A Short Introduction to Super-Resolution Microscopy

by J. Demmerle, C. Eggeling & L. Schermelleh


The past decade has seen a number of exciting new fluorescence imaging techniques emerge, many of which fall under the umbrella of "Super-Resolution Microscopy" or "Optical Nanoscopy". At Micron Oxford† and the Wolfson Imaging Centre Oxford†† we work to advance these techniques and to bridge the gap between complex technical developments and their biological applications in order to make cutting-edge imaging accessible to researchers in Oxford and beyond.

If you are a novice in the field you may ask:

In this introduction we will give a brief overview of the state of the field, and address some pros and cons of each method.

The Limits of Focused Light

Optical microscopy is one of the most versatile tools in a biologist's arsenal, since it allows the observation of the interior of 3D preserved fixed or living cell with minimal perturbation. However, from its invention in the 17th century until a few decades ago its spatial resolution was constrained by a seemingly impenetrable barrier: the diffraction limit. Simply put, the minimal distance at which two adjacent objects can still be discerned equals the wavelength used to observe them divided by twice the numerical aperture (NA) of the microscope (the NA roughly expresses the focusing strength of the microscope's objective lens). With visible light, the smallest practical wavelength is around 450 nm (blue emission), and the largest NA possible is around 1.4.

Even with advances such as confocal microscopy and deconvolution algorithms, only biological structures separated by more than about 200 nm in the lateral dimension (x and y axis) could be discerned. Resolution along the axial dimension (z axis) is even lower, being limited to roughly the wavelength of the emitted light, or 500 nm at best. Further, these values are only theoretical limits under optimal optical conditions. The effective resolution for typical biological samples is significantly decreased due to distortions from the sample itself such as light scattering, out-of-focus blur, spherical aberration, and poor signal-to-noise ratios.

Breaking the Diffraction Barrier – A Menagerie of Options

Imaging with sub-diffraction spatial resolution was postulated in the early 20th century, and while significant theoretical advances were made starting in the late 70's, it was not until the mid-90's when the first approaches for truly circumventing the diffraction barrier were implemented, and only in the last decade have these super-resolution technologies become available to the average researcher. Based on their different underlying principles these methods may be grouped into:

  1. structured illumination microscopy (SIM);
  2. targeted photoswitching microscopy such as stimulated emission depletion (STED) or reversible switchable optical fluorescence transition (RESOLFT) microscopy;
  3. localization microscopy such as (direct) stochastic optical reconstruction microscopy ((d)STORM) or (fluorescence) photoactivated localization microscopy ((F)PALM).

Using Moiré Interference: Structured Illumination Microscopy (SIM)

Super-resolution 3D-structured illumination microscopy (3D-SIM) works by the interaction of a high frequency three-dimensionally modulated illumination pattern with high frequency variations in the sample fluorescence caused by very small structures. This entails a lower frequency Moiré interference pattern that contains otherwise non-resolvable structural information encoded into the observed image. By imaging a series of these Moiré patterns in different positions and subsequent computational post-processing, the sub-diffraction sample information can be algorithmically decoded and reconstructed. By including information from above and below the focal plane both lateral and axial resolution is improved.

The principle of structured illumination microscopy is based on the Moiré effect generated when interfering a fine-striped pattern of excitation with sub-diffraction features in the sample emission. The SI pattern needs to be shifted and rotated to reconstruct super-resolution information from all spatial directions (adapted from Schermelleh et al., 2010).
Resolution
Two-fold improvement in x-y plane and along the z-axis (equivalent to an eight-fold volumetric improvement). With a 1.4 NA objective and depending on the wavelength 100-130 nm (xy) and 280-350 nm (z) can be achieved.
Pros
  • SIM can be used with conventional fluorophores (e.g., DAPI, Alexa, ATTO, GFP, YFP, etc.)
  • Very high sensitivity using EMCCD camera detection.
  • Massive contrast improvement (~3000-fold compared to wide-field).
  • 3D sectioning over 10 µm into a sample (ideal for cultured cells).
  • Fast pattern generation (OMX Blaze) and rapid image collection (e.g., by using recently developed sCMOS cameras) enable very high imaging speed and are suitability for live-cell imaging.
  • Fast recording of a large field-of-view (e.g. a whole cell).
Cons
  • Only moderate lateral resolution improvement compared to other methods.
  • Errors in system calibration, refractive index mismatch and/or poor sample quality can induce artifacts and severely compromise resolution.
Applications
  • Multicolor imaging of 3D structures with large field-of-view, e.g. cytoskeletal networks; centrosomes; nuclear organization within whole cells.
  • Live-cell applications.
  • Proper labeling contrast and meticulous optical aberration correction are of utmost importance.

The Targeted (Photoswitching) Readout: STED Microscopy

Stimulated emission depletion (STED) microscopy was the first optical super-resolution microscopy technique that increased the spatial resolution of fluorescence microscopy by a large factor; in principle it can reach resolution at the molecular scale. STED uses stimulated emission to inhibit fluorescence emission at predefined sample coordinates such that adjacent features emit sequentially in time. Typically, a laser beam inducing stimulated emission and featuring at least one zero-intensity point is overlaid with a regularly focused excitation beam confining the effective fluorescence signal to sub-diffraction dimensions. A common design is a doughnut-shaped focal intensity pattern of the STED beam. Scanning the co-aligned excitation and STED beams through the sample yields the final sub-diffraction resolution image, whereby the resolution can be adjusted by the intensity of the STED beam. Similarly, other fluorescence inhibition processes may be used to overcome the diffraction barrier, as long as two distinct transient states of fluorescence emission (e.g., a dark and a bright one) can be prepared. This can for example be achieved by ground-state depletion (GSD) or the use of photoswitchable fluorescence markers in the generalized concept called RESOLFT.

STED microscopy is based on stimulated emission depletion of a targeted region surrounding the excitation spot of apoint scanning laser beam. This leaves only a small sub-diffraction sized read-out area of emitted fluorescence to bedetected by a photomultiplier or photodiode (adapted from Schermelleh et al., 2010).
Resolution
If the resolution of STED is in principle unlimited, in reality the resolution is determined by the sophistication of the laser and the photophysical properties of the dyes. Current commercial setups achieve a lateral resolution of 50-80 nm with an axial resolution in the 500-800 nm range, equal to the conventional confocal axial resolution. The best lateral spatial resolutions achieved experimentally so far are 5 nm in solid states, 20 nm in fixed samples, 30-50 nm in living cells, and around 60 nm in 3D applications.
Pros
  • "What you see is what you get" - no algorithms required
  • High lateral resolution.
  • Can image over 20 µm deep into the sample.
  • Can be combined with "conventional" confocal imaging modalities.
  • The resolution can be tuned.
  • Can be combined with single-molecule techniques such as fluorescence-correlation-spectroscopy (STED-FCS) to measure fast molecular dynamics.
  • STED/RESOLFT currently provides the fastest sub-diffraction resolution recordings of small fields of view.
Cons
  • Fluorophore options are limited: two colors only, and UV is not an option.
  • STED requires high laser intensities energies, thus photobleaching/phototoxicity can be a problem. Advanced techniques such as RESOLFT can avoid this problem.
  • Speed scales with the scan size and can become relative slow for large field-of-views. In theory, techniques such as SSIM or parallelized RESOLFT can avoid this, but are not yet commercially viable.
  • Currently available commercial instruments are not yet optimal for 3D and live-cell imaging of larger objects.
Applications
  • Current commercial instruments are ideal for distinct structural features with little z-extension, e.g. vesicles; membranes; synaptic behavior.
  • Bright labeling and the right choice of dyes are critical.

The Stochastic Read-out: Single Molecule Localization with (d)STORM/(F)PALM

The principle of localization microscopy is based on detecting the positions of individual fluorescent molecules. These are switched between two distinct states of fluorescence (e.g. a "dark" and a "bright" state). Using a conventional wide-field or TIRF light path and EMCCD camera detection, only a small subset of molecules are allowed to switch into the "bright" state at the same time in a stochastic manner. Various strategies are used to collect fluorescence emissions from the fraction of the total fluorophores, in any one image. The temporal separation allows single isolated diffraction limited spots (originating from individual fluorophores) to be detected in the camera image. Their positions can then be determined with sub-diffraction precision in a post-processing step using single-molecule localization methods such as Gaussian fitting. Finally, super-resolution images are reconstructed from the projected positional information of thousands to millions of individual molecules detected in a series of hundreds to thousands of raw images. PALM, FPALM and STORM were the first techniques to generate super-resolution images in this manner, using special photo-activatable or photo-switchable dyes. More recently it has been shown that conventional fluorophores like Alexa or ATTO dyes and conventional fluorescent proteins like GFP can be also utilized for this method (dSTORM, SPDM, GSDIM). For more details see Rainer Kaufmann's section on localization microscopy.

The principle of localization microscopy is based on timely separated stochastic photoswitching of individual fluorophores and the subsequent determination of their spatial position with high accuracy. Projecting all detected spots generates a (pointillist) super-resolution image of the sample (adapted from Schermelleh et al, 2010).
Resolution
The spatial resolution of localization microscopy is theoretically unlimited but depends on the number of fluorescence photons detected per single molecule and the density of detected molecule positions. Localization accuracy is typically in the 10s of nanometers, while structural resolution can reach down to 30 nm laterally. The axial resolution is either unchanged (in conventional far-field) or 120-200 nm (in TIRF mode). Improvements to 50-100 nm axial resolution are possible using highly specialized setups (e.g. biplane detection, astigmatism or double-helix PSF imaging, or dual-objective setups).
Pros
  • Very high single molecule localization precision (typically between 10-20 nm).
  • High lateral structural resolution (routinely ~50 nm in x-y).
  • Quantification of single molecules (e.g., cluster analysis, single particle tracking).
  • Excellent delineation of small, dim, punctate, or filamentous objects such as discrete macromolecular complexes.
  • Relatively simple instrumentation.
  • Works with most common fluorophores, and often with standard sample preparation.
Cons
  • Long imaging times (typically several minutes per single-plane image).
  • Typically can image one plane only (no 3D sectioning).
  • May require special fluorophores and imaging media to realize photoswitching under optimal conditions.
  • Works best with TIRF setup and with fixed samples where you are confined to imaging regions within a few hundred nanometers off the coverslip.
  • Missing molecules or counting them multiple times without correcting for it in the post processing may lead to misinterpretations of the data.
Applications
  • Imaging cell surface or distinct features with little z-extension. Particularly useful for: membranes; receptors; adhesion complexes; protein clustering and particle tracking with high particle densities (sptPALM).
  • Getting quantitative/statistical information about protein distributions in sub-diffraction-limit structures that are inaccessible with other techniques.
  • Successful imaging depends on the reconstruction algorithm and requires the right choice of dyes, embedding condition and high labeling density.

It's Both Better and Worse Than It Seems

Two things should always be kept in mind when considering the super-resolution options at one's disposal. Firstly, the technology is constantly advancing, and some of the statements in this introduction may have become obsolete by the time you read it – for example, localization microscopy techniques are experiencing rapid improvements at multi-channel acquisition and z resolution, and STED setups are becoming easier to implement with advances in laser technology. Secondly, many of the resolutions and a few of the techniques cited here are ceilings, achieved in highly specialized labs with custom-built systems that may not be practical for the average researcher. Most of the techniques discussed here are available commercially, but performance can vary, and getting the best from any of these techniques requires knowledge and experience.

Trade-offs – There's No Free Lunch

You may be tempted to go and lobby for your very own super-resolution system immediately, but remember that there are many things to take into account. Resolution is only part of the equation - you may value multi-color imaging, in vivo observations, high-speed imaging, three-dimensional imaging, or low labeling intensities. Each priority comes with a compromise in another area, and the optimal technique will be entirely dependent on the specific demands of your experiment and how you want to spend your given photon budget. Of course, there are still many remaining challenges in the application of super-resolution imaging to biological specimens, as well as a lack of expertise and tools to evaluate the quality of the resulting data. We are beginning to address these issues with tools such our SIMcheck ImageJ Plugin suite to evaluate 3D-SIM data.

Trade-offs in (super-resolution) optical microscopy. Benefitting in one aspect inevitably compromises others.

Getting Started!

Once you ensure these issues are resolved, you can proceed with acquiring fantastic images. Your best bet is to visit your institution's imaging core and talk to the facility manager. If you are based at Oxford, then feel free to visit the Micron office in the basement of New Biochemistry or the Wolfson Imaging Centre at the WIMM. If you are based at another institution, feel free to send us an email. The Micron team is well equipped to help you decide which technology is most suited to address your question, what reagents you may need, how to prepare the samples, and how to create suitable image acquisition and post-processing workflows. It may seem like a lot of work, but once you're generating data and seeing structures at the sub-200 nm resolution, we think you'll be hooked.

Further reading

Funding

† funded by a strategic award from the Wellcome Trust.

†† funded by the MRC

Micron and WIMM Imaging Unit are also jointly funded by a next generation imaging grant from the MRC, BBSRC and EPSRC (NanO).