Star-Shaped Brain Cells Found to Replay Time Itself During Memory Formation

For decades, the story of memory has been told almost entirely through neurons. The hippocampus, the brain’s seahorse-shaped memory hub, was known to contain ‘time cells’—pyramidal neurons that fire in ordered sequences to stamp the passage of seconds and minutes onto our experiences. Now, a team at Boston University has revealed that astrocytes, the star-shaped glial cells long dismissed as mere support staff, produce their own time-compressed activity sequences during learning—and replay them when a memory is recalled. The finding, published in Nature Neuroscience, fundamentally expands the cast of characters involved in how the brain encodes the temporal structure of experience.

The research team, led by co-first authors Ryan A. Senne and Rebecca L. Suthard in the laboratory of Steve Ramirez, used one-photon calcium imaging to watch individual astrocytes in the dorsal CA1 region of the hippocampus in freely moving mice across multiple days. The animals underwent contextual fear conditioning, a standard paradigm in which a mouse explores a chamber and receives mild foot shocks, learning to associate the environment with threat. Because astrocytes are not electrically excitable in the way neurons are, the researchers relied on GCaMP6f, a genetically encoded fluorescent calcium indicator delivered via a virus with a promoter that targets astrocytes almost exclusively. Immunohistochemical validation confirmed the specificity was extraordinary: of thousands of fluorescent cells examined, 99.83 percent were astrocytes, with only a vanishingly small fraction of neurons labeled.

What the team observed when the foot shock was delivered was striking. Rather than a uniform surge of calcium across the astrocyte population, individual astrocytes activated in a reliable order, forming sequences that unfolded over seconds. Critically, the timing of these sequences was not linear. When the researchers modeled the distribution of peak activation times, the best-fitting model was a power law, indicating logarithmically compressed time—early moments in the sequence were represented with fine resolution while later moments were progressively squeezed together. This mirrors a property previously documented in hippocampal neurons, where internally generated time is known to be logarithmically scaled, and it suggests that astrocytes and neurons may share a common temporal grammar.

The sequences did not simply reflect the physical shock itself. When mice were later returned to the same chamber—now shock-free—the same ordered astrocytic sequences re-emerged, locked to spontaneously occurring high-activity events in the population. In a distinctly different chamber that the mice had never associated with danger, these sequences were absent or far weaker and less consistent. This context specificity is the hallmark of memory recall: the astrocytes were not merely reacting to sensory stimulation but reinstating a learned temporal pattern tied to a specific remembered place.

To detect these sequences rigorously, the team developed a nonparametric event-detection method. Calcium activity from all tracked astrocytes was summed across the population, and time windows in which activity exceeded one standard deviation above the session mean for at least two seconds were flagged as candidate sequences. This approach, validated against the unambiguous shock-evoked responses, allowed the researchers to identify spontaneous sequence events during recall sessions without imposing assumptions about their waveform or timing. Bayesian model comparison using the widely applicable information criterion then determined which statistical distribution best described the peak timing structure within each detected event.

Perhaps the most provocative finding concerns the relationship between astrocytic activity and behavior. Within the conditioned context, spontaneous astrocytic calcium events were associated with transitions between behavioral states, most notably the onset of freezing—the stereotyped fear response in which mice hold perfectly still. This coupling between astrocytic dynamics and behavioral state shifts was absent in the novel context. The observation resonates with a growing body of work showing that astrocytes are sensitive to neuromodulators such as norepinephrine, which surges during salient or threatening events and reshapes brain state. Astrocytes may thus serve as a slow, integrative layer that tracks and potentially regulates the global state of hippocampal circuits as animals move between vigilance, movement, and fear.

The technical achievement underlying these observations should not be understated. Astrocyte calcium signals are slower and morphologically messier than neuronal spikes, and tracking the same hundreds of cells across days of imaging requires sophisticated registration and segmentation pipelines. The team employed machine-learning tools for cellular segmentation, markerless pose estimation to quantify behavior, and clustering methods to classify the diverse kinetic shapes of astrocytic calcium waveforms. Their analysis revealed at least six distinct response clusters varying in peak latency and decay, hinting at functional heterogeneity within the astrocyte population that parallels the diversity long recognized among hippocampal neurons.

Notably, the researchers also controlled for the possibility that the sequences were simply an artifact of movement. Decoding analyses attempting to reconstruct the animals’ running velocity from astrocytic population activity largely failed, and only one animal showed any significant coupling between calcium dynamics and locomotion. The temporal structure, in other words, was not a disguised motor signal but something genuinely tied to the learned temporal and contextual features of the memory itself.

The study builds on a rapidly accelerating reassessment of glial function. Recent work has shown that hippocampal astrocytes encode reward location, that cortical astrocytes perform network-level encoding of local neurotransmitters, and that learning-associated astrocyte ensembles regulate memory recall. The new findings add a crucial dimension: time. If astrocytes carry compressed temporal sequences that are specific to a remembered context and linked to behavioral state, they may participate directly in constructing the ‘when’ component of episodic memory, not merely supporting the ‘what’ and ‘where’ handled by neurons.

The implications stretch toward disease and therapy alike. Astrocytes are implicated in Alzheimer’s disease, traumatic brain injury, and numerous psychiatric conditions, and memory disruption is a shared feature of many of them. If astrocytic temporal sequences are a genuine component of memory traces, then therapies targeting glial calcium signaling could one day modulate how memories are formed, strengthened, or dampened—a possibility of particular interest for disorders of maladaptive fear such as post-traumatic stress disorder. For now, the message from Boston University is clear: every time the brain replays a memory, it appears to do so with a full cellular orchestra, and the astrocytes have finally been handed their score.

Subject of Research: Astrocytic calcium activity sequences in the hippocampus during learning and memory recall

Article Title: Hippocampal astrocytic sequences emerge during learning and memory recall

Article References: Senne, R. A., Suthard, R. L., Cao, R., Coello, S., Monasterio, A. H., Ruesch, E. A., Buzharsky, M. D., Howard, M. W., & Ramirez, S. (2026). Hippocampal astrocytic sequences emerge during learning and memory recall. Nature Neuroscience. https://doi.org/10.1038/s41593-026-02448-0

Image Credits: AI Generated

DOI: 10.1038/s41593-026-02448-0

Keywords: astrocytes, hippocampus, memory, calcium imaging, time cells, fear conditioning, neuroscience, glial cells, memory recall, temporal sequences, CA1, Boston University