[intergalplan]

The work seemed to pointedly avoid micromanaging, by instead finding out how to open up the correct pathways between cells in a region and instruct them to become something else. No need to say "OK, these cells need to become a retina, these need to become a cornea" or to figure out how to change the body plan of the creature at the DNA level—just place a "become an eye" marker in the right spot, and boom, the cells self-organize to become an eye, and may even manage to connect themselves to the CNS such that it becomes a functioning eye!

The key insight seems to be that you can hack the (as the research put it, by analogy) software of a set of cells and let the hardware (DNA, cell mechanisms) worry about exactly how to do what you told them to do, so rather than, say, figuring out how to read DNA and tell what kind of organism will come out of it such that we can start adding or removing bits here and there (very, very hard), you can instead screw around with inter-cell signaling to hijack functionality that the cells already have, but just aren't using, to get them to do new things.

I thought there were several "holy shit" moments in the video. The cancer thing—yikes. Reconnect genetically-broken, actively dangerous cells to the inter-cell communication network and they stop acting crazy, while still having messed-up DNA? That's a surprising result, and a powerful demonstration of what the technique can accomplish (may not be therapeutically useful for a bunch of reasons I can think up, but it still seems like a notable, surprising result)

Another was the sticky "memory" of a modified organism, that is, the worm that'd had its tail cut off then been instructed to make a second head there instead, and then had that cut off and... grew a head again, without further prompting, demonstrating that body layout and gene expression is some combination of DNA and a kind of cell-memory or cell-state across the whole organism, not something driven entirely deterministically by DNA alone.

The part about tricking worms into growing heads for 100-150 million-year separated evolutionary relatives, complete with the correct-for-the-other-species brain, was nuts. The "instructions" are still there, just dormant, and if you confuse the cell-memory (if you will) of what it's supposed to be it'll sometimes decide to be the wrong species. WTF.

[ravi-delia]

On that last note, a similar thing happens with trees. Trees aren't really a distinct class of organism. The last common ancestor of some trees wasn't a tree, and the last common ancestor of some non-trees was a tree. As it turns out, with a few genetic tweaks lots of plants will slow their metabolic processes and construct a sturdy wood stem, and visa-versa.

Even cooler, perhaps, are eyeless cave fish redeveloping eyes quite quickly after a few generations in the light. It seems like a lot of things are like that. When we look at the fossil record it looks like things evolve in leaps and bounds rather than more gradually over time. Maybe that's because organisms are more built with prefabs rather than from scratch.