Deciphering the Link Between Neural Activity and Blood Flow in the Brain

Scientists often use blood flow in the brain as a stand-in for neural activity. Generally, increases in neural activity are accompanied by increases in blood flow. But an array of studies over the past two decades indicates that this relationship is not clear-cut and can vary across different brain regions and mental states. This could have implications for brain imaging techniques like functional MRI, or fMRI, which treat blood flow signals as proxies for brain activity.

A new mouse study in Nature has now brought clarity to this patchwork of findings by proposing that blood flow reflects the interplay between two groups of brain cells: one that ramps up when the brain becomes aroused and one that quiets down. These two populations are distributed throughout the brain, the study found, and together their activity can predict blood flow patterns across a wide range of brain regions and states.

“When an fMRI scan shows increased blood flow, we can now be sure it’s due to neurons firing,” says study co-author Matteo Carandini of University College London. “But we have a new ambiguity: which of the two types of neurons are firing.”

In other words, working backward from blood flow to neural activity can be a challenge, since different patterns of neural activity can produce similar blood-flow signals.

The research was carried out in the laboratory of Carandini and fellow University College London neuroscientist Kenneth Harris, both investigators with the Simons Collaboration on the Global Brain and members of the International Brain Lab (IBL). This study builds upon the IBL’s work to create a brainwide map of neural activity by relating it to blood flow.

Divergent Neural Dynamics

Previous studies of the connection between blood flow and neural activity tended to evaluate a single average measure of the rate of neuron firing in a brain region. This kind of bulk measure is incapable of distinguishing between neurons that have differing relationships to blood supply.

“If you look at the average activity, things can cancel out,” says Agnès Landemard, a postdoctoral fellow at University College London and the study’s first author. “You’re missing the richer signal.”

To gain a more nuanced view, Landemard and her colleagues used a combination of ultrasound imaging and thin silicon probes to record high-resolution data for both blood flow and the activity of thousands of neurons. As mice spontaneously passed in and out of periods of arousal — detected by the motion of their whiskers — the researchers could track the ebb and flow of blood and the dynamics of neuronal firing.

When a mouse became aroused, some neurons became more active and others less active. As the mouse transitioned back to a more restful state, the activity patterns of these two populations largely reversed. The researchers named the populations “Arousal+” and “Arousal−” neurons.

When the researchers tried to model blood flow dynamics using the bulk firing rate of all neurons together, the resulting prediction was poor, as expected. When they instead created a model with separate blood flow relationships for the two populations, their prediction improved markedly. It remained accurate as the team examined different brain regions and mental states. The variations that previous studies had found across brain regions are explained, the researchers propose, by the fact that different regions have different proportions of Arousal+ and Arousal− neurons.

“This shows that the relationship between neural activity and blood volume is more complicated than people thought, but it’s fixed across brain regions and states,” Carandini says. “We can go from neural activity to blood volume in a rock-solid way.”

A New Challenge for fMRI

What’s harder is to go in the other direction, from blood volume to neural activity. Neural activity, the new study shows, is inherently two-dimensional, incorporating separate information from Arousal+ and Arousal− neurons. It’s clear how to collapse this two-dimensional signal into a one-dimensional blood volume signal. But there’s no simple way to go in the other direction unless you have additional information, such as the relative proportion of Arousal+ and Arousal− neurons in the brain region you are studying.

Carandini likens the situation to hearing noise from a distant sports arena and wondering what is happening in the game. If you hear a roar from the crowd, maybe the home fans are cheering and the away fans are groaning — or maybe the reverse is true. “You know that something happened, but you can’t say what,” Carandini says.

That’s bad news for neuroscientists who rely on fMRI to gain insight into neural activity. “The fMRI researchers would like to just look at blood volume and infer what neurons are doing,” Carandini says. “We’re saying you can’t do that.”

The new work has given rise to a host of questions about the two neuronal populations. Apart from their involvement in arousal, “we don’t know what they do and what they are,” Landemard says. She plans to examine how neuromodulators such as serotonin influence the activity of Arousal+ and Arousal− neurons, in the hope of resolving some of these mysteries.

Perhaps the biggest mystery is why these two populations appear in every part of the brain, including regions such as the visual and auditory cortices, which are not traditionally associated with arousal. “It’s really bizarre that all regions have neurons that go up with arousal or down with arousal,” Carandini says. “We don’t know why, but we’re looking forward to learning more.”