Hacker News Reader

Comment by burning_hamster | Hacker News Reader

> "The neurons do not fire action potentials, and do not express any voltage-gated sodium channels."

This an old and incorrect belief that largely derives from the difficulty of putting electrodes into their teeny, tiny neurons. Close relatives of C elegans that are larger (and hence more easily experimented on) do have action potentials, and for some neurons in C elegans, we also have good evidence of action potentials [1, 2]. Absence of evidence is not evidence of absence.

[1] Lockery SR, Goodman MB. The quest for action potentials in C. elegans neurons hits a plateau. Nat Neurosci. 2009 Apr;12(4):377-8. doi: 10.1038/nn0409-377. PMID: 19322241; PMCID: PMC3951993.

[2] Jiang, J., Su, Y., Zhang, R. et al. C. elegans enteric motor neurons fire synchronized action potentials underlying the defecation motor program. Nat Commun 13, 2783 (2022). https://doi.org/10.1038/s41467-022-30452-y

View on HN

Well, the 'canonical' action potential is mediated by sodium currents, so it's maybe not surprising that people concluded that C elegans don't have APs given that a) they don't have any genes for voltage-gated sodium channels, and b) when people had recorded from C elegans neurons (it's hard but not impossible), they had never seen action potentials. (So it's not like no one had looked, and then had concluded that they don't exist. They looked and didn't see them.) In the paper that originally reported APs in C elegans (Liu et al 2018), they were looking in a specific neuron (AWA), and they had to elicit a 'plateau potential' by depolarizing the cell for a while before the spikes were revealed, riding on top of the plateau.

The APs discovered by Liu et al (2018) are generated by calcium, not sodium currents, so one could even argue that they aren't action potentials in the strict sense. Also, they seem to be rather difficult to elicit, and it's still not clear whether neural computation in C elegans is mostly AP-mediated, or if APs are the exception rather than the rule.

Liu, Q., Kidd, P. B., Dobosiewicz, M. & Bargmann, C. I. C. elegans AWA olfactory neurons fire calcium-mediated all-or-none action potentials. Cell 175, 57–70 e17 (2018) https://doi.org/10.1016/j.cell.2018.08.018

> are generated by calcium, not sodium currents, so one could even argue that they aren't action potentials in the strict sense

Does the underlying chemistry define if its an action potential or not? I thought an AP just needed a voltage differential regardless if its from calcium or sodium.

Given that we have a simulator of this worm right there (which includes it moving), can it really be up to debate whether it uses action potentials or not?

I'd think the simulation has to get it right, and so needs to simulate action potentials if the worm has them, or not simulate them (but whatever the worm has instead) if not, right? Or could the simulation still be incorrect and only based on current assumptions, but getting this wrong still allows some worm-like behavior?

I really wish the readme/FAQ would talk a bit more about the worm and the simulation, rather than have 80% of their content be about Docker, though, so that I could learn more what cells it actually simulates.

Not necessarily, because you could also simulate the worm without neurons at all. It's the closeness of the simulation to the real thing that demands that it is done right and the question effectively is: is this simulation close enough that if such a detail would be wrong that it would fail?

One way to answer that would be to add and remove such mechanisms to see if it would lead to different behavior.

That isn't immediately true, with enough fitting parameters you can capture the effect underlying behaviour without explicitly capturing it, or even without knowing it exists.

"With four parameters I can fit an elephant, and with five I can make him wiggle his trunk." - John von Neumann [0]

[0]: https://doi.org/10.1119%2F1.3254017

Better information on what it actually simulates is here: https://docs.openworm.org/projects/

As of this year there are 3 known neuron classes in C. elegans that do exhibit action potentials. The rest exhibit graded potentials.

I suspect slarsOn is the founder of the project (also my old roommate). (Stephen, maybe post a top-level comment/introduction? Also, hi!)

How did researchers before that explain what the neurons do if they believed they did not have action potentials? Did they believe communication was done solely through chemical messaging?

You don't need spikes to have computation. Deep networks don't have spikes.

Classical action potentials are just one mechanism of INTRAcellular communication - You could think of it as a special case of signaling via chemical concentration, where the chemical is cations and the propagation is faster+more directed than diffusion. INTERcellular signaling is only rarely mediated directly by voltage. Also, action potentials are most "useful" for propagating a signal rapidly over a long distance - It kind of accelerates and error-corrects (= reverses diffusive broadening) voltage signals down a linear path. Action potentials are so well known mostly because they show up in stuff that's easy to observe (long motor neurons) and they're easy to quantify

Somewhat related, there is a roughly inverse correlation between neuron count and "computational power per neuron", "older and simpler" critters' neurons are more likely to be "less specialized" and more likely to use hundreds of different chemicals for transmitting intercellular signals, while "newer and more advanced" critters' neurons are more likely to be "specialized" and use just one chemical for transmitting intercellular signals

Neural computing without action potents is commonplace. Computational interactions among cells and neurons in retina are almost all graded potentials that modulate transmitter release or conductances through gap junctions. Retinal ganglion cells of course do generate conventional spikes—to pass a data summary to midbrain. hypothalamus, and dorsal thalamus.

Action potential are almost strictly INTRAcellular events (minor exception being ephaptic effects) that are converted in a surprisingly noisy way into presynaptic transmitter release and variable postsynaptic changes in conductances.

Action potential are a clever kludge necessitated by being big and having long axons and needing to act quickly.

Reply on news.ycombinator.com