Mammalian brains are known to be comprised of densely interconnected neurons, but a remaining mystery in neuroscience is how tools which capture relatively few components of brain activity have enabled scientists to predict behavior in mice. To better capture and understand neural activity in mice, professor Alipasha Vaziri and his team at The Rockefeller University used large-scale recordings and light-beads microscopy (LBM), a volumetric, two-photon imaging technique developed by the Vaziri lab in 2021. LBM increases imaging speed by eliminating the “dead-time” between sequential laser pulses, when no neuroactivity is recorded, and by removing the need for scanning. LBM breaks one laser pulse into 30 sub-pulses that are aimed at the mouse brain. Each sub-pulse reaches a different depth in the brain. The same amount of fluorescence is induced at each depth through a cavity of mirrors that staggers the firing of each pulse in time and ensures that each pulse reaches its target depth via a single microscope focusing lens. With LBM, the only limit to the rate at which samples can be recorded is the time it takes the fluorescent tags to flare. Consequently, broad swaths of the brain can be recorded within the same amount of time it takes a conventional two-photon microscope to capture just a smattering of brain cells. Light-beads microscopy (LBM) developed by the Vaziri lab at The Rockefeller University has enabled a 100-fold increase in the number of neurons that can be simultaneously recorded. Courtesy of Rockefeller University. The researchers integrated LBM into an optical microscopy platform and imaged the activity of one million neurons at cellular resolution across the mouse cortex. The team used multiple cameras to observe the mice from different angles while they engaged in spontaneous behaviors. After demonstrating the efficacy of LBM, Vaziri’s team was ready to use it to address fundamental questions in neurobiology. “We had a tool that could allow us to make discoveries that other technologies could not,” Vaziri said. “So, we tried to ask questions that only such a tool could answer. To wit: how much more information are we extracting as we keep recording from more and more neurons, and what does that information represent?” Combining LBM with advanced data analysis, computational modeling, and machine learning techniques, the researchers studied the neural activity of mice as the mice spontaneously moved and reacted to their environment. When the researchers imaged neural activity at cellular resolution simultaneously across the mouse cortex, they found a correlation between the unbounded scaling of dimensionality and the number of neurons in populations of up to one million neurons. They found that while about half the neural variance could be explained by 16 behavior-related dimensions, the remaining dimensions lacked immediate behavioral correlates. These higher dimensions represented fine-grained activity patterns of neural activity that spanned the entire mouse cortex. More than 90% of the dimensions were not connected to any spontaneous movements or sensory inputs in the mice. Thousands of these dimensions, containing more than half the cumulative neural activity of the mice, were spread across the brain in space and time without forming distinct clusters in any specific region of the brain, and ranged in time from minutes to less than seconds. The reason for this continuous neural activity in the mice is a mystery. “We still don’t know, but it’s definitely a signal that is distinct from noise,” Vaziri said. “It could offer a window into a variety of complex internal states or neurocomputation.” In addition to displaying continuous activity, the neurons were firing everywhere across the cortex. The researchers do not know the reason for the firing, but Vaziri believes that these neurons could underlie a brain-wide network of correlated neural fluctuations, possibly related to internal-state dynamics such as hunger or motivation. Previous research has shown that neural activity linked to animal movements is streamlined into a low-dimensional subspace. However, the inability to record from a multitude of neurons has influenced previous assumptions about the true dimensionality of brain dynamics, Vaziri said. Large-scale, cellular-resolution recording is needed to uncover the full neural substrates of neuronal computations. “It was only thanks to LBM’s capability that we could discover that more than 90 percent of the remaining dimensions contained reliable signals that were distinct from noise, not required for behavior, and not explained by environmental stimuli.” Vaziri said. Vaziri and his team are helping researchers at Stanford University, University College London, and other institutions to replicate LBM technology in their own neuroscience labs. The researchers are making the data from the current study available to the research community for analysis. The team also hopes to make LBM applicable for other types of research. “For example, we’d like to welcome research groups that work with different model systems beyond mice — insects, nonhuman primates and so on — so, we need to have versions of LBM that are more versatile, robust, and user-friendly,” Vaziri said. The research was published in Neuron (www.doi.org/10.1016/j.neuron.2024.02.011).