Realistically, how deep into tissue can you go while still achieving high-quality single molecule localization data? I’m sure this will vary based on the tissue, but just wondering what a reasonable ballpark would be. Thanks!
Based on my experience, I’d expect things to deteriorate pretty quickly—maybe you could go as far as ~5–6 µm? A lot would depend on the nature of your sample and your optical setup. E.g. if you were using a lattice light sheet this would be quite different than imaging using HILO or epi-illumination.
There are a couple of other challenges for SMLM + tissue / thick samples to that it can be helpful to keep in mind:
Spherical aberration, which will become more severe with depth and can be partially accounted for by objective/immersion choice and/or the use of adaptive optics.
Drift correction can be especially tricky as it can be difficult to place/visualize fiducial markers when working with tissue samples. If your target is abundant across the field of view you can probably get by with redundant cross-correlation but for more sparse targets this approach does not work very well.
Thank you, Brian, for such a thoughtful answer! I’ve definitely seen that I’m depth-limited by spherical aberration with SIM imaging, so that would likely be a limiting factor with other techniques as well.
I agree with Brian regarding the choice of microscope, and how it interprets the axial position of each localization. However, with regard to drift-correction, I am a bit more optimistic. I have imaged samples which are over 10 um thick, and had fiducials on top of the tissue, as well as on the coverslip, and that seemed to be sufficient. Of course, having fiducials at smaller axial intervals will improve your structural resolution, but for most microscopes and users, the reduction in resolution will come from the optics.
Hope it helps,
I agree with @bbeliveau about the useful limit being within a few microns inside the tissue, as well as spherical aberration being the limiting factor by degrading the PSF. This of course will depend on the tissue type and sample preparation.
Classical STORM has been performed on brain slices in the first microns (classic example is Dani et al. Neuron 2010, see also some tips and tricks in Barna et al. Nature Protcols 2016). Various tricks have been published on the imaging and algorithm side (see Bon et al. Nat Meth 2018, Mlodzianoski et al. Nat Meth 2018, Li… Ries Biomed Opt Express 2019 for some examples), which could help in getting a bit deeper.