Super-resolution microscopy pushes fluorescence imaging well below the classical diffraction limit of about 200 nm. This primer walks through the four families available at the ICMS Advanced Microscopy Facility, the physical trick each one uses, and how to choose between them for a real experiment.
The diffraction limit
An optical microscope cannot focus light to a point. A point emitter is imaged as a blurry spot whose width is set by the wavelength of light and the numerical aperture (NA) of the objective. Ernst Abbe wrote this down in 1873: the smallest resolvable distance is roughly λ / (2·NA). For visible fluorescence (λ ≈ 500 nm) and the best oil-immersion objectives (NA ≈ 1.4), that floor sits around 200 nm laterally and 500 nm axially. That is fine for whole cells and large organelles. It is hopeless for nuclear pores, synaptic vesicles, receptor clusters, and viral particles — the structures that most modern cell biology actually cares about, which all live below 100 nm. Super-resolution methods break this wall by exploiting the photophysics of the fluorophore, not the optics of the lens.
STED
Stimulated Emission Depletion (STED) microscopy was proposed by Stefan Hell and Jan Wichmann in 1994 (DOI:10.1364/OL.19.000780). The trick is to overlay the focused excitation spot with a donut-shaped depletion laser. This second beam drives any fluorophore in its bright ring back to the ground state by stimulated emission before it can fluoresce spontaneously. Only the molecules sitting at the donut’s central zero survive to emit. Scan that effective point across the sample and you get a sharper image — the same way you would by using a smaller pencil to draw the same picture.
Resolution scales as Δr ≈ λ / (2·NA·√(1 + ISTED/Isat)). In practice, 30–80 nm lateral resolution is routine, with sub-20 nm achievable on photostable samples. Because STED is a real-time scanning method, the image appears as it acquires — no frame stack, no reconstruction, no buffer chemistry. That makes it a strong choice for thick fixed samples and for live-cell work when the dye and power budget allow.
At AMF, STED is delivered by the Abberior STED ExpertLine.
SMLM (STORM and PALM)
Single-Molecule Localization Microscopy (SMLM) builds a super-resolution image one molecule at a time. The key insight is statistical: two fluorophores closer than the diffraction limit cannot be told apart when they emit simultaneously, but if you persuade them to take turns — only one within any diffraction-limited region emitting per camera frame — each can be localised by Gaussian fitting to a precision of a few nanometres, much smaller than the spot itself. Repeat for thousands of frames and you reconstruct an image as a point cloud of single-molecule positions.
The two foundational papers came out within months of each other in 2006: PALM by Eric Betzig and colleagues (DOI:10.1126/science.1127344), using photoactivatable fluorescent proteins; and STORM by Mike Rust, Mark Bates, and Xiaowei Zhuang (DOI:10.1038/nmeth929), using a photoswitchable cyanine-dye pair. The practical workhorse since then has been direct STORM (dSTORM), introduced by Markus Heilemann and colleagues in 2008 (DOI:10.1002/anie.200802376): standard organic dyes are coaxed into long, well-defined dark states by a reducing thiol buffer, simplifying labelling enormously. Lateral resolution lands at 10–30 nm; localisation precision can be 5–10 nm with bright dyes and a long acquisition.
SMLM is the most flexible super-resolution family in the toolbox — multicolour, 3D via PSF engineering, and live-cell modes are all in routine use — but it does not feel quick. A single dataset is typically tens of thousands of frames over several minutes, plus reconstruction. At AMF, SMLM is delivered by the ONI NanoImager (a compact two-colour benchtop SMLM) and the Abbelight SAFe-360 on a Nikon Ti2 body (three lasers, three channels, TIRF/HiLO/epi modes, live-cell incubation).
DNA-PAINT
DNA-PAINT replaces the photochemistry of STORM with hybridisation kinetics. The target is decorated with short single-stranded DNA “docking” sequences; a dye-labelled complementary “imager” strand floats freely in solution. When the imager finds its docking strand it sticks for a few hundred milliseconds, fluoresces brightly, and falls off again. The blinking comes from binding and unbinding, not from coaxing a dye in and out of dark states.
Three things follow. Any bright, photostable dye works, because no photoswitching is required. Many targets can be imaged in the same sample by sequentially exchanging imagers (Exchange-PAINT) — Ralf Jungmann and colleagues demonstrated this in 2014 (DOI:10.1038/nmeth.2835) and routinely reach ≈10-plex. The binding kinetics are calibrated, so molecules can be counted absolutely (qPAINT). The cost is acquisition time: each round takes tens of minutes, and high-plex experiments run for hours. At AMF, DNA-PAINT runs on the Abbelight SAFe-360, which is engineered for it.
MINFLUX
MINFLUX (MINimal photon FLUXes), introduced by Francisco Balzarotti and colleagues in 2017 (DOI:10.1126/science.aak9913), is the newest of the four. It localises a fluorophore not from where its photons land on the camera, but from where the minimum of a donut-shaped excitation beam must sit to switch it off. Because position is encoded in beam placement rather than photon centroids, MINFLUX needs roughly 22× fewer photons than camera-based SMLM — pushing single-nanometre precision into routine reach, with molecules ≈6 nm apart resolvable. It is currently the closest optical fluorescence imaging gets to the molecular scale.
AMF does not yet have MINFLUX. It is on the instrument roadmap; if your project genuinely needs sub-5-nm precision, talk to the imaging scientist early so we can plan the right path.
Trade-offs and when to use which
No single technique wins everywhere. The honest summary:
| Technique | Resolution | Speed | Live cells | Labelling | At AMF |
|---|---|---|---|---|---|
| Confocal (baseline) | ~250 nm | Fast | Yes | Standard dyes | Nikon A1, Nikon AX-NSPARC |
| STED | 30–80 nm | Fast (real-time) | Yes (with care) | Specialised STED dyes | Abberior STED ExpertLine |
| SMLM (dSTORM/PALM) | 10–30 nm | Slow (frame stack) | Limited | Photoswitchable dyes + buffer | ONI NanoImager, Abbelight SAFe-360 |
| DNA-PAINT | ~5–10 nm | Slow (kinetic) | Fixed only (typically) | DNA-conjugated antibodies | Abbelight SAFe-360 |
| MINFLUX | ~1–3 nm | Moderate | Yes | Switchable dyes | Not yet — ask |
We also do not currently offer live-cell Structured Illumination Microscopy (SIM) on site. If your question is about dynamics at ~100 nm rather than structure at 20 nm, ask before booking — SIM elsewhere on campus or confocal-plus-deconvolution may serve you better.
A worked example
You want to image a single membrane receptor in a fixed mammalian cell and ask whether it forms nanoscale clusters.
Fixation: paraformaldehyde preserves protein epitopes; avoid methanol if the antibody is conformation-sensitive. Labelling: primary plus dye-conjugated secondary antibody is the default, but the antibody sandwich (~20 nm) limits how close two true molecules can appear — a directly-labelled nanobody (~3 nm) gives a more honest cluster map if available. Fluorophore: AlexaFluor 647 remains the dSTORM workhorse — bright, photostable, well-behaved in standard thiol buffer. Instrument: the ONI NanoImager for fast turnaround and a single colour; the Abbelight SAFe-360 if you need three channels, TIRF-quality background suppression, or a path to DNA-PAINT later. Expected output: a localisation point cloud at 10–20 nm precision, ready for cluster analysis (DBSCAN, Voronoi, or Ripley’s K).
For more on choosing an instrument, see Choose a microscope. For acquisition steps, see the imaging SOPs.