Peter Passaro writes:
> They do nasty things like tearing cell membranes. If
> you freeze an object very quickly the crystals are
> probably kept in a less organized state and of a
> diameter which is not particularly dangerous to
The types of damage to biological caused by temperatures below their optimum are manifold, and still largely unknown. Proteins denaturate, as a result some superstructures are damaged or fragment, lipid bilayers undergo phase transitions, as the result integral membrane protein structure and/or their distribution get disturbed. Further down the slow thermal descent (only tiny specimens can be cooled rapidly) extracellular ice growth begins, increasing the cytosol concentration (which in itself is denaturating to many proteins). Randomly oriented ice crystals anisotropically deform tissue by pushing it along their main growth axis. If ice is removed by freeze-substitution (washed out with, say, isopropanol at dry ice temperature), the resulting tissue strongly resembles an irregular mesh, with about 100 um holes iirc. Though difficult to assess without watching the dynamics process/using cell markers, the whole tissue looks scrambled (randomly, nonlinearly distorted). If frozen tissue is warmed, ice crystals melt suddenly, resulting shrunken cells full of highly concentrated cytosol to be surrounded by solute of low ionic strength -- causing cells to rupture due to osmotic shock, resulting membrane debris closing up into vesicles. You think you see lots of intact structures, but a lot of them are just artefacts. (Notice that most of these points, grave as they are, are largely irrelevant to the gros of the patients, where factors like system-wide damages due to the original disease/age, hour-long normothermic ischaemic damage (go neuro, people, the periphery is junked, anyway), clotting and spotty perfusion, mechanical damague due to brain swelling etc. dominate overwhelmingly).
Of course we do not need to warm the patient for reconstruction, in fact the only way to do a mapping at molecular scale is to do it in solid state, using the structure as blueprints only. Since the scan is destructive, you essentially rebuild the structure anew minus the suspension damage (using damn smart filters), whether physically, or as model in machina (at least slicewise,as you will need for applying smart filtering, anyway). However, you are still facing a severe, very probably irreversible information loss due to the 'turbulent' deformations (irreversible mapping: a given distorted structure could have been originated by a large number of structures, which are all more or less equally probable).
To avoid such information lossage we can use vitrification, fixation (with a complex perfusate/protocol which would have to be developed) and subsequent storage at low temperatures or uploading in vivo ('gradual uploading'), which has the major advantage of supplying lots of low-level operation signatures for what is an essentially indefinite duration (which, at the top level, you can even tweak or correlate with sensomotoric information). If the duration (hours to decades) and resolution (say, a sample/um^3) of the sampling is good enough, you essentially don't need to make a molecular map anymore.
Unfortunately, you need devices made with the mechanosynthetic paradigm, which is not yet validated.