency, and ease of use. Moreover, they can be operated in pulsed mode, which is of great advantage for fluorescence microscopy. Pulsing means that the laser light is not emitted continuously (as with continuous-wave (cw) lasers), but as short, periodic pulses, which are usually a few 100 picoseconds long with a pause in between of a few ten nanoseconds. This way, the power is concentrated on the short pulses, greatly alleviating the energy burden on a specimen, a critical prerequisite for live cell imaging as both photobleaching and phototoxicity are reduced. Pulsed lasers are also a prerequisite for fluorescence lifetime imaging, which is based on measuring the time between a light pulse and the arrival of photons.

abberior’s STEDYCON, MIRAVA POLYSCOPE, and INFINITY microscopes are all equipped with diode lasers for excitation that get on with a power below 100 microwatts.

Most recently, fiber-based pulsed white light lasers have also gained traction. Their biggest plus is their versatility: They do not emit at a single or few wavelengths, but over the full spectrum. But wait – did you pay attention? White lasers seem to violate the definition of lasers as their light is not monochromatic. However, they generate white light by a so-called non-linear optical process from monochromatic light. And their light is unidirectional and coherent, in line with the definition. So we can speak of white lasers as lasers, after all. 

With a tunable filter any desired wavelength can be selected from a white laser. Sounds good, doesn’t it? Well, only up to this point, for cost and complexity for both the laser and the required tunable filters are high. And when the only laser breaks down, the whole microscope is instantly unusable. 

In superresolution microscopy, white lasers have another pivotal drawback: The fact that a wide range of wavelengths is emitted from the same source means that chromatic aberrations cannot be compensated for all wavelengths and objective lenses, which significantly reduces image quality.

Stimulated emission depletion (STED) microscopy introduces a new level of complexity to the topic of lasers as here fluorophores are not only excited but also deexcited. The light to shut down undesired fluorescence is delivered by a STED laser, and the requirements for this job are quite distinct from those for excitation. Most importantly, STED lasers must deliver much higher intensity light and allow pulsed operation. The former guarantees that fluorescence is de-excited efficiently in the area of the STED beam while the latter ensures that the energy is focused in time, precisely right after the excitation pulse when it achieves maximum effect on the excited dye. Any photons arriving at the sample outside this narrow time slot are entirely useless and only do harm by bleaching fluorophores or damaging the sample – which is why cw STED lasers are not a practical option (Fig. 2).

In the past, pulsed titanium-sapphire lasers were often used for STED microscopes, but compact, reliable and affordable fiber-lasers have made them largely obsolete for this application. The STED lasers installed in abberior microscopes are pulsed diode lasers with powers between 1 and 3 watts.

Superresolution techniques like PALM and STORM, the most important types of single molecule localization microscopy (SMLM), in turn come with other requirements for their lasers. Since they are essentially wide-field techniques, the excitation light is constantly spread over the full field of view. This requires high-power, typically fiber-based lasers, in the range of 1 to 5 watts to achieve the local power density required for excitation. If you want to learn more about SMLM, we recommend this article.

MINFLUX is the single fluorescence microscopy method that reaches molecular resolution. It also holds the temporal resolution record in the field. One would expect this high-end microscopy technique to need particularly sophisticated lasers. The requirements indeed differ from those posed by STED but are not necessarily stricter. Commonly, MINFLUX instruments use a diode laser with a power of several hundred milliwatts.

A photon is like the Road Runner with symptoms of ADHD: Full of energy, it dashes through space with – indeed! – the speed of light, but it’s easily distracted by anything it passes on its way and will instantaneously divert from its straight path. Only that the photon’s ADHD is truly incurable: It is in the photon’s wave properties and those we cannot change. The very physics of light dictates that it is bent whenever it transits at an angle between media with different refractive indices. This phenomenon is called refraction, and the refractive index is a value for the medium’s optical density, indicating how strongly light is deflected. When refraction diverts light rays from their destined path, it smears the focus and leads to murky images. These are aberrations.

The degree by which light is diverted does not only depend on the refractive index, which usually is a function of the light’s wavelength. That’s why droplets of water in the air can split the sun’s light into its spectral parts and create a rainbow. Usually, the shorter the wavelength (and the higher the energy), the stronger the diversion.

In nature we find refraction creating sights that please the eye. For a light microscopist, refraction is what makes lenses work. However, refraction can also become a nuisance. On the way from the objective lens into the sample and out again, the light has to pass air or an immersion medium, the cover slip, and the mounting medium. Even when those media are closely matched according to their refractive indices, residual unwanted refraction easily leads to aberrations.

Then there is the sample itself: Biological samples are chronically unordered, often inhomogeneous, and countless structures refract the light on its way. And the deeper you focus into the sample, the stronger the effect. Eventually, you will be left with hardly any signal at all.

Do we have to stay at the sample’s surface then, where aberrations can at least be controlled to some degree? Do we have to accept that an image will be turbid and that valuable information hidden deeper in the sample has to remain unattainable? And that’s that?

Certainly not! Scientists pondering the aberration problem for quite a while came up with a solution: It should be possible to compensate for aberrations by introducing a corrective in the beam path that has the opposite effect. Since aberrations are a function of the sample, immersion and embedding media, and wavelength, the operator must be able to adapt the correction for it to be useable in practice.

One option to this end is to use an objective lens containing movable groups of lenses that can be adjusted to correct for aberrations. However, these correction collar objectives suffer from a number of relevant limitations: As they work mechanically, their settings are poorly reproducible and some imprecision will always remain. Furthermore, they fail when confronted with irregular aberrations as introduced by e.g. sample inhomogeneities. They are also rather slow and even when motorized they cannot dynamically adapt their correction while focusing through the specimen. This means that you have to decide where in your sample you want to restore your signal: On top, somewhere in the middle, or at the bottom? The rest will always remain dark and murky. In consequence, correction collar objectives are poorly suited for 3D scans.

And then there is the issue that the objective best suited to your embedding medium might not be available with a correction collar.

Another problem is that correction collar objectives push the focus out of its original position as soon as their lenses are shifted to compensate for aberration, a phenomenon familiar from zoom objectives in photography.

In summary, correction collar objectives can in principle compensate for aberrations, but this only works well for static images of sufficiently homogeneous samples and for a very limited depth range. Their shortcomings are even more evident when it comes to superresolution microscopy. When taking images with nanometer resolution, the coarse imprecision of a mechanical system becomes unacceptable.

So let’s dismiss this solution and hand over the stage to adaptive optics: This strategy takes a different approach to put diverted rays back on track by applying a deformable mirror whose surface can be shaped as desired. 140 actuators pull on the membrane to make it resemble the distortions introduced by the sample, but in the opposite direction. In other words, the mirror’s surface is a negative of the wavefront error and can thus cancel aberrations.

The correction achieved with a deformable mirror is highly precise and can be adapted to different situations quickly, for example to varying focusing depths. As the actuators are controlled electronically, this happens within milliseconds so that the correction can dynamically follow the focus during z scans, guaranteeing perfect correction along the complete axis. And this is possible on the fly and without influencing the focus. Adaptive optics with a deformable mirror thus outperforms correction collar objectives in every relevant aspect and image acquisition runs completely autonomously for bright, high-resolution images from the top down deep into the sample.

The diffraction limit dictates the smallest possible size of an excitation PSF in an ideal situation. In practice, imperfections in the optical train, from the laser sources, up to and including the sample, and back to the detectors, compromise the imaging properties of the microscope and deform its otherwise diffraction-limited PSF. The sample is of special consideration. While imperfections in the optics are static, and therefore easily measured and corrected, inhomogeneities in the specimen are unpredictable and vary from sample to sample and even within a single imaging field of view^1, 2. These imperfections and inhomogeneities give rise to optical aberrations, which – if left uncorrected – cause the excitation PSF of the microscope to become more diffuse, compromising the system’s resolution and excitation efficiency. Additionally, fluorescence that would otherwise be perfectly focused at the pinhole, becomes smeared out and can be blocked by the pinhole, leading to inefficient detection.

Super-resolution microscopes have especially high demands for optical perfection. In a STED microscope, the main concern is the PSF of the STED laser. If the STED beam picks up aberrations on its way to the focal plane, the center of the STED-PSF will have a finite, non-zero intensity. This aberrated STED-PSF will then de-excite fluorescence entirely, rather than just confine it to the center of the ring, thus resulting in heavy losses in signal and resolution.

The problem is especially evident with three-dimensional-(3D-)STED. While the zero-intensity center of the two-dimensional STED doughnut is somewhat robust against aberrations, the center of the 3D-STED PSF quickly becomes non-zero even if only minor aberrations are present ^{3, 4}.

A major source of aberrations in microscopy are regions in the sample with non-uniform refractive index numbers. When light encounters a change in refractive index, the rays are bent and continue to travel in a different direction, a phenomenon known as refraction. While designing lenses, refraction is a desired effect and manufacturers take great care in optimizing refraction so that lenses produce perfect (i.e. diffraction-limited) PSFs. Unfortunately, subsequent arbitrary changes in refractive index along a microscope’s optical beam path compromise this precise focusing ability.

One of the most prominent causes of an unwanted change in refractive index is the interface between the coverslip and the embedding medium of the sample, where an index mismatch can cause spherical aberrations and defocus (Fig. 1). Defocus changes the apparent focusing depth and can cause distance measurements along the z-axis to yield wrong results, while spherical aberrations give rise to a non-optimal PSF shape with characteristic long tails and multiple maxima.

Fig. 3 Layer of fluorescent beads under index-mismatched conditions (left column) and corrected with abberior’s adaptive optics module, RAYSHAPE aberration correction (right). Note how brightness and resolution decrease with focusing depth when no correction is applied. All images are scaled to the same brightness.

Modern, high-NA oil objective lenses are designed to take into account the coverslip-immersion medium interface. However, the refractive index (nOIL = 1.518 at 23°C) is assumed to be constant after that, i.e. in the sample. Thus, embedding in Mowiol with a refractive index of 1.40–1.49 will cause aberrations and yet, this is still one of the best mounting medium recommendations for an oil immersion objective, second to embedding the sample in TDE ¹¹, which is not always an option, e.g. for live-cell experiments. Similarly, the reason water immersion objectives lenses exist is that they closely match the refractive index required for live-cell work, and similar arguments hold for glycerol and silicone oil immersion lenses. Nevertheless, despite the close match between these immersion media and the specimens for which they were intended, “close” is often not good enough, and even a slight mismatch can give rise to aberrations.

An additional cause of aberrations are refractive index inhomogeneities in the specimen itself, for example transitions between lipid- or DNA-enriched regions and the rest of the cell. Most aberrations become more severe when focusing deep into a sample, such as thick tissue. This is the reason why multiphoton microscopy typically profits from aberration correction as well.

Optical aberrations and their effects on PSF shapes can be approximated using Zernike polynomials ⁵ to model the corresponding deformation of the wavefronts ¹. The different orders of the polynomials are assigned to known optical aberration modes such that any arbitrary aberration shape can be easily described by decomposing it mathematically into its constituent Zernike modes.

Optical aberrations can be corrected via the implementation of adaptive optics. Before the STED and excitation beams enter the objective lens, they are “pre-aberrated” by an adaptive element that induces the same amount of aberration as the sample, but in the opposite direction. Consequently, when the pre-aberrated beams pass through the aberrating sample, the two sets of aberrations – the first induced the adaptive element and the second by the sample – cancel each other out. Furthermore, the emitted fluorescence, which also gets aberrated as it passes through the sample, is corrected by the adaptive element, too. This way, diffraction-limited conditions are restored at the focus and at the detection pinhole (Fig.2).

Reliable and systematic correction is made possible by the integration of adaptive optical elements such as deformable mirrors into the beam path. Deformable mirrors are coated membranes with a reflective surface whose shape can be controlled by actuators behind the membrane. The aberration compensation is achieved by applying the correct mirror shape, which is half the negative of the distortions introduced by the sample. It’s half the sample distortions, because the rays pick up one half on their way to the mirror and the other half on their way back. A suitable deformable mirror must have extremely high reflectivity from the ultraviolet to the infrared range to avoid losses, and enough actuators (more than 100) so that complex aberrations can be accurately rendered on its surface. Moreover, the response time of deformable mirrors must be sufficiently fast (down to ten milliseconds) so that corrections can be made dynamically within a single image acquisition. When implementing deformable mirrors into a microscope setup, great care must be taken to ensure that the surface shape exactly matches the desired deformation. This involves intricate calibration procedures ⁶ that take into account membrane stiffness, actuator coupling, actuator response and drift, etc.

Deformable mirrors offer several advantages over objective correction collars. For example, deformable mirrors can correct arbitrary aberrations while correction collars can only correct for spherical aberration. In fact, certain aberrations, e.g. sample tilt ⁷, have been shown to compromise the effect of correction collars such that adjusting them makes the resulting image worse rather than better. In addition, deformable mirrors offer much faster response times and thus can be adjusted even between the lines of an image scan. They also have much longer lifetimes due to their non-mechanical nature, and they do not introduce moving parts into the beam path that can cause additional aberrations. Note, that movements of the mirror actuators are miniscule (< 1 µm) compared to movements of lens groups in an objective lens (mm).

Put into practice

Correction of the wavefront distortions has the potential greatly increase signal and resolution. As a rule of thumb, a greater refractive index mismatch, an increased imaging depth and a demand for more (super-) resolution all warrant the use of adaptive optics. While one can get away without aberration correction for confocal imaging close to the cover slip or in a sample embedded in Mowiol using an oil lens, imaging experiments in thick (> 100 μm) samples (Fig. 3, 4) or 3D-STED experiments at just a few microns below the sample surface ⁸, will not yield usable results and call for the use of adaptive optics.

To determine the exact amount of correction to apply on an adaptive optics system, several algorithms have been proposed ^9, 10. Luckily, the most prominent type of aberrations, those caused by a refractive index mismatch, can be predicted using only focusing depth and the amount of refractive index difference. These aberrations increase linearly with depth and can be easily corrected as the sample is refocused, or during a volume or xz-scan, once the user has established them for a certain focusing position. This way, the imaging brightness of fluorescent beads at an imaging depth of 250 μm can be improved by up to a factor of five (Fig.3).

Performing a 3D-STED experiment 100 μm deep inside a complex sample such as Drosophila larvae inevitably requires adaptive optics. In this scenario, the severely aberrated PSF of the STED beam de-excites spontaneous fluorescence everywhere, leaving no useful signal and certainly no resolution. Here, using good adaptive optics makes all the difference between being able to get results or none at all.

¹ M. Schwertner, M. Booth, and T. Wilson, “Characterizing specimen induced aberrations for high NA adaptive optical microscopy,” Opt Express 12, 6540-6552 (2004).

² M. Schwertner, M. J. Booth, M. A. A. Neil, and T. Wilson, “Measurement of specimen-induced aberrations of biological samples using phase stepping interferometry,” Journal of Microscopy
(Oxford) 213, 11-19 (2004).

³ T. J. Gould, D. Burke, J. Bewersdorf, and M. J. Booth, “Adaptive optics enables 3D STED microscopy in aberrating specimens,” Opt Express 20, 20998-21009 (2012).

⁴ S. Deng, L. Liu, Y. Cheng, R. Li, and Z. Xu, “Effects of primary aberrations on the fluorescence depletion patterns of STED microscopy”, Opt Express 18, 1657-1666 (2010).

⁵ V. N. Mahajan, “Zernike circle polynomials and optical aberrations of systems with circular pupils,” Appl Opt 33, 8121 (1994).

⁶ M. Booth, T. Wilson, H. B. Sun, T. Ota, and S. Kawata, “Methods for the characterization of deformable membrane mirrors,” Appl Opt 44, 5131-5139 (2005).

⁷ R. Turcotte, Y. Liang, and N. Ji, “Adaptive optical versus spherical aberration corrections for in vivo brain imaging,” Biomed Opt Express 8, 3891-3902 (2017).

⁸ J. Heine, et al. “Three dimensional live-cell STED microscopy at increased depth using a water immersion objective.” Review of Scientific Instruments 89.5 (2018): 053701.

⁹ M. Booth, D. Andrade, D. Burke, B. Patton, and M. Zurauskas, “Aberrations and adaptive optics in super-resolution microscopy,” Microscopy 64, 251-261 (2015).

¹⁰ M. J. Booth, “Adaptive optical microscopy: the ongoing quest for a perfect image,” Light-Sci Appl 3, e165 (2014).

¹¹ Staudt, Thorsten, et al. “2, 2′‐thiodiethanol: a new water soluble mounting medium for high resolution optical microscopy.” Microscopy research and technique 70.1 (2007): 1-9.

In our knowledgebase you can read a lot about “How the donut changed the world” and the different ways how resolution beyond the diffraction limit can be achieved in fluorescence microscopy, with MINFLUX capable of resolving individual molecules. We talk a lot about optical physics, mathematical functions, and stuff like this. Which is all good and true.

However, what we have not touched on so far is the difference between resolution and localization precision: Superresolution microscopy achieves a resolution in the nanometer range. This resolution, however, applies only to the detected fluorescent molecules. A resolving power of 3 nanometers does not necessarily mean that the proteins labeled with the fluorophores can be localized with equal precision. When your fluorescent dye is too big or too far away from your protein of interest, you will never be able to localize it with a precision that is higher than this offset. But there are ways to reduce the offset between target protein and fluorescent label. And one of these are nanobodies, also called VHH fragments or single-domain antibodies.

Nanobodies are particularly small and simple-built antibodies derived from certain types of antibodies only found in camelids, a group of animals including camels, llamas, and alpacas. To understand the advantages of using nanobodies in superresolution microscopy, we first need to shortly recap how immunofluorescence staining works. For a more detailed introduction to the topic, check out our article “Let the cells shine with immunofluorescence labeling”.

In immunofluorescence staining, the biomolecule of interest – usually a protein – is labeled with a fluorophore with the help of antibodies. Typically, antibodies of immunoglobulin type G (IgG) are used. There are two different labeling strategies (Fig. 1): Direct immunofluorescence staining uses a single antibody which binds to the protein of interest and carries a fluorophore. Indirect immunofluorescence staining, in contrast, requires two antibodies that are applied sequentially: Like in direct immunofluorescence staining, the so-called primary antibody binds to the protein of interest. However, it does not carry the fluorophore. This is introduced by the secondary antibody which binds to the primary antibody.

In most settings, indirect immunofluorescence labeling will be the strategy of choice as it has some advantages over direct immunofluorescence labeling. One is a brighter fluorescence signal when multiple, mostly polyclonal (what’s that again?) secondary antibodies carrying fluorophores bind to a single primary antibody, boosting sensitivity.^1,2

So with indirect immunofluorescence labeling, we get a brighter signal, which is good. However, with indirect immunofluorescence labeling we also get a molecular complex of primary and secondary antibodies with a proud size of 30 nm in diameter, which is not so good: It means that the fluorophore may be something like 20 nm away from the protein. That is the so-called linkage error. It is no issue as long as you are imaging with a conventional confocal microscope with a maximum resolution of roughly 250 nm, where less than ten percent linkage error hardly matters. Superresolution microscopy, in contrast, resolves details of 20 to 30 nm in the case of STED or PALM/STORM; with MINFLUX, we even talk about 2 to 3 nm!

Which brings us back to the initial problem: In superresolution microscopy, a big antibody complex means a significant offset between target and label, limiting the achievable localization precision. Another drawback of such large labeling complexes is that they might block each other’s access to the protein targets when these are densely packed – which is often the case in the busy and complex environment of cells or tissues.^3,4,5

By now, you probably guessed how nanobodies can help here. It’s all in their name: Compared to conventional IgG antibodies, which measure 12 nm in length, nanobodies are tiny at 3 nm. The difference becomes even more obvious when we look at the molecular weight (this is a measure for the mass of a molecule, given in the quantity Dalton): An IgG antibody (150 kDa) outweighs a nanobody (12 to 15 kDa) by a factor of 10.²

Nanobodies are so small because practically they are just fragments of the camelid antibodies, which already have a simpler structure than conventional antibodies: Conventional antibodies consist of four peptide chains, two heavy and two light ones, and the region that binds the target structure – the antigen-binding domain – is composed of two chain regions, the variable regions of the heavy chain (VH) and of the light chain (VL). Camelid antibodies lack the light chains; consequently, their antigen-binding domain consists of a single region only, called variable heavy region of the heavy chain (VHH). Nanobodies are derived from this VHH (Fig. 2).² So, in fact, a nanobody is an antibody reduced to the essentials.

Due to their small size, nanobodies can significantly reduce the linkage error. But how much? And how can they be employed best? One option is to replace the fluorophore-coupled secondary antibody with a polyclonal secondary nanobody. This nanobody can bind to multiple sites of the primary antibody, including epitopes that cannot be recognized by conventional secondary antibodies. As a result, signal amplification remains high. The complex of primary antibody and secondary nanobodies is only slightly larger than a single antibody (Fig. 3).

Thus, with polyclonal secondary nanobodies, the linkage error is reduced while strong fluorescence signal is maintained. For instance, tubulin fragments stained with secondary nanobodies coupled to abberior STAR RED have a significantly smaller apparent filament diameter (67 nm) than tubulin filaments stained with secondary antibodies (77 nm) (Fig. 4). The linkage error can be reduced even further when fluorophore-coupled nanobodies are used for direct immunofluorescence. However, here the same limitations apply as for direct immunofluorescence with conventional antibodies: no signal amplification and a limited availability of suitable nanobodies.

Increasing resolution can open up opportunities unavailable to biological research before, as we discuss in “Superresolution for biology: when space, time, and context matter”. Another screw one can adjust in fluorescence microscopy to get more information out of ones image is the number of different proteins labeled in a single sample. Here, as well, a lot depends on the antibodies used.

As detailed above, indirect immunofluorescence staining is usually a two-step process where the sample is first incubated with the primary antibody and subsequently with the secondary antibody. This limits the number of target proteins that can be labeled since every primary antibody has to originate from a different host species to exclude cross-reactivity with the secondary antibodies. In theory, this cross-reactivity could be avoided by pre-incubating every individual primary antibody with its specific secondary antibody before adding the pre-assembled complex to the sample. However, this approach is not feasible when using conventional secondary antibodies as these are polyvalent (not to be confused with polyclonal), meaning that every antibody has two antigen-binding domains (one at each arm’s end). Consequently, it can bind to two target primary antibodies, which is also what happens in case you pre-incubate the two together. And as not only one, but many secondary antibodies do so, this results in the formation of huge antibody aggregates that are no longer able to deliver a specific immunostain.

Again, nanobodies are the solution, because they differ from conventional antibodies not only in size, but also in structure: As nanobodies consist of a single antigen-binding domain, they bind only a single target, i.e. they are monovalent and can easily be pre-mixed with primary antibodies (Fig. 5). Et voilá, antibody aggregates are history.

With secondary nanobodies, it is now possible to use several primary antibodies raised in the same species to label a variety of target proteins, a strategy called multiplex staining.³ A 3-color STED image with excellent multicolor staining is shown in figure 6.

Nanobodies have quickly developed into versatile tools in molecular biology. For fluorescence microscopy, they already solve more than one problem. And there is more to come, for sure.

Naturally, a normal cell or tissue sample does not come along with fluorophores, meaning that you have to introduce them into the specimen before acquiring images. There are different strategies to bring the dye to the sample, e.g. co-expression of a fluorescent protein or click chemistry. The most versatile and therefore most common strategy, however, is immunofluorescence. In case you always wanted to know how immunofluorescence works and which properties of antibodies make it so powerful and at the same time define its limits, read on!

In immunofluorescence labeling, the fluorophore is brought to the biomolecule of interest, usually a protein, with the help of antibodies. And since antibodies play such a central role here, it is worth to take a closer look at them before diving into the technical aspects of immunofluorescence.

Antibodies or immunoglobulins are an integral part of the immune system. They are produced by specialized immune cells, the B-cells, when these encounter a foreign molecule, e.g. some part of a virus or bacterium. The antibodies help the immune system to recognize and fight this molecule and the pathogen it belongs to. Two properties make antibodies perfectly suited for this task: specificity and affinity, meaning that they are particularly good at finding exactly their counterpart, and once they have bound to it they never let go again. For these reasons, antibodies are great not only for our immune system to identify and label intruders, but also in biological research to detect, mark, or isolate proteins of interest.

Most people will probably recognize an antibody depiction when they see one, since it is so familiar: Antibodies are Y-shaped proteins, formed by four chains: two so-called heavy chains and two light chains (Fig. 1). The antibody’s stem and part of the arms are identical for all antibodies of one class – e.g. immunoglobulin type G (igG), the most common antibody – and are therefore together called the constant region. The outermost part of the arms, however, differs from antibody to antibody. It is called the variable region. The variable region of the heavy chain (VH) and the variable region of the light chain (VL) together form the antigen binding domain. It is this part of the antibody which recognizes and binds the target structure (“epitope”), hence determining the antibody’s specificity.¹

For scientific applications, antibodies are produced in animals, mostly mice, rats, rabbits, or bigger animals like donkeys and goats. Naturally, an animal’s B-cells react to a foreign molecule by starting mass assembly of antibodies for tagging it. We can hijack this immune response simply by injecting an animal with a protein of interest, whereupon the animal’s B-cells will happily produce antibodies specifically against this protein. Exactly what we need!

An important differentiation for antibodies used in immunofluorescence is their clonality: They can either be mono- or polyclonal (Fig. 2). The antibodies produced by an individual B-cell are all absolutely identical, meaning that they are of the same type (e.g. IgG) and all recognize and bind to the same epitope.

When a pool of antibodies is produced by clones of this B-cell, they are said to be monoclonal, meaning all of them recognize the same epitope. Imagine having several keys for the same building, which are exactly the same and which all open the exact same lock. The second option are antibodies which recognize the same protein, but different epitopes of it, because they originate from a pool of different B-cells. These antibodies are called polyclonal. This would be akin to having keys for the same building (as in “same protein”), but they all “recognize” different locks on different doors.

If you know that someone has a “monoclonal” key, you know exactly which door he used to get in as the key will fit in this door’s lock only. Similarly, monoclonal antibodies exhibit superior specificity and very low cross-reactivity, i.e. they are unlikely to bind to anything other than their epitope. In immunofluorescence, this results in low background noise. However, they give rather low signal, since only one antibody can bind per molecule, and are more expensive to produce. Moreover, as they are all the same, when attaching a fluorescent label happens to change their binding behavior, it will do so for all antibodies, meaning they will all show decreased functionality.

In contrast, someone with a “polyclonal” key set has more flexibility as to which door she uses, but also, you can’t be so sure which one it is. Correspondingly, polyclonal antibodies are comparably inexpensive, exhibit a higher signal as they can bind more than one epitope of the target protein, and tagging with a fluorophore is unlikely to affect the binding behavior of all of them. On the other hand, the properties and quality of polyclonal antibodies may vary between badges and they are more likely to cross-react with proteins other than the one of interest, producing more background noise.¹

With this information in mind, let’s go back to immunofluorescence staining. There are two different labeling strategies (Fig. 3): Direct immunofluorescence staining uses a single antibody which binds to the protein of interest and carries a fluorophore. Indirect immunofluorescence staining, in contrast, requires two antibodies that are applied sequentially: Like in direct immunofluorescence staining, the so-called primary antibody binds to the protein of interest. It mostly originates from mice or rabbits. However, it does not carry the fluorophore. This is introduced by the secondary antibody, which is usually produced in donkey, goat, or rabbit by immunizing these animals with antibodies from mouse or rabbit. The primary antibody binds to the constant region of the primary antibody.

At first glance, direct IF labeling appears much more attractive as it offers an easy and fast staining protocol. Why not go for it? There are two main drawbacks: First, the number of commercially available fluorophore-coupled primary antibodies is limited as for every protein an individual antibody has to be produced and coupled to a fluorophore. And chances are that none of the antibodies available is directed against your protein of interest. Second, in direct immunofluorescence labeling any target protein will be bound by maximum one fluorophore-coupled antibody and will therefore produce a comparably weak fluorescent signal.

For these reasons, indirect immunofluorescence labeling is usually the strategy of choice. Here, two monoclonal or even more polyclonal antibodies carrying fluorophores bind to a single primary antibody, which amplifies the signal and boosts sensitivity. ^2,3 This way, the primary antibody can focus on recognizing a very specific target of interest, while the secondary antibody merely has to be targeted against the primary antibody.

With indirect immunofluorescence researchers have a particularly versatile technique at hand to label proteins with high specificity and sensitivity. The most important advantage when using antibodies in immunofluorescence is that you can target virtually any protein as long as there is an antibody at hand that recognizes this target.

So much for the advantages. But every method has its limits, doesn’t it? Indeed this is also the case for immunofluorescence. For one thing, the number of different proteins one might label in a single sample is restricted, because for every primary antibody used you need a different host species. And, as primary antibodies are usually generated in rabbits and mice, it is usually not possible to stain more than two proteins at once: the secondary antibody would not be able to differentiate between e.g. two mouse antibodies bound to two different target proteins and you would consequently have no chance to label these two targets with different fluorophores.

For another thing, coming in at around 30 nm, the size of the complex of primary and secondary antibodies might pose a problem: In the tightly packed environment of a cell, the complexes might block each other’s access to epitopes. Moreover, their size means that the fluorophore might be displaced from the labeled protein by something like 20 nm. As long as resolution in light microscopy was restricted to the Abbé limit of about 250 nm, this so-called linkage error was negligible. With the invention of superresolution microscopy, however, 20 nm offset between target and label mean that otherwise sharp images get blurred.

Luckily, a new type of antibody provides a solution for both problems. But this is the topic of another article.

Virus particles vary in size across different virus species. One of the smallest viruses, the parvovirus, has a particle size of approx. 20 nm while the largest known virus particles of the Mimivirus family possess a capsid size of up to 500 nm, which already compares to the size of a bacterium (Fig. 1). Therefore, due to their size, virus particles and their substructures can be studied in detail using superresolution microscopy methods, such as stimulated emission depletion (STED) microscopy or single-molecule localization methods, which offer resolution capabilities of about 20 nm. Numerous studies using super resolution imaging techniques have already expanded the scientific knowledge of the Human Immunodeficiency Virus (HIV-1) ^{1, 2}, Influenza A virus ^{3, 4}, Herpes Simplex Virus (HSV) ⁵, Respiratory Syncytial Virus (RSV) ⁶, Vaccinia Virus ⁷, Adeno-Associated Virus (AAV) ⁸, Nipah Virus (Niv) ⁹, Adenovirus ¹⁰ and many more.

In this article, we provide a brief overview of superresolution microscopy, focusing on STED microscopy and its contributions to virus imaging and virus research. Developments in STED microscopy instrumentation are rapid and, while the technique has already seen widespread use in scientific research, we are confident that it’s potential can only increase in the future. One should not forget that the first ever images on a home-built STED microscope were acquired only less than two decades ago. Since then, tremendous amounts of engineering and innovation have made STED microscopy what it is today: a reliable, easy-touse super resolution imaging technique that can be operated as intuitively as a confocal microscope, by everyone.

The potential of superresolution microscopy extends far beyond STED. The novel technique MINFLUX ²² offers isotropic resolutions up to 2 nm – the size of a single fluorescent molecule. Already STED microscopy was able to change the way we look at virus particles, so the impact MINFLUX will have on scientific research cannot be underestimated. This new level of resolution is expected to have a tremendous impact on virology research. MINFLUX tracking can be performed at extremely high speeds with true molecular precision, making it possible to observe rearrangements inside a single protein over time, or to follow single virus particles over a long time with unprecedented precision and speed.

Stimulated emission depletion (STED) microscopy delivers resolutions better than 20 nm. Due to significant progress in STED instrumentation and the commercialization of new, photostable dyes, STED is now widely accessible. However, although STED offers unlimited resolution in theory, photobleaching comes into play in practice. Mediating its effects can require tuning down the resolution or recording fewer time steps and smaller 3D-volumes than desired. Besides adaptive illumination, a solution to circumvent the physical limitations of current labeling technology is the adaptation of the PAINT-concept (point accumulation for imaging in nanoscale topography) to STED microscopy. PAINT is based on the application of exchangeable labels that only temporarily bind to their target structures. During acquisition, these labels are constantly replenished from a large pool in the imaging medium, providing stable sample brightness even at the highest resolutions.

The resolution of a STED microscope scales with the intensity of the STED laser. Unfortunately, an increase in resolution is therefore often connected to an increase in photobleaching. The available fluroescence photon budget thus can hinder time lapses as well as volume imaging. This is because with time lapse imaging, the sample area is illuminated multiple times and the number of fluorophores is concomitantly reduced each time. Similarly, when imaging volumes, adjacent z-planes are imaged in succession, so that each plane is subjected to multiple rounds of illumination and therefore photobleached to a certain extent each time. As a result, microscopists often need to sacrifice either resolution, the number of time steps or the number of z-planes recorded in order to reduce photobleaching to an acceptable level.

The situation described above was particularly true for early implementations of STED microscopy, when very strong sample illumination met a small number of photo-stable fluorophores. Today, major developments in STED instrumentation, such as the use of optimized pulsed STED lasers ¹, highly efficient detection via APDs, and Adaptive Illumination ^{2, 3, 4} now allow a strong reduction of sample illumination. Currently, state-of-the-art STED microscopes readily offer resolutions reaching <20 nm and allow the acquisition of live cell movies and image stacks. At the same time, a vast range of optimized, bright, and photostable fluorophores for STED imaging is now commercially available and comes with many conjugation chemistries for diverse applications ^5-9.

Nevertheless, the achievable resolution and number of image frames are still ultimately limited by photobleaching, albeit at a much higher level than only a few years ago.

On the other hand, PAINT allows to suppress or bypass the photobleaching process and therefore offers additional resolution and extended time-lapse or volume imaging. With conventional PAINT microscopy, individual, transiently binding dyes are located precisely using single molecule localization microscopy ^{10, 11}. Most importantly, fluorophores are constantly exchanged: bleached fluorophores are replaced with fresh ones from the large reservoir of the imaging medium using labels with transient binding properties. This allows for fast and continuous exchange of fluorophores during image acquisition.

Recently, Spahn et al. ¹² have successfully adopted exchangeable fluorophores for use with STED microscopy. In contrast to classical PAINT, which requires very few dyes bound at a single time, staining for STED was optimized to provide high densities of bound dyes while allowing fast replenishment from the imaging medium. Interestingly, this can be implemented with a large range of staining techniques, including simple lipophilic dyes, DNA stains, toxin- and peptide-conjugated dyes as well as with immunostaining with DNA-labeled antibodies.

PAINT is based on special labeling conditions under which dye molecules are constantly exchanged. This means that bleached dyes are constantly replaced with fresh ones, allowing repeated scanning of the image area and accumulating signal until a bright image has been formed.

The kinetics of dye exchange determine the exact imaging mode and speed, because bleached dyes need to be replaced before the next accumulation step is started. This results in two different imaging approaches for fast and slow dye exchange rates. It is important to note that a larger image area reduces the overall acquisition speed, which in turn provides more time for the dyes to be replenished between repeated scans. Essentially, a larger image area has a similar effect as faster dye exchange rates.

Approach 1: Fast Dye Exchange Rate

For dyes with a fast exchange rate, nearly full recovery of the stain will occur while a single line is scanned during the acquisition of a STED image. This provides a constant, high signal in each line scan and allows for multiple line accumulations before moving to the next line. Note that for larger images, each line scan takes longer, resulting in improved stain recovery. The brightness of the final image is generally high and depends mostly on the number of line accumulations. This approach is suited for time lapse acquisition in live cells with stable intensity over extended periods of time. The achievable scan speed depends on the staining brightness, the dye exchange rate, and image size.

Pros and cons 

+ Best suited for large images with high resolution

+ Compatible with live cell imaging

+ Robust against stage drift

– Slow total imaging speed requires autofocus

Approach 2: Slow Dye Exchange Rate

Stains with slow exchange rates are more suitable to repeated acquisition of full image frames without line repeats. As full frame acquisitions take longer than individual line scans, more time is available for a complete replenishment of the dye before the next frame starts. Usually, after several scan cycles equilibrium is reached and bleaching and recovery balance out. Then, the signal in each frame is accumulated, for the bright final image. A drawback is potential motion-blurring due to the longer signal accumulation time compared to a quick succession of line accumulations with Approach 1. This effect becomes less noticeable in smaller images due to faster frame acquisition.

Pros and cons 

+ Best suited fixed cells with low staining intensity 

+ Small organisms like Bacteria

+ Motion blur: Less suited for live cell imaging and sensitive to stage drift

– Slow total imaging speed requires focus lock

A crucial requirement for the STED-PAINT approach is the exchangeability of dye molecules. While traditional immunolabeling is not directly compatible with PAINT, other common stains readily exhibit transient binding and unbinding, allowing dye exchange between a bound fraction and the surrounding buffer. This includes fluorescent dyes conjugated to toxins or drugs that readily stain structures such as DNA, tubulin, actin, lipophilic dyes for membrane staining and others (Tab. 1). Depending on the cell-permeability and toxicity of the label, both live and fixed cell imaging is possible with the presented approach.

In addition, it is beneficial to use fluorogenic dyes, i.e. dyes which remain non-fluorescent in solution and turns fluorescent only when bound to the target. This reduces the background signal originating from the imaging buffer. However, many non-fluorogenic stains with high target affinity can also provide good signal-over-background.

Popular examples for dyes compatible with live-cell STED microscopy are the commercially available silicon rhodamine (SiR) or abberior LIVE dyes.

Stains exhibit large differences in binding behaviour with different on- and off-rates. In addition, in live-cell applications, the appropriate stain concentration may be limited, reducing replenishment from the imaging buffer pool. As an example, SiR-tubulin stains show rather slow replenishment at concentrations compatible with overnight imaging (often around 0.1 – 1 µM, lower if spindle apparatus is observed), allowing repeated acquisition of STED imaging with stable intensities every 2-5 minutes. As rule of thumb, labels with KD values in the lower µM-range (_~ 1 – 10 µM) and off-rates of 1 – 100 s-1 should be well suited for STED-PAINT applications. Therefore, a large number of stains are potentially compatible with STED-PAINT and many more applications can be expected in the near future.

We thank Prof. Dr. Mike Heilemann and Dr. Christoph Spahn for very fruitfull cooperation and support with STED-PAINT samples and imaging. 

Labeling for

Since the discovery of the diffraction barrier of light microscopy in the late nineteenth century¹, it has been accepted that conventional far-field optical microscopy cannot resolve structural details finer than half the wavelength of light. Superresolution fluorescence microscopy techniques, such as stimulated emission depletion (STED) microscopy^{2, 3}, overcome this limitation. Here, we present robust sample reparation protocols for STED microscopy on mammalian cells. These protocols can be used as a starting point to adapt existing labeling strategies to the requirements of STED microscopy.

Protocol I

Immunofluorescence-labeling of cultivated adherent mammalian cells – methanol fixiation

1. Cultivation of cells

Cells are typically seeded on cover slips for 12–36 h before labeling (Note 3, 4)

2. Fixation and blocking

Sample is fixed for 5 min with ice-cold methanol (abs.) with cells facing upwards (Note 5, 6, 7, 8, 9, 10). Then, cover slips are washed with PBS (Note 11, 12, 13). Finally, unspecific binding sides are blocked with PBT for >15 min at RT, e.g. in a petri dish (Note 14).

3. Incubation with primary antibodies

Coverslips are removed from the petri dish and excess PBT is drained by placing the edge of the coverslip on a piece of tissue. Then, coverslips are placed in a humid chamber, covered with 25 µl to 100 µl (depending on the diameter of the coverslips) of diluted antibody solution (in PBT) (Note 15), and incubated for 1 h at RT.

4. Washing

The coverslips are removed from the humid chamber. The antibody solution is drained by pipetting off the liquid and then placing the edge of the coverslip on a piece of tissue. Then, the coverslips are placed in a fresh petri dish containing PBS (Note 13; 16) followed by incubation > 5 min at RT (2x). An additional blocking step with PBT is possible after this step (optional).

5. Incubation with secondary antibodies

Coverslips are removed from the petri dish and excess PBS is drained by placing the edge of the coverslip on a piece of tissue. Then, coverslips are placed in a humid chamber, covered with 25 µl to 100 µl (depending on the diameter of the coverslips) diluted antibody solution (in PBT) (Note 15), and are incubated for 1 h at RT.

6. Washing

The coverslips are removed from the humid chamber. Antibody solution is drained by pipetting off the liquid and then placing the edge of the coverslip on a piece of tissue. Then, coverslips are placed in a fresh petri dish containing PBS (Note 13, 16, 17). Incubate for at least 5 min at RT (3x).

7. Embedding, Storage, Stability

Finally, the coverslips are removed from the petri dish; excess PBS is drained by placing the edge of the coverslip on a piece of tissue. Then, the coverslips are mounted using the favored embedding medium.

1. Cultivation of cells

The cells are typically seeded on cover slips for 12–36 h before labeling (Note 3, 4)

2. Fixation, Extraction and Blocking

Coverslips are fixed with formaldehyde solution with cells facing upwards (Note 5, 6, 7, 8, 9, 10) for 5 min. Then, cells are extracted using 0.1 – 0.5 % Triton X-100 in PBS (Note 11) also for 5 min. The coverslips are washed with PBS (Note 12, 13). Finally, unspecific binding sites are blocked with PBT for >15 min at RT (Note 14) in a petri dish.

3. Incubation with primary antibodies

Coverslips are removed from the petri dish; excess PBT is drained by placing the edge of the coverslip on a piece of tissue. Then, coverslips are placed in a humid chamber, covered with 25 µl to 100 µl (depending on the diameter of the coverslips) diluted antibody solution (in PBT) (Note 15), and are incubated for 1 h at RT.

4. Washing

The coverslips are removed from the humid chamber. The antibody solution is drained by pipetting off the liquid and then placing the edge of the coverslip on a piece of tissue. Then, the coverslips are placed in a fresh petri dish containing PBS (Note 13, 16). Incubation for at least 5 min at RT (2x). An additional blocking step with PBT is possible after this step (optional).

5. Incubation with secondary antibodies

Coverslips are removed from the petri dish; excess PBS is drained by placing the edge of the coverslip on a piece of tissue. Then, coverslips are placed in a humid chamber, covered with 25 µl to 100 µl (depending on the diameter of the coverslips) diluted antibody solution (in PBT) (Note 15), and are incubated for 1 h at RT.

6. Washing

The coverslips are removed from the humid chamber. The antibody solution is drained by pipetting off the liquid and then placing the edge of the coverslip on a piece of tissue. Then, the coverslips are placed in a fresh petri dish containing PBS (Note 13, 16, 17). Incubation for at least 5 min at RT (3x).

7. Embedding, Storage, Stability

Finally, the coverslips are removed from the petri dish; excess PBS is drained by placing the edge of the coverslip on a piece of tissue. Then, the coverslips are mounted using the favoured embedding Medium. In addition, mounted coverslips may be fixed with nail polish completely or at several small points (Note 18, 19).

8. Storage of samples

Ready-made samples should be stored at 4 °C (Note 21). Most samples are stable for a rather short time only. For best results (Fig. 5), samples should be imaged within one week.

1. Cultivation of cells

The cells are typically seeded on coverslips for 12–36 h before labeling (Note 3; 4)

2. Fixation, Extraction and Blocking

Samples are fixed with formaldehyde solution with cells facing upwards (Note 5, 6, 7, 8, 9, 10) for 5 min. Then cells are extracted using 0.1–0.5 % Triton X-100 in PBS (Note 11) for 5 min. The coverslips are washed with PBS (Note 12, 13). Finally, unspecific binding sides are blocked with PBT for >15 min at RT (Note 14) in a petri dish.

3. Incubation with fluorophore-labeled phalloidin

Coverslips are removed from the petri dish; excess PBT is drained by placing the edge of the coverslip on a piece of tissue. Then, coverslips are placed in a humid chamber, covered with 25 µl to 100 µl (depending
on the diameter of the coverslips) diluted phalloidin solution (in PBT; Note 15), and are incubated for 1 h at RT.

4. Washing

The coverslips are removed from the humid chamber. The antibody solution is drained by pipetting off the liquid and then placing the edge of the coverslip on a piece of tissue. Then, the coverslips are placed in a fresh petri dish containing PBS (Note 13, 16). Incubation for at least 5 min at RT (2x).

5. Embedding, Storage, Stability

Finally, the coverslips are removed from the petri dish; excess PBS is drained by placing the edge of the coverslip on a piece of tissue. Then, the coverslips are mounted using the favoured embedding medium. In addition, mounted coverslips may be fixed with nail polish completely or at several small points (Note 18, 19).

6. Storage of samples

Ready-made samples should be stored at 4 °C (Note 21). Most samples are stable for a rather short time only. For best results (Figs. 5, 6), samples should be imaged after not much more than one week.

The most important rule for immuno-labeling is that specimens must not be allowed to dry out.

1. Dissolution of fluorphore-labeled phalloidin: The vial contents should be dissolved to yield a final concentration of 200 units/ml, which is equivalent to approximately 6.6 µM.
2. Certain embedding media may affect the photophysical properties of some fluorophores, leading to changes in brightness and bleaching behaviour, for example. A shift in the excitation and emission spectra is also possible.
3. Seeding of cells may take place earlier if required. Depending on the doubling time, cultivation time, cell density and others, cells may grow in microcolonies.
4. Cells grown in very high densities, i.e. a confluent layer, may lead to high background signal during imaging.
5. In contrast to other protocols, specimen/cells are not washed with buffer or similar before fixation.
6. Cells are not subjected to a pre-extraction procedure before fixation.
7. The fixation is performed using an excess of fixative.
8. For fixation, the coverslips are placed into a fresh petri dish containing fixative. Alternatively, fixation may be performed in the same petri dish the cells were grown in.
9. For fixation, the growth medium is removed completely. Then, cells are submerged in fixative.
10. The fixation should be performed with freshly prepared or frozen fixative. It is best to use prewarmed fixative. Ready made 37% formaldehyde/ methanol should not be used.
11. Between fixation and blocking, an extraction step may be required. For formaldehyde fixation it is crucial to extract the cells using detergents. For methanol fixed cells, extraction is not required, however a short wash in PBS/ 0,1% Triton X-100 may be advantageous for the labeling.
12. Coverslips with fixed cells may be stored several (1–2) days at 4 °C in PBS. However, the quality of the labeling may be affected by the storage of the samples.
13. Washing should be performed using excess PBS.
14. Blocking should be performed using excess PBT solution.
15. For antibody incubation, coverslips should not be placed with cells facing downward on parafilm or similar films. Although this might reduce the amount of antibody needed for labeling, cells may be affected while removing them from the parafilm.
16. After antibody incubation, coverslips should be washed in different petri dishes. Otherwise cross contaminations might occur.
17. If high background labeling of the specimen occurs, it can be further reduced by incubation with PBS/ Triton X-100 and/or PBT.
18. If non-hardening embedding media are used, coverslips have to be sealed using Twinsil or nail polish.
19. If TDE is used as embedding medium, twinsil may not be used since the TDE will hinder its polymerization and hardening.
20. The use of TDE is not possible for phalloidin labeled cells.
21. Storage of samples at -20 °C should be avoided, because ice crystals may affect the quality of the samples.

BSA
DMF
DMSO
PBS
PBT
RT
SDS
STED
TDE

Bovine Serum Albumin
N,N-Dimethylformami
Dimethylsulfoxide
Phosphate buffered saline
PBS/ BSA/ Tween 20
Room Temperature
Sodium Dodecyl Sulfate
Stimulated Emission Depletion
2,2′-Thiodiethanol

Biological research has penetrated deep into the realm of the smallest. Today, we study individual cells rather than average their behavior across tissues. We identify unknown microbial species based on novel DNA sequences alone. We’ve even explored quantum underpinnings of photosynthesis and the magnetic compass of migrating birds. The resolution with which we see life has made leaps and bounds, in no small part due to advances in microscopy. But when does superresolution have the greatest impact on how we understand life?

Fascinating biology takes place at scales invisible to the naked eye. And with molecular biology driving this century’s boon in biotechnology and biomedicine, it is no understatement to say that superresolution matters in biology. Fact is that the interactions and processes of life transpire at scales below the diffraction barrier.

A range of light microscopy technologies reach and surpass the light diffraction barrier of about 200 nm, among them stimulated emission depletion (STED) and MINFLUX. As methods, they vary widely in complexity of setup, sample preparation, data processing after image capture, and capacity for live-specimen imaging. Together, however, they grant access to the nanoscale world increasingly nearing the resolution of an electron microscope. Thus, today a biologist who needs to visualize a cellular structure or molecule in the low nanometer scale can contemplate numerous alternatives to highly specialized methods that require expertise, specialized equipment, time, and sample destruction. As we’ll see, those alternatives are essential to the exploration of biology in more ways than one.

For decades, electron microscopy dominated with mind-blowing images of subcellular components and molecules. Now however, superresolution technologies make that detailed imagery standard in light microscopy. Consider mitochondria. The folds of their inner membrane – or cristae – is where they synthesize ATP. Mitochondria and their cristae are highly dynamic, adjusting in shape and quantity. Visualizing cristae can help characterize how those changes correlate with the metabolic state of a cell but requires a resolution far below the average distance between cristae (_~100 nm). Using newly developed fluorescent dyes, researchers have applied stimulated emission depletion (STED) microscopy to image the characteristic stripes of mitochondria at resolutions as high as 35 nm. With MINFLUX, resolutions as high as 2 nm can be reached, which allows to tell apart individual molecules.

Size and distance measurements are important to characterize subcellular environments, but with them alone we cannot unravel life in all its facets. Life is about movement and interactions, and the timescales at which those happen span from nanoseconds for protein conformational changes to minutes for DNA replication. Clearly, the temporal resolution of imaging is decisive in making meaningful biological observations at the molecular level. Electron microscopy meets its limits here as samples need to be fixed and cannot provide any information about temporal correlations.

A handful of superresolution methods accommodate the temporal resolution required to observe processes of life (Fig. 2). STED, for example, offers fast imaging speed (roughly 1 frame per second). Using photostable dyes that endure long-term imaging, Yang and colleagues documented changes in the shape of individual mitochondria over a 10-minute and a 60-minute timeframe. The movies show the formation of bubbles in mitochondria undergoing fission. Those bubbles fill rapidly with cristae and then separate to form individual small mitochondria.¹ Other studies have used superresolution methods to track proteins, like the movement of AMPA receptors in neurons, the translocation of membrane-bound glycophosphatidylinositol (GPI) molecules, and even the diffusion of Hfq protein as it binds mRNA in E. coli. Finally, MINFLUX takes it to the next level: Thanks to its photon efficiency, it can resolve molecular movements with a temporal resolution in the range of 100 µs while providing highest spatial detail, making it possible to track individual lipids diffusing in biological membranes or the conformational changes of kinesin-1 as it walks along microtubules. ^3,4,5

The long-term imaging of biological processes at high resolution is not just about imaging speed. It’s also about keeping your sample alive. That is, after all, the context of biological research. The difficulty here lies in preventing specimen destruction through preparation techniques or the toxic effects of exposing live cells to high-energy or lengthy excitation cycles. Creative solutions to this problem can be found in the literature. For example, event-triggered STED limits photobleaching and toxicity by linking automated STED imaging to the detection of a cellular event, like protein recruitment or vesicle trafficking. ⁶ In a particular exciting demonstration of live imaging, STED microscopy was used to record dynamic changes in dendritic spines of cortical somatosensory neurons in a mouse. The mouse was alive during the procedure and the fine details of these neural structures were revealed with a spatial resolution of under 70 nm.⁷

Looking into the future, our understanding of biological processes to leverage them for applications will increasingly rely on methods that offer more than exceptional spatial resolution. They must also match the time scale of biological events and capture molecular interactions in vivo. Advances in superresolution microscopy are at the cusp of unifying those three dimensions into one easy-to-use technology. Our view of life is looking clearer than ever before.

In other articles on resolution – “What is resolution” and “How to measure resolution” – we talk about how a wide point spread function (PSF) reduces the resolution of a microscope as its blur makes two points indistinguishable in the resulting image. The physical width of the PSF cannot be smaller than the diffraction limit but superresolution technologies circumvent that boundary to generate an effective PSF width down to 20 nm or less. One workaround strategy, relies heavily on high photon flux from the fluorophore to localize it based on a likelihood fit to the shape of its PSF. That is, more photons per emitter is good. Another workaround strategy narrows the width of the PSF by restricting the area where a fluorophore can emit. A narrower full width at half maximum (FWHM) of the PSF is good.

The interaction between these two parameters – FWHM and photon number per emitter – is worth a closer look.

Take two emitters separated by a distance of 150 nm and image them multiple times. Now, let’s look at what happens to the intensity profiles of their PSFs when we alter the FWHM and the number of photons emitted by the fluorophores. In the graphic below, each colored line is a separate imaging run.

A low photon emission rate results in a noisy PSF, dramatically limiting how well the two emitters can be discriminated. That noise, however, becomes less relevant in separating the two PSF when you have a narrower FWHM. Look at the graphic from left to right. What is surprising is that for a small FWHM, only a handful of photons are sufficient to clearly distinguish the two PSFs. Perhaps an alternative: The narrow FWHM teases apart the signals of the two fluorophores despite the low and inconsistent detection of photons between runs.

Resolution also improves as the number of photons detected from each emitting fluorophore rises. Look at the graphic from top to bottom. The individual measurements become more consistent to reveal a distinct dip between the two PSF intensity profiles. It can be clearly seen that if the FWHM is too large, even a higher number of photons than in our example will no longer improve the resolution.

Key, however, is that the discrimination power of the microscope improves more readily with the narrower FWHM, even with lower photon flux from the fluorophores. These are two levers at your disposal to reach higher resolution. Which will give you greater access to the nanoscale world? As so often in life, it depends!

One thing is clear, however: the narrower the FWHM, the fewer photons are needed for a high-resolution image.

The resolving power of a microscope is its ability to discriminate two points in a sample. Standard criteria equate resolution with the width of the point spread function (PSF) that blurs details in images. Practically speaking, those measurements are done on the intensity profile of a point’s PSF, taking the full width at half maximum peak or the FWHM (Figure 1)¹ . Modern microscopes, however, reach unprecedented resolution levels. At those scales, variability is a dominant force and that has consequences for how we measure resolution. For more information take a look at ”Part one – What is resolution?“

The FWHM is a fair approximation of resolution, especially when the point imaged is substantially smaller than the PSF and the signal of that point vastly outweighs noise and background. But at high resolution, neither condition really holds. As resolution goes up, you have to work with increasingly smaller points, which in the case of fluorescence leads to notoriously dim signals that are difficult to distinguish from background. In fact, the flux of photons detected by superresolution microscopes is awash in a sea of variability that distorts the signal itself and obfuscates measurements. There’s noise from the instrumentation and there’s background, like fluorescence from optical components and from molecules outside of the plane of interest. There’s also noise from stochastic fluctuations in photon detection which garbles PSF intensity profiles.

Things get messy at high resolution. And thus, measurements become a mission to minimize variability.

The three horsemen of uncertainty

A thought experiment: you image a random distribution of fluorescent points to measure the resolution of a microscope. Three attributes of your sample play a role in your measurement: the distances between the randomly distributed points, the size of the PSF that each point casts, and the brightness of each point. Why? Because they can all inject variability into your measurements. Depending on how close they are to one another, fluorescent dyes can influence each other’s emission (e.g., quenching) and consequently alter the size and shape of the PSF. Differences among PSF measured – due to point size variation, for example – decreases the reproducibility of measurements. And last but not least, dim florescent points emit fewer photons making PSF intensity profiles more vulnerable to muddling by detection noise.

To get ideally accurate and precise measurements of resolution, you would do well to control these attributes. Go ahead. Stop the thought experiment and throw out the specimen you used. Now, go get a standardized fluorescent dye consisting of probes that are highly reproducible in size, shape, brightness, and distance from one another. These “standard probes” should be small so that what you measure is to the greatest extent possible just PSF. The standard probes should also be bright – emit lots of photons – to overcome detection noise and generate reliable PSF intensity profiles.

OK. Hand over those standard probes, you say.

Yeah, well. It’s not that easy. The truth is none of the currently used standards manages to mitigate variability of all three attributes. Table 1 lists standard probe types used for resolution determination. Each has features that make it more appropriate for certain but not all microscopy systems.

At high resolution, point-like emitters like single molecules are a suboptimal standard. They emit a weak fluorescent signal that can be obscured by background and noise. They randomly switch between bright and dark states – something called fluorescence intermittency or blinking – and their emission is influenced by their dipole orientation. Taken together, this means that the emission intensity of a point-like emitter is generally irreproducible making resolution measurements unreliable.

Let’s take a look at two alternatives.

One option is to go a bit larger than point-like emitters: use a fluorescent bead of known size that is smaller than the resolution of the microscope. These beads are highly standardized and packed with dye molecules that together emit strongly for a good signal-to-noise ratio. The spherical shape of the bead and the high number of emitters also eliminate any orientation effects. Finally, bead size is held within a narrow diameter range and thus, this standard probe type generates highly consistent PSF.

Those are the advantages. An inconvenience, perhaps, is precisely that size. Because a bead has a diameter, its image PSF is a composite (the convolution) of the bead itself and the effective PSF of the microscope (Figure 2). We can measure the FWHM of the image PSF. However, it is the FWHM of the effective PSF that gives us the microscope resolution. That means that we have to figure out the width of the effective PSF from the image PSF by accounting for the size of the bead.

Luckily, that calculation has been worked out for STED. In cases where the bead is much smaller than the effective PSF, the influence of the bead size is negligeable. That is, the FWHM of the image PSF is the microscope’s resolution. In cases where the bead size is comparable to the effective PSF, however, the contribution of the bead diameter to the image PSF becomes more dominant. But we know the size of the bead! With that information, we can calculate the true resolution.²

² Harke, B. 2008. 3D STED microscopy with pulsed and continuous wave lasers. PhD thesis. University of Göttingen.

Another option is to hold constant the distance between standard probes. For that, your standard requires structure – something that holds apart two fluorescent probes at a fixed distance. Maintaining that kind of three-dimensional relation between points happens to be a useful feature of DNA origami. Like paper, DNA origami folds into nanostructures and can incorporate discrete binding sites for fluorescent molecules arranged in a programmed geometry. So, you can create rods holding fluorescent probes at a fixed, known distance from one another.

The fixed distance between each pair of fluorescent probes provides an additional reference point to benchmark the resolution of a microscope. Thus, you know a trait of your specimen that you can look for in the microscope image to assess fidelity. In that sense, this standard probe type is more informative about the performance of a system than PSF size alone. However (and there’s always a drawback), the handling of DNA origami is difficult. Additionally, dim fluorescence leads to a low signal-to-noise ratio and quenching between adjacent probes further distorts measurements.

Using fluorescent beads or DNA origami to measure resolution is not an option for all microscope systems. Their limitations in the three attributes we discussed are especially pivotal for superresolution systems where the behavior of fluorescent molecules is integral to achieving sub-diffraction resolution itself. Put another way, if superresolution emerges from the interplay of microscope and molecule, how do you differentiate microscope from molecule performance when measuring resolution?

This duality is less problematic for stimulated emissiond depletion (STED) than other superresolution technologies. STED achieves sub-diffraction resolution by restricting the area where a fluorescent molecule can emit photons to a diameter of about 20 nm. That is, the resolution is the restriction, as long as you use a bright probe that is smaller than the resolution.³ But consider single-molecule localization microscopy. There, superresolution comes from calculating the coordinates of fluorescent molecules. The precision of that localization depends on analyzing consistent, reliable PSFs of sparsely distributed fluorescent probes. Sound familiar? Consistent PSF. High photon emission. Distance. Three out of three attributes, which current standard probe types just can’t do. In fact, resolution of SMLM systems is often measured using Fourier ring correlation (FRC), which forgoes measuring FWHM of PSFs altogether.

As superresolution microscopy moves further into the sub-nanometer scales, the interplay microscope-molecule is increasingly central to performance. Breakthroughs will not come solely from creativity in the design of microscopes but also by optimizing the chemistry of dyes. That chemistry is the next frontier in superresolution technologies.

Human ingenuity commands admiration. We coax living entities – from elephants to viruses – into doing our bidding. We source out physical and mental work to machines. We sling people across the globe in thin-walled, stratospheric aircrafts. We receive pictures of the universe from a telescope 1.5 million kilometers from the Earth. Although it is the blink of an eye in the grand scheme of things, human history is a maelstrom of innovation.

All kinds of human motives drive that ingenuity. Need. Pleasure. Curiosity. Sharing. Greed. Fueling the adoption of its products, however, is one quality: ease-of-use. The press of a button that unleashes transactions, knowledge, and emotions is the name of the game.

The forces that drive technology adoption are no different in the science lab. Highly specialized instrumentation is limited to core facilities or niche research groups. Specific techniques may require an occasional unorthodox device in some labs, but a staple set of equipment and methods appears on nearly every benchtop, and its acquisition is driven by ease-of-use. Ease of installation. Ease of proficient use. Ease of results. By consequence, if you want to transform new technology into an everyday tool, you need to design it for one thing: ease-of-use.

That thought sparked the engineering feat that brought superresolution microscopy into the average science lab. With stimulated emission depletion (STED) microscopy and other techniques already enabling sub-diffraction resolution, it seemed unreasonable that researchers curtail their ability to visualize sub-cellular structures by using diffraction-limited microscopes. Clearly, adding STED to the standard constellation of lab equipment would require more than outstanding resolution. The problem, developers reasoned, was that Nobel-prize technologies are like Formula 1 race cars that only high-performance drivers can maneuver. So, their strategy was to distill that Nobel-winning “race car” into a shoebox that everyone could “drive.” STED would become a powerful upgrade to any widefield microscope with plug-and-play operation, intuitive handling, and rapid outcomes. The product was STEDYCON.

So, what does it look like?

At first glance, STEDYCON is a blue-and-metal casing about the size of a shoebox that sits on the camera port of a microscope. It is compact, unobtrusive, and sturdy. The plain exterior presages its ease-of-use but belies what lies within. Carefully planned engineering and design transform your microscope into a multicolor confocal and 2D STED system. It’s like plugging your scope into an amplifier and being rewarded immediately with power.

To begin with, the engineers tackled the big nuisances that make superresolution microscopy finicky. STEDYCON requires no alignment of the excitation and depletion lasers. They are aligned by design. With a novel optical arrangement that sends all laser beams through the same fiber, the system is more stable than other STED microscopes where beams travel separately. That inherent stability shortens the time to initiate imaging and simplifies installation, which takes only minutes. Furthermore, STEDYCON includes a scanner technology that allows it to operate with any widefield microscope, equalizing performance across platforms and precluding the need for dedicated equipment.

The engineers also revamped the autofocus. Unlike conventional optical instruments that use a dichroic mirror to couple the autofocus laser with the imaging beam path, STEDYCON runs the two beams parallel but segregated. Thus, autofocus in STEDYCON never interferes with imaging, which can produce fluorescence loss or distortions. With the simple press of a button, STEDYCON guarantees drift-free focus for extended imaging sessions up to several days. Also, as STEDYCON is equipped to control motorized stages, it automates the recording of multiple positions in a sample or a tile scan. The user is thus released from sitting at the microscope and can easily capture the whole sample or navigate across it to find objects of interest.

Secondly, STEDYCON makes superresolution microscopy a research technique for everyone. With minimal training, even novices to microscopy can produce high-quality STED images with a few intuitive manipulations of a browser-based control system with a user-friendly interface. Whichever the chosen procedure, users commonly visualize structures at a resolution of 30 nm. The same applies to lifetime imaging, which is fully integrated into the software with TIMEBOW.

Finally, STEDYCON makes no compromises on sensitivity or dynamic range of detection. The avalanche photo detectors (APDs) in STEDYCON have superior quantum efficiency. Thus, they reliably collect photons even under low-signal conditions, like when labeling densities are minimized to preserve physiological conditions in samples. This feature ensures clear images and exceptional signal-to-noise ratio, even under high-signal conditions.

The greatest testament to the performance and value of STEDYCON is its growing user base. Being frame-agnostic, STEDYCON complements microscopes of various makes and models. Exceptionally stable, it breaks with convention by operating flawlessly in dedicated labs, trade fair floors, living rooms, and even campsites. And STEDYCON users are a new generation of innovators leveraging a level of microscopy previously reserved for “Formula 1 drivers”. Swayed by easy access to unprecedented detail, they add diversity to the where, when, and how superresolution is used and uncover a myriad of insights hidden behind the low resolution of standard microscopes. There, at the frontiers of science, those discoveries unlock new lines of research as the boundaries to how we study structure, movement, and interactions at the smallest scale are lifted. STEDYCON’s ease-of-use in a shoebox is a portal to breakthroughs in an infinite world.

Resolution is one of those concepts that everyone feels like they understand, but then turns out to be annoyingly muddled. The term itself is used rather indiscriminately to mean different things. There’s image resolution, angular resolution, spectral resolution, and so on. Generally, all these terms reflect the ability to reveal detail in an object, but there are nuanced differences. So, let’s start by stating clearly what this article is about. For light microscopy, resolution is the smallest distance between two points of a specimen that a microscope makes distinguishable. Specifically in the context of fluorescence microscopy, resolution defines a spatial interval between two fluorophores. The closer the two fluorophores can be while remaining discernable in the resulting image, the smaller that distance and thus, the greater the detail resolved. The big R of resolution is actually a tiny d for distance.

So, just how small can d be?

First, we need to know what limits resolution. Several components of a microscope contribute to its resolving power. Some of them impact your ability to use resolution. For example, a detector (camera, sensor) with low pixel count will not capture the detail achieved by high resolving power. The opposite is also true. The highest pixel-count camera in the world will not add detail to an image produced by a microscope with low resolution. Similarly, magnification enlarges the image of a specimen, but creating detail lies squarely¹ with the microscope’s resolving power.

The truly limiting factors of resolution are the light used to examine a specimen and the ability of a microscope’s optical components to gather and focus that light. There is a simple, insurmountable reason for that dependence: light passing through an objective diffracts. As a result, the image of an emitting fluorophore is blurred, spreading beyond its actual size. As the blurred images of two fluorophores overlap, they become indistinguishable. The larger the blur, the further apart two fluorophores must be to tell them apart. Bingo. That blur, called the point spread function (PSF), restricts the resolution of a light microscope.

At the limit with Rayleigh, Sparrow, and Abbe

The PSF and its troubling consequences for resolution in optical systems have vexed scientists for centuries. Mathematician and astronomer George Biddell Airy first explained the smeared-out visage of a point of light in 1835. In fact, the PSF of an ideally focused point of light made with a perfect lens is named after him: the Airy pattern, which is a central and bright Airy disc surrounded by dim concentric rings (Figure 1).

By the turn of the 20^th Century, physicists were proclaiming just how close the PSF of two points could get before they become indistinguishable. Three definitions or “limits” stand out. To understand and compare them, it is worth mapping the relative light intensity across the middle of a PSF as projected onto an image plane (Figure 2). The center disc contains the bulk of light intensity (roughly 84 %) while the remainder is distributed in peaks and troughs of decaying amplitude corresponding to the concentric rings that extending out from the center.

The first limit is the Rayleigh criterion. John William Stutt, 3^rd Baron Rayleigh, stated that two light points of equal strength are resolved when separated at minimum by the width of the Airy disc (Figure 3A). This spatial interval allows a 20–30% dip in light intensity between the two PSF, which is discernable with the naked eye. A second version of the resolution limit came later from physicist Carroll Mason Sparrow who defined it as the distance at which the light intensity remains constant between the two PSF (Figure 3B). And perhaps the most famous limit is that described by Ernst Abbe who demonstrated that the resolution of any light microscope will never exceed half the wavelength of light (Figure 3C).

Figure 3A.

Rayleigh’s
resolution limit

Minimum resolvable separation of two points is the diameter of the central disc in a PSF.

Figure 3B.

Sparrow’s
resolution limit

Minimum resolvable separation is when the light intensity plateaus between the two points.

Figure 3C.

Abbe’s
resolution limit

Diffraction limits resolution of a light microscope to about half the wavelength of visible light.

Common to all three definitions for the limit of resolution are two parameters that Abbe identifies in his seminal paper² as the culprits of diffraction-limited resolution. Abbe was the first to introduce the concept of numerical aperture (NA) – a measure that combines the angle of the cone of light that can enter and leave an objective and the refractive index of the medium in which it operates to characterize its ability to accept and focus light. Abbe explained that the size of a PSF is dictated by the numerical aperture of the microscope objective and the wavelength of light used to image a specimen. Large numerical apertures and high-frequency light produce smaller PSF, which in return shortens the resolvable separation d between two points or fluorophores.

A practical approximation of Abbe’s resolution limit allows empirically measuring the resolution of a fluorescence microscope. If you image a fluorescent bead, you can measure the full width of its PSF central peak at half the maximum intensity (FWHM, Figure 4). Measuring resolution of a microscope is a topic for another article – Voilà.

² Abbe, E. 1873. Beiträge zur Theorie des Mikroskops und der mikroskopischen Wahrnehmung. Archiv für Mikroskopische Anatomie 9:413–468.

Abbe’s resolution or diffraction limit is fundamentally unassailable. Sure, you could look to use objectives of larger numerical aperture and higher frequency light. But unfortunately, the numerical aperture of modern fluorescence microscopes is at a practical maximum (about 1.4 for oil immersion objectives) and using ultraviolet or x-ray light damages specimens. That means that there is nothing we can do to our microscope system to reduce d any further. But there’s a second player in the generation of an image: the fluorophore. Manipulating the on/off states of fluorophores is another lever we can use to shrink d and thus, a workaround that modifies our approximation of the resolution limit by adding a third parameter.

\(d\approx 0.50\frac{\lambda }{\textrm{NA}}\cdot \textrm{workaround}\)

Stimulated emission depletion (STED) microscopy is one example. In STED, the excitation laser beam used to trigger fluorophore emission is superimposed with a second, donut-shaped de-excitation beam that suppresses that excitation. As a result, only fluorophores at the center of the donut-shaped beam are allowed to fluoresce. Increasing the light intensity of the de-excitation beam constricts the area in which fluorophores are allowed to fluoresce to a substantially smaller diameter than Abbe’s diffraction limit (roughly 200 nm). In this way, fluorophores can be much closer to one another and still be discriminated by the microscope.

The intensity of the de-excitation light is part of the “workaround” parameter here and can be used to approximate the diameter of the narrowed area of fluorescence based on the response of the fluorophores to the de-excitation light. Integrating that approximation into our measure of resolution:

\(d\approx 0.50\frac{\lambda }{\textrm{NA}}\cdot \frac{1}{\sqrt{1+\frac{I}{I_{sat}}}}\)

³ Harke, B. et al. 2008. Resolution scaling in STED microscopy. Opt. Express 16: 4228–4244.

STED and SMLM either shine lots of light on a specimen or capture lots of light from a specimen to manipulate the on/off states of fluorophores and thus, overcome the blur of the PSF. And although both can theoretically improve resolution infinitely, their “workarounds” are precisely what limits their resolving power. MINFLUX, a next-generation superresolution technology that achieves resolution in the single-digit nanometer scale, avoids that limit altogether by completely revamping fluorophore localization. But that too is a topic for another article.