STED microscopy sees details on the nanoscale

The first viable concept based on this insight was stimulated emission depletion (STED) microscopy (Optics Letters 19 780). In its single-point scanning variant, a STED microscope probes the object with a diffraction-limited laser spot exciting the molecules to a fluorescent state. By scanning this spot across the specimen, an image can be assembled computationally by collecting the fluorescence point by point.

Without any further additions, the resolution of the system would be diffraction-limited. However, in the STED microscope, a second red-shifted beam of light, known as the STED beam, is focused in such a way as to form a doughnut featuring a central zero intensity point. Molecules exposed to this doughnut spot are quenched by stimulated emission and, as a result, cannot fluoresce. Consequently, fluorescence occurs preferentially at the doughnut centre.

Fig. 1

Importantly, as the intensity of the STED beam increases, the fluorescence decays nearly exponentially towards zero (see figure 1). As a result, the diameter of the fluorescent spot, d, scales with the intensity of the STED beam according to the equation shown in figure 1.

Since the image is obtained by scanning, the spatial resolution is now given by the reduced fluorescent spot diameter. This means that the resolution improvement by STED is all-physical and no prior information about the sample or the dye concentration is required.

Fig. 2

Figure 2 shows how d scales for three values of I divided by I_SM. Without STED (I equals zero) the formula is just that of Abbe. However, as I goes to infinity d converges towards zero (Nature Biotechnology 21 1347). Thus, STED microscopy leads to the insight that the limiting role of the wavelength can be radically overcome. While d still scales with the wavelength (accounting for the fact that STED relies on propagating waves) the resolution is no longer limited by the wavelength in use.

A practical STED microscope shares many elements with a confocal one (figure 2): a high aperture lens, a dichroic mirror for separating the fluorescence, a scanning device, and a detector. In addition, a second dichroic mirror couples the STED beam into the light path.

To achieve the doughnut-shaped focus, the STED beam first passes through a wavefront shaper modulating the beam's wavefront such that a doughnut is formed upon focussing. We note, though, that STED is not a conceptual extension of confocal microscopy, because it does not require a confocal pinhole. Its resolution is just determined by the doughnut and the ratio I divided by I_S. That said, a pinhole is advantageous to reject stray light. Thus, a confocalized STED microscope preserves most of the advantages of a confocal microscope while adding subdiffraction resolution. Depending on the phase filter used, the increase in resolution may be lateral (in the focal plane) and/or along the optic axis.

So far, excitation and STED have mostly been effected with synchronized pulse pairs of approximately 100 and 300 ps duration, respectively. While pulsed illumination is particularly effective because it is adapted to the lifetime of the fluorophore, continuous wave (CW) illumination has also been shown viable recently (Nature Methods 4 915). The use of CW beams greatly simplifies the implementation of a STED microscope.

Fig. 3

STED microscopy in action
Figure 3 (upper row) shows an agglomeration of 20 nm-sized fluorescent beads on a coverslip imaged in the confocal and the CW STED mode. Whereas the confocal image exhibits a resolution of 250 nm, that of the CW STED counterpart is 29 nm. The middle and bottom rows show image pairs of the proteins syntaxin and synaptophysin recorded from the same place in a mammalian cell (Proceedings of the National Academy of Sciences USA 103 11440). Linear deconvolution was additionally applied to the CW STED images to further enhance the details; however, deconvolution is optional.

The comparison demonstrates that the resolution of the confocal microscope is insufficient to map the exact protein distributions, while the STED images can reveal the distinct formation of these proteins. The image pair in the lower row was acquired using pulsed lasers, whereas that in the middle row is in the CW mode.

While the resolution is no longer fundamentally restricted by the wavelength of light, the current technical limitations are set by the laser power available for STED, the bleaching resistance of the dyes, and the (im)perfection of the doughnut 'zero'. As these parameters are continually optimized, even sharper images can be expected in the future. STED microscopy has been implemented recently as a fast beam-scanning commercial microscope system (www.llt.de) that should greatly promote its wider adoption.

Conclusion and outlook
Today, it is clear that the limits set by diffraction in lens-based fluorescence microscopy have been overcome. Key to this accomplishment was the incorporation of the spectroscopic properties of fluorophores, specifically of their states, into image formation (Optics Communications 106 19).

This insight is also critical to more recent approaches to super-resolution imaging like RESOLFT (reversible saturable/switchable optical fluorescent transitions, Current Opinion in Neurobiology 14 599), PALM (photoactivated localization microscopy, Science 313 1642) and STORM (stochastic optical reconstruction microscopy, Nature Methods 3 793) that switch markers between a fluorescent and a non-fluorescent (conformational) state and back.

Featuring a practical resolution down to 20 nm, STED microscopy has been used to solve biological problems. For example, STED has been applied in the neurosciences to settle a long-lasting question about the fate of proteins contained in the membrane of synaptic vesicles (Nature 440 935).

The nanoscale resolution of STED microscopy has also unraveled the spatial arrangement of the protein bruchpilot in the synapses of the fruit fly neuromuscular junction (Science 312 1051). Recently, STED has been used to determine the size of syntaxin 1 protein clusters required to proper neuronal function (Science 317 1072). Applications are expected to increase as the technique becomes commercially available.

It is interesting to note that, while fluorescence is a very powerful readout mode, the spatial confinement of a signal-giving state could, in principle, also be extended to signals other than fluorescence (Science 316 1153). Therefore, it would not come as a total surprise if the far-field diffraction barrier was broken for other microscopy contrast modalities.

• This article originally appeared in the December 2007 issue of Optics & Laser Europe magazine.