Long-sought walking circuit found in fruit flies

Many animals, from cockroaches to cats, can walk without input from the brain. Yet scientists have struggled to pinpoint the responsible rhythm-generating circuit, or central pattern generator (CPG), in the spinal cord in any organism. 

“You look at textbooks, and people just draw a circle and write CPG in it,” says John Tuthill, professor of neurobiology and biophysics at the University of Washington. 

Tuthill says he had even started to doubt the CPG’s existence—until he and his team used connectome-based modeling to pinpoint one in Drosophila melanogaster. The network of just three neurons in the ventral nerve cord produces the oscillatory neuron activity required for walking, the team reported in March at the Computational and Systems Neuroscience (COSYNE) annual meeting and in a preprint posted on bioRxiv in April.

Another putative CPG in fruit flies surfaced in an independent team’s study posted on bioRxiv earlier this month. This circuit also generates rhythmic motion without peripheral input, connectomics and optogenetics showed.

“Now we know the neurons involved, which is a big step forward,” says Graziana Gatto, professor of neurobiology at the University Hospital Cologne, who was not involved in either preprint. These studies “capture the way neuroscience is going right now. Mapping the connectome has been a huge revolution,” she adds. 

A

The neurons in the recurrent neural network-based model they created can send feedback signals to one another in loops, and their activity depends on both present input and previous states. Stimulating descending neurons caused motor neurons to oscillate spontaneously, without any training or fine-tuning of the network, the new study shows.

The fly connectome lacks information about neurotransmitters and other biophysical properties of neurons, Brunton says, so she and her colleagues took a “brute-force approach” to assign various biologically plausible properties to neurons in the simulated network. A large majority of the simulations, which vary slightly, produced similar results. And the precise number of neurons in the connectome didn’t affect oscillations.

Most of the neurons descending from the simulated central brain to the ventral nerve cord don’t produce oscillations, the researchers found by activating each descending neuron one at a time. DNg100, which is already known to drive forward walking in flies, produced the strongest rhythmic activity. Optogenetically stimulating DNg100 drives forward walking, whereas optogenetically stimulating DNb08 drives rhythmic waving-like leg movements in living flies, Tuthill’s team discovered.

“When John showed me that it works, I was like, ‘Wow, this is amazing,’” says Salil Bidaye, research group leader at the Max Planck Florida Institute for Neuroscience, who was not involved in the April preprint but previously pinpointed DNg100 in forward locomotion and worked on the preprint posted earlier this month. 

A three-neuron CPG, which is downstream of DNg100 and composed of two excitatory neurons and one inhibitory neuron arranged in a loop, is sufficient to produce the oscillatory activity, the researchers found by computationally eliminating neurons until the rhythm stopped.

That arrangement goes against the half-center model, says Rune Berg, associate professor of neuroscience at the University of Copenhagen, who was not involved in either new preprint. That “traditional view of rhythm generation” proposes that rhythmic output emerges from groups of neurons that reciprocally inhibit one another, like a seesaw. “There’s no reciprocal inhibition in [the putative fly CPG] network, which is a bit surprising,” Berg says.

DNb08 controls a different rhythmic circuit, composed of two of those three neurons and three additional ones, suggesting that this descending neuron might drive a different behavior that involves rhythmic movement, such as climbing or grooming. 

This is the “only time in my career where modeling and theory was leading our insights significantly,” Tuthill says. 

B

These descending neurons converge on a network of five neurons, one of which overlaps with those in the three-neuron circuit that Tuthill and his colleagues described, Bidaye’s new preprint shows; two neurons in the network are also part of the five-neuron circuit Tuthill’s study describes, downstream of DNb08. The congruity “was really encouraging,” Bidaye says. 

This five-neuron motif repeats across all legs and is directly upstream of leg motor neurons. The five-neuron circuit sends commands to commissural neurons that inhibit the opposite-side circuit, Bidaye and his team found, providing a putative mechanism for left-right leg alternation in walking flies. Commissural neurons also receive input from the brain, the study suggests.

But both studies pinpointed neurons that the other didn’t. Neither group knows exactly why, but both point to the differences in their experimental approaches. The modeled circuit, for example, didn’t show any left-right coordination, and only some motor neurons exhibited strong oscillatory behavior. “That tells us that it’s not the whole story,” Bidaye says. Brunton and Tuthill are currently searching for neurons that coordinate left-right movement, they say. 

But Bidaye’s tracing strategy may have similarly missed components of the circuit, because it looked only at direct connections between descending inputs and neurons in the cord, says Ronald Calabrese, professor emeritus of biology at Emory University, who was not involved in either of the new preprints. 

B

Bidaye and his team used high-speed cameras to track the motion of fly legs while headless flies walked on a regular or slippery surface, an experimental setup that prevented the legs from coordinating via peripheral feedback. They performed similar experiments with flies suspended in the air and flies that had their legs amputated at the midpoint of the femur.

Optogenetically stimulating three descending neurons—DNg100, DNg097 and MDN—drove walking under all conditions (amputated flies moved the parts above the amputations). On solid ground, DNg100 and DNg097 drove forward walking, and MDN drove backward walking. Further experiments suggested that proprioceptive input was necessary to coordinate all six legs when DNg097 and MDN received optogenetic stimulation.

Stimulating multiple descending neurons allowed the fly to walk faster, the study showed, and slowing down required proprioception.

“The data are beautiful,” Calabrese says. “They’ve really pinned things down with the behavioral analysis with what’s central and what’s peripheral.”

But it’s unclear whether the fly findings apply to mammals. 

Rhythmic movement generation may instead be an emergent property of a large high-dimensional network rather than one dedicated circuit, some studies in mammals suggest. 

But these two views of rhythm generation aren’t exclusive, Berg says. “A CPG is just a circuit that produces some rhythmic motor output, It doesn’t mean that it’s the only thing producing that output.”

Still, the field has a long way to go to find the CPG in mammals, Gatto says. “But now we know what neurons are involved. We can finally test all of these theories in a controlled system and start to apply it in mammals.”