Capturing the Whole Brain in Action

New technologies that record the entire brain as animals think, move and make decisions could provide insight into global dynamics

Andrew Leifer and collaborators can capture activity from more than 100 neurons in the worm brain. Credit: Leifer lab
Andrew Leifer and collaborators can capture activity from more than 100 neurons in the brain of a moving worm.

For the most part, scientists record neural activity at very small or very large scales. Functional magnetic resonance imaging (fMRI) can image the whole brain, but it captures an indirect and relatively slow proxy for activity, and it offers poor spatial resolution. Electrophysiology techniques, by contrast, can track millisecond-scale activity, but typically only in tens or hundreds of cells. Neither technique can give scientists a comprehensive picture of what the brain is doing. Researchers recording from small groups of neurons often worry they are missing a vital part of neural signaling, particularly how circuits in different brain regions communicate.

“We need to look at large-scale dynamics across multiple areas at the same time to understand how they talk to each other,” says Misha Ahrens, a neuroscientist at the Howard Hughes Medical Institute’s Janelia Research Campus in Ashburn, Virginia. Ahrens is a Simons Collaboration on the Global Brain (SCGB) investigator who studies whole-brain imaging in zebrafish. “No behavior is implemented by a single brain area, except maybe some reflex arcs,” he says.

An emerging class of technologies allows scientists to record neural activity at near single-cell level across the entire brain as animals behave and make decisions. Scientists hope that such an exhaustive map of brain activity will reveal how neurons and circuits function as a unit and how they drive behavior. “We have good general principles for how a single neuron works, but we haven’t found universal principles about how collections of neurons work together and how they affect cognition and behavior,” says Andrew Leifer, a neuroscientist at Princeton University and an SCGB investigator. Leifer has developed whole-brain imaging techniques for studying freely moving worms. “If we could record from many neurons, ideally all of them at the same time, we might be better situated to find the general principles,” he says.

At the 2016 SCGB annual meeting in September, three researchers described their approaches to whole-brain imaging in different animals — Leifer in C. elegans; Sophie Aimon, a postdoctoral fellow at the University of San Diego, in fruit flies; and Florian Engert, a neuroscientist at Harvard University, in zebrafish. Their findings suggest that these techniques, so far applied only in simple organisms with small brains, can indeed reveal activity patterns that other imaging modalities miss. The approach is beginning to reveal how circuits produce behavior.

The Squirming Worm

With its small size and transparent tissue, the microscopic worm C. elegans is ideal for whole-brain imaging. The animal has just 302 neurons, about 150 of which make up its tiny worm brain. Last year, Leifer and his collaborators described in the Proceedings of the National Academy of Sciences a new approach for imaging a worm’s brain as it crawls around. The technique combines calcium imaging with automated tracking of the worm’s movement. “With these kinds of optical recording technologies, we can record from nearly all the neurons in the head,” Leifer says. The researchers routinely record 100 to 120 neurons. They are continuing to develop and refine algorithms that reduce artifacts caused by movement and compression of the worm’s soft head.

One of the unique aspects of Leifer’s approach is that it works in freely moving animals. That’s particularly important for studying how neural dynamics drive behavior. “In C. elegans, most behavior has to do with motion, so it’s important for me to study the worm as it moves, awake and unrestrained,” Leifer says.

C. elegans’ sinusoidal crawl makes it relatively easy for researchers to quantify its behavior. Scientists have previously shown that a large fraction of the worm’s crawl can be described by four distinct postures, which resemble sine and cosine waves at different spatial frequencies. Researchers can describe any posture that the worm makes by simply summing these four modes. “Now you have a language for talking about the animal’s behavior that’s convenient mathematically,” Leifer says.

With that behavioral language in hand, researchers are trying to reverse engineer how different patterns of neural activity generate behavior. They search for specific activity patterns that relate to the animal’s speed, turning status and direction. Then they build decoders to predict aspects of how the animal is moving or is about to move based on its neural activity. “What’s exciting is looking at all neural activity and predicting completely unrestrained behavior,” Leifer says.


This video shows the patterns of brain activity in a freely moving worm sped up 5 times. The top panel shows the location of all of the neurons in a stereotyped coordinate system. Their size represents the activity of the neuron as measured from the calcium indicator. The bottom left panel shows the animal’s behavior. The green line indicates the worm’s posture. The bottom right is the raw video of the calcium signal we see in the head of the worm.

Leifer and his collaborators also want to discern the extent to which single neurons (versus collections of neurons) drive behavior in worms. Previous research has suggested that certain worm behaviors are driven by individual neurons. But Leifer says those findings might reflect the technology used in the experiment rather than biology. “Whole-brain recordings will hopefully shed light on that,” he says.

Simply tracking neural activity isn’t enough — scientists want to control it as well. Leifer and his collaborators have developed optogenetics tools to disrupt specific neurons and circuits, which will help uncover their role in behavior. “The future will be about combining recordings with other experiments where you drive neural activity,” Leifer says. “That allows you to not just generate hypotheses, but to test them.” Most studies of this sort have targeted small numbers of neurons or focused on stereotyped behavior, such as an arm reach or eye movement. But Leifer’s team is developing tools to control larger groups of cells.

The Fascinating Fly

Sophie Aimon, an SCGB fellow, is studying another laboratory stalwart, the fruit fly.

Flies are more complex than worms, both in their anatomy and their behavior. The fly brain has about 100,000 neurons, three orders of magnitude more than C. elegans. “It makes more complex decisions and helps bridge the gap between simple organisms and more complex ones,” Aimon says.

Sopie Aimon is developing whole brain imaging techniques for fruit flies.
Sopie Aimon is developing whole brain imaging techniques for fruit flies.

Aimon uses a relatively new technique known as light field microscopy to capture neural activity across the fly’s entire brain. The technology is similar to standard fluorescence microscopy, which measures emitted light from a single plane of the brain. Both monitor neural activity via fluorescent probes that detect changes in calcium levels or voltage. But in light field microscopy, a series of microlenses separate light rays from different angles. Scientists can then use information about the position and angle of the rays to reconstruct the whole volume of the brain. “Light field microscopy allows us to image the whole brain really fast, much faster than anything I know of,” she says. Its resolution comes close to the millisecond timeframe of action potentials — a fly’s brain can be imaged in just a few milliseconds.

The technique is much cheaper than other approaches for live imaging, such as confocal or two-photon microscopy, but it has a lower spatial resolution. Aimon and her collaborators have developed computational techniques, including principle component analysis and independent component analysis, to disentangle signals from a few to tens of neurons, getting close to single-neuron resolution.

Proof of principle experiments recording flies’ responses to light and odor are promising. Neural activity recorded via light field microscopy generally matches that recorded using electrophysiology or imaging of small brain regions. But Aimon also sees some novel patterns, such as neural oscillations that haven’t previously been reported in flies. She notes that this might result from imaging the neuropil — a dense network of nerve fibers — instead of the cell body, the target of electrophysiology. “Since no one has looked at whole-brain dynamics in flies, it’s all kind of new,” Aimon says.

Like Leifer, Aimon is employing her imaging technique to study the link between whole-brain activity and behavior. For light field microscopy, the fly’s head is fixed for imaging. But the animal is positioned on a rotatable ball so that it can still walk, groom and make other movements. Aimon has found that fly brains are quite active while the animal is walking but show little brain activity during grooming. Though it’s unclear why this is the case, she speculates that perhaps walking requires more thought, such as making navigation choices, whereas grooming is a relatively mechanical activity.

Aimon aims to link different networks to specific behaviors and to figure out what drives flies to switch their behavior. Why does a fly start or stop walking or change direction, for example, all without any external stimulation? Because Aimon can capture this entire repertoire of behaviors under the microscope, “whole-brain imaging will be a good tool for understanding where these behaviors come from,” she says.

So far, Aimon has employed both calcium imaging and voltage-sensitive probes. The former generates the best signal-to-noise ratio, allowing researchers to visualize activity from a few hundred brain structures. Voltage-sensitive probes can detect only the most active structures — researchers can currently track activity from a few tens of them — but have better temporal resolution. Early experiments with these probes hint at interesting spatial temporal patterns undetectable with other techniques. For example, Aimon has recorded a wave of activity that propagates simultaneously in two parts the brain when the fly turns, as well as activity that flip-flops from high to low in two distant brain regions.

Aimon’s goal is to record signals from the whole brain at a speed that rivals electrophysiology’s fast timescale. “This should help bridge the gap between the current fast, local methods and slow, large-scale methods, and maybe reveal new effects that are only visible in this fast and large-scale range,” she says.

A Fish Tale

Florian Engert is using  two-photon calcium imaging to study the whole brain in immobilized zebrafish, a technique he first published with Ahrens in 2013. Engert and his collaborators are now using that technique to try to understand a complete behavioral repertoire, from sensory input to motor output. They are focused on the the fish’s optomotor response, a reflex that stabilizes the animal’s position with respect to its surroundings. In simple terms, when a fish sees leftward motion, it turns left.

To study this reflex, scientists show an immobilized fish a moving grating — the animal thinks it’s swimming and periodically beats its tail to maintain its pace. Researchers track the animal’s response to the changing visual scene by measuring activity in the tail, a proxy for the fish’s desired swimming speed and direction.

Neural activity as the fish views leftward motion towards each eye. Neurons that respond best to lateral motion are shown in green. Those that respond best to medial motion are shown in purple.
Neural activity as the fish views leftward motion towards each eye. Neurons that respond best to medial motion are shown in green. Those that respond best to lateral motion are shown in purple. Cell 2016

In research published in Cell in November, Engert and his collaborators developed a network model of the zebrafish optomotor response and then used whole-brain imaging to test predictions from the model. For example, one of the model’s 10 cells responds to leftward motion, driving the system to turn left. It responds more strongly when both eyes see leftward motion. The researchers found cells with similar properties in the real fish via imaging.

One of the insights that arose from the model was the importance of inhibition, such as how the circuit stops a fish from swimming in the wrong direction. “If you want to turn left, you don’t just instruct muscle do to the appropriate action,” Engert says. “You also have to suppress the action in the wrong direction.”

Whole-brain imaging has also helped the researchers determine which aspects of the environment are most important in guiding behavior. It turns out that fish care whether movement is directed toward or away from them rather than side to side. “The fish has an inward or outward channel, not a left or rightward channel,” Engert says. “That’s something that emerged from the iterative process of modeling and imaging.”

Engert and his collaborators now plan to use the same approach to study other behaviors, such as phototaxis, in which fish swim toward light, and predator avoidance, when fish avoid looming objects. The researchers will build models for the different behaviors and then try to integrate them into one larger model. They could ask, for example, how the animal would react if given conflicting chemical and visual information. “We want to look at different behaviors that, when put together, give complex behaviors that have to do with decision-making,” Engert says.

To date, these techniques are limited to fairly simple organisms. But the researchers hope that they will be able to elucidate general neural principles at work in these small brains and then search for them in more complex animals. “Whole-brain imaging gives you two things — confidence and resolution,” Engert says. “Confidence that the model is correct, and the single-cell resolution you need to understand circuits.”

 

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Save

Recent Articles