Hail Hydra! Doing neuroscience without a brain
We’ve reached the milestone of recording every neuron in a brain at the same time
What’s the best way to understand a brain? A brain, any brain at all. In the absence of a theory telling us what we should be looking for, many believe the answer is: record everything. Every neuron. So that rules out our brain, obviously. With around 86 billion neurons, understanding our brain is as far out of our reach as understanding quantum chromodynamics is to my cat Charlie (and he’s the smart one; to Bob it’s as far as understanding how the cat flap works). So we need to aim at something, some creature, with a brain that’s a little more modestly proportioned.
And how will we know when we’ve recorded every neuron? Ideally because we can see every neuron we’ve recorded from. But that’s impossible: how can we find a creature with a tiny brain in which we can record and see every neuron?
Enter Hydra. A cnidarian, with whom our shared common ancestor is likely more then 500MYA; a worm-like thing with tentacles, about 1cm long with two stunningly useful features. For one, all its neurons are spread throughout its entire body, not irritatingly grouped into clusters like in C. elegans, or flys, or us. (Technically, this is cheating as Hydra therefore has no “brain”, just a “nerve net”. But all we care about are whether it has neurons that link together and control the creature, which it does). And the other thing: it is completely see-through. That means we can video the neurons by simply videoing the animal (as we can see into the animal) and (as the neurons are all spread out) we can easily separate individual neurons when we analyse the video. The key then is to work out how to video not just the neurons, but the activity of those neurons.
Christophe Dupre and Rafael Yuste did just that. They engineered their Hydra so that the Hydras’ neurons expressed a protein that emits green light in proportion to how active a neuron is. So by watching their videos they could work out each neuron’s activity in each frame of the video just by recording how green it was. Of essentially every neuron in its body. And all while the creature was wriggling around under the microscope.
Also useful about Hydra, parts three and four. They can reproduce asexually by budding; which creates a clone, so you have a set of practically identical individuals to record from. The slight wrinkle here is that how many neurons a Hydra has is not fixed: it depends on the age of the creature, and ranges from a few hundred to a few thousand. So genetically identical creatures do not have identical nerve nets. Oh, and another useful feature is that they are seemingly immortal — so you can record from them for a longggg time.
(One for the cognoscenti: It turns out that Sydney Brenner himself recommended this team study Hydra to crack the problem of recording every single neuron. Because sometimes the problem with C Elegans is it’s just too complicated).
What did Dupre and Yuste find? Simplicity itself.
They found three different networks of neurons spread throughout their Hydra. Two in the outer layer, one on the inside. (Hydra is, roughly speaking, a cylinder with a thick shell and tentacles on the top. Yep, it’s a Cheese String). Each network changed its activity for one specific behaviour: one network for getting longer when exposed to light; one for getting narrower; and one for getting shorter. And a set of neurons at the base of the tentacles whose activity changed when Hydra nodded.
Most interesting here was that the network controlling “get shorter” consistently reduced its activity to get the Hydra to shrink. What this suggests is that the constant activity of this network maintains body length. Or, to put it another way, it maintains the Hydra’s posture. Though I’m willing to bet Hydra do not suffer from bad backs, on account of not having one.
This simplicity was most evident in how the activity in each network was organised. Each frame of their video was 100 milliseconds long. And if one neuron of a network was active in that frame, then every single other neuron in the network was also active in that frame. Every single neuron in the network at the same time. So instead of having to keep track of the activity of hundreds of neurons, they just needed to keep track of three numbers: is network one on? is network two on? is network three on? And they could simply plot out the yes/no answers to these questions to see the entire nerve net’s activity.
This is a staggering example of dimension reduction: from hundreds of neurons, to three yes/no questions. For others, like me, who tackle more complex brains, this was the most cool part. And also induced jealousy: there is an outside chance that we may one day genuinely understand something so simple. It would be so cool if the brains of more complex creatures were so simple, but they’re not.
Though maybe Hydra has more insight to yield here. It has more complex behaviours in its repertoire — feeding, for example. And these need co-ordination between these three different networks, as well as sensory information being used by them to guide behaviour. So now we can chase down their complexities: how are the networks linked within them to get that dramatic synchrony? And between them to co-ordinate actions?
(And the Achilles heel of this study is that its results are just changes in activity, just correlation. So clinching insights will come from testing causality: by activating networks, and checking they do directly control behaviours; by knocking out networks, and checking the behaviour goes away. But given the stark one-to-one correspondence between each network’s activity and a behaviour, it seems unlikely these experiments would radically change the conclusions reached by Dupre and Yuste. Still, in science, one can never know where surprises are in store. Otherwise, they wouldn’t be surprises).
Dupre and Yuste’s study is a startling display of how rapidly neuroscience is advancing. We now have a cast-iron demonstration of recording (almost) all the activity from (almost) every single neuron of a creature. Is this “understanding”? Not yet. It is observation, not testing. It is data, not theory. Yet it can’t but sharpen our thoughts about what we mean by “understand a brain”.
Want more? Follow us at The Spike