Video rate stimulated emission depletion (STED)

Stefan Hell has published a new paper on video-rate tracking of single synaptic vesicles in axons using STED. STED is an imaging method that gives an improvement over the usual diffraction limit of λ/2NA by a factor of 3 - 4*. STED achieves this improvement by using a pulsed probe beam, followed by a pulsed "depletion" beam. By using diffraction, the depletion beam can be contoured like a donut around the imaging beam, and can induce stimulated emission in a region surrounding it. The remaining excited fluorophores now have a reduced spatial extent, and by collecting fluorescence that occurs after the depletion beam, the effective point spread function size is reduced.

This paper is the first example of using STED for video tracking, as opposed to imaging of static objects, and it's a nice application. But STED hasn't been very widely adopted in general, for two basic reasons, in my opinion:

1) It is incredibly complicated to implement. It requires two co-aligned lasers, which have to be spatially and temporally modulated in sync. I saw Stefan Hell speak about this work at UIUC (I believe he was there giving a job talk), and I remember thinking that there was no way on God's green earth that anybody else was going put in the effort to duplicate this. Now, this is coming from someone who works in an optical trapping lab, where "aligning obscenely complicated optics" is part of breakfast, and this hasn't stopped plenty of people from building optical traps. But, see below.

2) A factor of four improvement just isn't all that great. From a single-molecule perspective, this is similar to one of the problems that I saw with SHRImP: Most macro-molecular complexes are about 10 nm in size max, which is within the range of FRET and below the range attainable by these types of "diffraction-limit-improved" approaches. On the other hand, most cells are at least a micron, and eukaryotic cells, which have internal structures, are typically more like 10 - 50 microns. Most of these structures can be imaged perfectly well with conventional microscopy, and there's not a huge number of interesting things too big to measure with FRET and too small to measure with conventional confocal scanning. Now, I obviously have a physicist's bias here, and I'm sure there are plenty of people who know more cell biology than I who would disagree**. But, the fact is that I don't really see people rushing to implement STED in their laboratories.

Now, optical trapping requires probably as much or more complication than STED, at least the sorts of super fancy optical traps we have here in the Block lab. But the difference, in my mind, is that optical trapping gives you an entirely new tool for measuring and manipulating. It's completely orthogonal to fluorescence approaches, and therefore it has an immense amount of power to go where other techniques cannot even hope to tread. STED has all the complication, but far fewer benefits as a result. Furthermore, optical trapping naturally appeals to physicists, with its emphasis on forces and mechanical transitions. STED is a tool that would primarily appeal to cell biologists, but it requires a physicist to build it. Hence, there is probably a dearth of people who both a) have questions that STED can answer, and b) have the expertise to actually build such an instrument.

On the other hand, it looks like you can now buy a commercial off-the-shelf STED microscope from Leica, so who knows? Maybe this will be the next big thing.

*They claim the diffraction limit is about 260 nm. With a typical oil immersion objective, NA = 1.45, and Cy3/TAMRA is probably one of the most popular dyes, with an emission peak of around 570 nm, such that the diffraction limit = 570/(2*1.45) = 197 nm. However, in this paper, they used Atto 647 (emission peak = 673 nm) and a 1.4 NA objective, which would result in a diffraction limit of 673/(2*1.4) = 240 nm. I believe they're calculating the "diffraction limit" in this case as the FWHM of the observed point spread function (PSF). But, there are two sort of tricky things here to note. The first is that their claim of a 260 nm diffraction limit, instead of the more typical 200 nm, makes their data look better by contrast. The second is that they claim an 18-fold improvement over the diffraction limit, but this is in terms of focal area, not point spread function width. This is also a tricky way to make their numbers seem a bit better than they are, because the interesting parameter is the linear length scale, which tells you how far apart two objects can be and still be resolved. They only reduce the point spread function width itself by a factor of about 4.

**Although I did just organize a Fat Alberts club here in the Block lab so us ignorant physicists can start learning more about cell biology. I just presented chapter three, "Proteins", last week, and we're on to DNA this week. Woo!