Sparks Brain Preservation

I've just learned from the highly reliable source of ChatGPT that dendritic spines begin retracting within minutes of ischemia and are entirely lost within about an hour. This leaves smooth dendrites, dispersed receptors, debris, and orphaned boutons. This feels like fairly thorough information theoretic loss to me. If this is true, then all cryonics and fixation cases performed so far are not recoverable. The only way to survive this seems to be fixation within minutes after death, and even that seems to be nearly impossible due to typically poor perfusion. I look forward to learning more about this. It seems pretty important and I can only hope I'm missing something.

We reviewed this literature in our paper from 2023:

https://pmc.ncbi.nlm.nih.gov/articles/PMC10294569/#S7.4

One relevant quote:

“It is challenging to reconcile differences in the decomposition rate of dendrites observed across these studies. However, problems with dendrite visualization in studies finding that dendrites degrade relatively faster could help to explain this divergence, whereas it is more difficult to imagine how the literature suggesting dendrites are stable for longer periods of time could be systematically flawed.

In summary, dendritic swelling can occur early in the PMI (Rees, 1976; Roberts et al., 1996). Additionally, there is evidence for clumping or beading of dendritic components (Lavenex et al., 2009; Williams et al., 1978). However, total loss of dendritic cell membranes is not a commonly reported early phenomenon in the PMI. As a result, dendrite tracing in volumetric microscopy data may be possible even in brains with relatively longer PMIs if robust visualization methods are used. Certain morphological staining methods, such as the Golgi-Cox impregnation method, are more resistant to PMI artifacts than others. Experiments labeling for particular biomolecules, such as MAP2 or SMI32, should not be considered dispositive of the state of dendrite morphology more generally.”

It doesn’t really make sense to me how dendritic spines could retract so quickly in global cerebral ischemia. That seems like an active process requiring ATP and ATP is rapidly depleted in the postmortem period. There is also data suggesting this “retraction” is reversible up to a fairly long time period with reperfusion, which doesn’t really make sense if it is true information loss.

However, this clearly needs more study. One of the things we noted in our review is that nobody has ever done a time course study of the effects of the postmortem interval using volume electron microscopy. The best way to do this study, in my opinion, would be to do it in rats: perform euthanasia, wait a variable number of minutes/hours, remove the skull, take a biopsy sample, immerse it in formaldehyde/glutaraldehyde, wait until fixed, embed it, and then do volume EM on it. Doing it right would involve a sufficient number of rats in each group to each account for the inevitable biological variability.

ChatGPT is doubling down. It claims that after a few minutes of degradation, any residual spine has been reduced to a bump. This represents a complete loss of information because it becomes a one-to-many problem. There is no longer any way to know which bouton the spine remnant previously connected with. The studies that show longer persistence actually just represent a semantic disagreement because they show remnants that do not contain enough information to infer original connectivity. The fast-collapse studies seem to be robust. ATP is required to construct and maintain actin scaffolds, but actin depolymerization is not dependent on ATP at all. Actin depolymerization and dendritic spine collapse are thermodynamically favored when ATP is lost so collapse happens within seconds to minutes and continues rapidly until there is no more structural actin remaining. Recovery seems to have a sharp cutoff of about 6 minutes.

Yes, I fully realize I've gone down a ChatGPT rabbit hole, and that it's behaving like an idiot savant. Still...

Interesting! It would be helpful if you could link to the chat or ask chatgpt what the main sources it is basing these claims on.

Some of the articles in this space visualize redistribution of fluorescent proteins and call it “spine loss.” And then the spines “come back” at mostly the same locations if there is rapid enough reperfusion. I don’t see how would the neuron know where to rebuild the connection if the information was lost. So maybe it relates to the visualization method. Example article: https://pubmed.ncbi.nlm.nih.gov/18272696/

But it’s also possible that there are other articles that I am not aware of.

I started exploring scenarios where ChatGPT might have gotten it wrong. Like maybe the information is substantially encoded at the epigenetic level and somewhat random individual synapses could be recreated. That's a a hard no. If the spines all completely retract in under 5 minutes due to loss of ATP, then what we are left with is a synapse-free connectome that is fairly stable for about 6 hours, just like I always envisioned. But it's a blank slate. All the boutons are still largely intact, but here's why that's not helpful. Dendritic shafts are crowded with passing axons. The boutons on those axons don't even all correspond to previous synapses. The axons seem to have boutons all over the place just in case a dendritic spine wants to connect. With that many boutons and that much density, any attempt to recreate connections will just be random noise. It seems like the gross structures of the axons and dendrites is fairly stable throughout life and even after death for hours. But memories are all encoded at the level of synapses which are far more fragile than I ever imagined they could possibly be. Memories seem to be entirely dependent on ATP to maintain structure. Once they degenerate within 5 minutes, then the mind is lost.

If this is true, it would change our business model completely. I'll give it a few days to think about it because most new information of this kind is typically wrong and misunderstood. But if it's true, then we may need to shift completely to a Medical Aid in Dying (MAID) model, and we would probably even need to induce premortem hypothermia. The significance of this information could be absolutely game changing. I'm also obviously concerned that MAID laws are subject to change on a whim, possibly even in reaction to what we choose to do.

The information is definitely not encoded in the nucleus, that's true.

So your claim is that all dendritic spines retract and synapses are lost within 5 minutes of complete global cerebral ischemia?

This seems pretty inconsistent with the available evidence. See, for example, the section of our review on synapses: https://pmc.ncbi.nlm.nih.gov/articles/PMC10294569/#S7.6

The current scientific understanding of dendritic spine retraction is about 10 years old. Prior to that it would have been more reasonable to think that the synapses persisted for hours, just like the neurons themselves. (this was not in reply to your link above. We posted simultaneously)

Ugh. ChatGPT can just be so dumb sometimes while sounding so smart. It's now admitting that the synapses and connections are stable. But it's still not backing down on its claim of total information loss. I don't really understand that claim. It seems to now be claiming that it's impossible to tell what the original synaptic weights were. It threw a lot of detail at me regarding all the different molecules that have been disturbed and how it's now completely impossible to know what the original weights were. I'm skeptical. So I think I've hit the end of this rabbit hole. I have no way to interpret that.

Claude gave me a straight answer. The spine does retract somewhat, but the connectome is not lost. Some of the weight information is probably lost, but much molecular detail remains at the synapse that might allow inference. Finally, we don't know how much is lost. It depends whether you're talking to an optimistic or pessimistic neuroscientist.