For the first time, scientists have discovered that sleep can be detected by patterns of neural activity lasting just milliseconds, 1,000 times shorter than a second, revealing a new way to study and understand the fundamental brainwave patterns that govern consciousness. They have also shown that small regions of the brain can momentarily “flicker” and wake up while the rest of the brain remains asleep, and vice versa, from wakefulness to sleep.
These results, described in a new study published in the journal Neuroscience of Natureare a collaboration between the labs of Assistant Professor of Biology Keith Hengen at Washington University in St. Louis and Distinguished Professor of Biomolecular Engineering David Haussler at UC Santa Cruz. The research was led by doctoral students David Parks (UCSC) and Aidan Schneider (WashU).
Over the course of four years of work, Parks and Schneider trained a neural network to study patterns within massive amounts of brainwave data, discovering patterns that occur at extremely high frequencies that have never been described before and challenge long-held, fundamental understandings of the neurological basis of sleep and wakefulness.
“With powerful tools and new computational methods, there is much to be gained from challenging our most basic assumptions and reexamining the question of what a state is,” Hengen said. “Sleep or wakefulness is the single most important determinant of your behavior, and everything else follows from that. So if we don’t understand what sleep and wakefulness actually are, it seems like we’ve missed the boat.”
“As scientists, we were surprised to find that different parts of our brain actually take short naps when the rest of the brain is awake, even though many people may have already suspected this in their spouse. Perhaps it’s the lack of gender bias that’s surprising,” Haussler joked.
Understanding Sleep
Neuroscientists study the brain using recordings of electrical signals of brain activity, called electrophysiological data, by observing voltage waves as they rise and fall at different rates. These waves contain the spiking patterns of individual neurons.
The researchers worked with data from mice at the Hengen lab in St. Louis. These free-living animals were equipped with a very lightweight helmet that recorded brain activity in 10 different brain regions for months, tracking the voltage of small groups of neurons with microsecond precision.
This amount of information generated petabytes of data, a million times larger than a gigabyte. David Parks led the effort to feed this raw data into an artificial neural network, which could detect extremely complex patterns, differentiate between sleep and wake data, and find patterns that human observation might have missed. A collaboration with the shared university IT infrastructure Located at UC San Diego, it allowed the team to work with such a large amount of data, which was on the scale of what large companies like Google or Facebook could use.
Knowing that sleep is traditionally defined by slow waves, Parks began feeding the neural network smaller and smaller chunks of data and asked it to predict whether the brain was asleep or awake.
They found that the model could tell the difference between sleep and wakefulness from just a few milliseconds of brain activity data. This shocked the research team because it showed that the model couldn’t rely on slow waves to learn the difference between sleep and wakefulness. In the same way that listening to a thousandth of a second of a song doesn’t tell you whether it has a slow beat, it would be impossible for the model to learn a beat that occurs over several seconds just by looking at isolated milliseconds of information.
“We’re seeing information at an unprecedented level of detail,” Haussler said. “We previously thought that nothing would be there, that all the relevant information was in the slower frequency waves. This study shows that if you ignore the conventional measurements and just look at the details of the high-frequency measurement over just a thousandth of a second, there’s enough data to tell whether the tissue is asleep or not. This tells us that something is happening on a very rapid scale – it’s a new clue to what might be happening during sleep.”
Hengen, for his part, was convinced that Parks and Schneider had missed something, because their results were in complete contradiction to the fundamental concepts that had been instilled in him during his many years of studying neuroscience. He asked Parks to provide more and more evidence that this phenomenon was real.
“It made me wonder how much of my beliefs are based on evidence and what evidence would I need to see to overturn those beliefs?” Hengen said. “It really felt like a cat-and-mouse game, because I was asking David [Parks] “He kept asking me over and over to produce more evidence and prove things to myself, and he said, ‘Look at this!’ It was a really interesting process as a scientist to watch my students tear down these towers brick by brick, and to have to accept that.”
Local models
Because an artificial neural network is fundamentally a black box and doesn’t report what it learns, Parks began removing layers of temporal and spatial information to try to understand what patterns the model could learn from.
Eventually, they got to the point where they were looking at chunks of brain data that were just a millisecond long and had the highest frequencies of brain voltage fluctuations.
“We took all the information that neuroscience has used to understand, define, and analyze sleep over the last century, and asked whether the model could still learn under these conditions,” Parks said. “That allowed us to look at signals that we hadn’t understood before.”
By analyzing this data, they were able to determine that the model was detecting the fundamental element of sleep, namely the ultra-fast pattern of activity between just a few neurons. Crucially, these patterns cannot be explained by traditional, slow, generalized waves. The researchers hypothesized that the slow waves might act to coordinate the fast, local activity patterns, but they ultimately concluded that the fast patterns are much closer to the true essence of sleep.
If we compare the slow waves traditionally used to define sleep to thousands of people in a baseball stadium doing the wave, then these fast patterns are the conversations between just a few people who decide to participate in the wave. These conversations are essential for the overall wave to occur and are more directly related to the atmosphere in the stadium – the wave is a secondary result.
Observing the flickers
As the researchers looked more closely at hyperlocal activity patterns, they began to notice another surprising phenomenon.
Looking at the model predicting sleep or wakefulness, they noticed what looked like errors at first glance: For a split second, the model detected wakefulness in one region of the brain while the rest of the brain remained asleep. They observed the same thing in wakefulness states: For a split second, one region fell asleep while the other regions were awake. They call these cases “flickers.”
“We were able to look at the individual times when these neurons fired, and it was pretty clear that [the neurons] “We’re moving to another state,” Schneider said. “In some cases, these flickers may be limited to a specific brain area, maybe even smaller than that.”
This prompted researchers to explore what flickers might mean about sleep function and how they affect behavior during sleep and wakefulness.
“There’s a natural assumption about this: Let’s say a small part of your brain goes to sleep while you’re awake. Does that mean your behavior suddenly looks like you’re asleep? We started to see that that was often the case,” Schneider said.
By observing the behavior of the mice, the researchers observed that when one region of the brain fell asleep while the rest of the brain was awake, the mouse paused for a second, almost as if it had fallen asleep. A flicker during sleep (a region of the brain “waking up”) translated into the animal’s tics during sleep.
The flickers are particularly surprising because they do not follow the established rules dictating the brain’s strict cycle of moving sequentially between wakefulness, non-REM sleep, and REM sleep.
“We see alternating REM sleep and wakefulness, REM sleep and non-REM sleep. We see all these possible combinations, and they break the rules that you would expect based on a hundred years of literature,” Hengen said. “I think they reveal the separation between the macro state (sleep and wakefulness at the whole animal level) and the fundamental unit of state in the brain (the fast and local patterns).”
Impact
A better understanding of the high-frequency patterns and fluctuations between wakefulness and sleep could help researchers better study neurodevelopmental and neurodegenerative diseases, both of which are associated with sleep dysregulation. Both Haussler and Hengen’s lab groups want to better understand this connection, with Haussler hoping to study these phenomena in more detail in brain organoid models, pieces of brain tissue grown on a lab bench.
“This potentially gives us a very, very sharp scalpel with which we can address these disease and disorder questions,” Hengen said. “The more fundamentally we understand what sleep and wakefulness are, the more we can address the relevant clinical and disease-related issues.”
At a fundamental level, this work helps advance our understanding of the brain’s many layers of complexity as an organ that dictates behavior, emotion, and much more.