Nature News · Feb 11, 2026 · Collected from RSS
Article Open access Published: 11 February 2026 Olena Bukalo ORCID: orcid.org/0000-0001-7344-36011, Ruairi O’Sullivan1, Yuta Tanisumi ORCID: orcid.org/0000-0002-2178-898X2, Adriana Mendez ORCID: orcid.org/0000-0003-0418-96081, Chase Weinholtz1, Sydney Zimmerman1, Victoria Offenberg1, Olivia Carpenter1, Hrishikesh Bhagwat ORCID: orcid.org/0009-0005-8421-77531, Sophie Mosley1, John J. O’Malley3, Kerri Lyons4, Yulan Fang5,6,7, Jess Goldschlager1, Linnaea E. Ostroff5,6,7, Mario A. Penzo ORCID: orcid.org/0000-0002-5368-18023, Hiroaki Wake2, Lindsay R. Halladay ORCID: orcid.org/0000-0003-2232-67094,8 & …Andrew Holmes ORCID: orcid.org/0000-0001-7308-11291 Nature (2026)Cite this article 3358 Accesses 49 Altmetric Metrics details AbstractBrain systems mediating responses to previously encountered threats provide critical survival functions. Fear memory and extinction are underpinned by neural representations in the basolateral amygdala (BLA)1,2,3,4,5,6,7, but the contribution of non-neuronal cells, including astrocytes, to these processes remains unresolved. Here, using in vivo calcium (Ca2+) imaging and causal astrocyte manipulations, we find that BLA astrocytes dynamically track fear state and support fear memory retrieval and extinction. By combining astrocyte manipulations with in vivo BLA neuronal Ca2+ imaging and electrophysiological recordings, we show that astrocyte Ca2+ signalling enables neuronal encoding of fear memory retrieval and extinction, and readout through a BLA–prefrontal circuit. Our findings reveal a key role for astrocytes in the generation and adaptation of fear-state-related neural representations, revising neurocentric models of critical amygdala-mediated adaptive functions. Similar content being viewed by others MainEnvironmental stimuli experienced during threatening events become associated with danger. On subsequent encounters with these stimuli, a fear memory is retrieved through engagement of dedicated brain circuitry that generates internal fear states and mobilizes defensive behaviours1,2. In regions including the BLA, stimulus-elicited fear states are neurally represented and readout by anatomically defined neuronal subpopulations to downstream effector sites3,4,5,6,7,8. However, it is unclear whether non-neuronal BLA cell types enable memory-related representations and the dynamic representational adaptations that occur as memories undergo extinction.Astrocytes are a highly abundant type of glial cell that make extensive contacts with neurons9. Although relatively quiescent under basal conditions, astrocytes exhibit complex intracellular Ca2+ dynamics, respond to neurotransmitters and modulators, and regulate transmission through the release of Ca2+-evoked gliotransmitters10,11,12,13,14,15,16,17. There is growing evidence that astrocytes interact with neurons through dynamic fluctuations in Ca2+ activity to encode sensory information, gate state transitions and mediate experience-dependent behavioural adaptations, including those related to fear18,19,20,21,22,23,24,25,26,27,28,29,30,31,32,33,34,35,36. These findings implicate astrocytes in fear, but there remains a critical gap in our understanding of how astrocytes might support fear memory-related neuronal coding and plasticity through modulation of BLA neuronal representations and behaviourally mediating circuit outputs.Astrocytes track memory retrieval and extinctionWe first monitored BLA astrocyte activity as mice underwent a cued fear memory paradigm by virally expressing a Ca2+ indicator (cyto-GCaMP6f) in the astrocyte cytosol and performing in vivo fibre photometry (Fig. 1a). Immunostaining for astrocytic and neuronal markers confirmed selective GfaABC1D-cyto-GCaMP6f expression in astrocytes (around 2% neuronal expression) (Fig. 1b; details of all statistical tests and results are provided in Supplementary Table 1). Fear conditioning (F-Con) involved three co-presentations of a tone-conditioned stimulus (CS) and footshock unconditioned stimulus (US) and was followed by 2 days of repeated (25×) CS-alone extinction sessions in a novel context (Ext1, Ext2) (Fig. 1c). CS-elicited freezing was high during early extinction (E-Ext, first extinction trial block), that is, during fear memory retrieval, then decreased significantly by late extinction (L-Ext, final extinction trial block) and retrieval (E-Ret), before returning to the pre-extinction levels on retesting in the conditioning context (fear renewal, F-Ren) (Fig. 1d).Fig. 1: Astrocytes track memory retrieval and extinction.a, In vivo Ca2+ fibre photometry. b, Example fibre placement and GCaMP expression in astrocytes (GFAP) and neurons (NeuN) (two-tailed paired t-test, *P < 0.0001). n = 16 sections, 6 mice. Scale bars, 200 μm (left) and 10 μm (right). c, Experimental timeline. Cxt, context. d, Freezing across testing phases (one-way repeated measures analysis of variance (ANOVA) with Å Ãdák’s post hoc test, *P < 0.0001, comparing E-Ext (Ext1 block 1) versus L-Ext (Ext2 block 5), and E-Ext versus E-Ret). n = 17 mice. e, Population trace of CS and shock-related Ca2+ activity during F-Con with three trial-wise heat maps shown below. n = 17 mice. f, AUC analysis of shock-related Ca2+ activity during F-Con (two-tailed paired t-test versus CS, *P = 0.0080). n = 17 mice. a.u., arbitrary units. g, Population trace of CS-related Ca2+ activity during E-Ext, L-Ext, E-Ret and F-Ren; five trial-wise heat maps are shown below. n = 14–17 mice per stage. h, CS-related Ca2+ activity across test phases (one-way repeated measures ANOVA and two-tailed unpaired t-tests, comparing E-Ext versus L-Ext (*P = 0.0231), E-Ext versus E-Ret (*P = 0.0262), L-Ext versus F-Ren (*P < 0.0001), E-Ret versus F-Ren (*P = 0.0002)). n = 14–17 mice per stage. i, The relationship between CS-related Ca2+ activity and freezing across test sessions (two-tailed Pearson’s correlation, *P = 0.0261). n = 14–17 mice per stage. j, Simultaneous in vivo Ca2+ fibre photometry recordings of astrocytes (cyto-GCaMP6f) and neurons (hSyn-GCaMP6f) in opposite hemispheres of the same mice. k, The latency to peak CS-related Ca2+ activity in neurons and astrocytes during E-Ext (two-tailed unpaired t-test, *P = 0.0021). n = 5 mice per group. l, In vivo two-photon (2P) Ca2+ imaging. m, Example GRIN lens placement (left) and GCaMP expression in astrocytes (S100β) (right). Scale bars, 200 µm (left) and 20 μm (right). n, Example raw and AQuA38 processed images (left). Scale bars, 50 µm. Right, example astrocyte events and associated spontaneous Ca2+ activity. o, The number of Ca2+ events per CS across test phases (one-way repeated-measures ANOVA, Pre-Con versus E-Ext (two-tailed unpaired t-test, *P = 0.0154), E-Ext versus L-Ext (two-tailed unpaired t-test, *P = 0.0376)). n = 257 events, 6 mice. p, Decoding CS presentation from astrocyte Ca2+ activity (one-tailed paired t-test, *P = 0.0265). n = 110 events, 6 mice. q, Example CS-related Ca2+ traces illustrating changes induced by fear and extinction learning and quantification of different subpopulation of Ca2+ transients changed by extinction. n = 80–94 events, 6 mice. Fluorescence values were z-scored to the 5-s pre-CS period (photometry) or the entire recording session (two-photon imaging). The horizontal lines above the traces in e, g and k denote the permutation-test-determined significant difference from chance for astrocytes (dark green), neurons (light green), or between astrocytes and neurons (orange). Data are mean ± s.e.m.Source dataFull size imagePhotometry revealed robust Ca2+ responses to US, and not CS, presentation during F-Con, and then to CS presentation during E-Ext (Fig. 1e–g). Paralleling changes in freezing across test phases, CS-related astrocyte Ca2+ activity (area under the curve (AUC) and transient frequency) decreased with extinction (that is, L-Ext, E-Ret) and recovered on F-Ren (Fig. 1g,h, Extended Data Fig. 1a–d and Supplementary Video 1). Test-stage-wise changes in astrocyte Ca2+ activity reflected overall freezing levels during CS presentation (Fig. 1i), rather than episodes of inactivity per se—given astrocyte Ca2+ increased at both movement onset and offset23,27,37,38,39 during E-Ext (Extended Data Fig. 1g). Thus, these data show that amygdala astrocyte Ca2+ activity tracked learned changes in the CS-related fear state.Comparable photometry data were obtained by measuring Ca2+ in astrocyte processes using the membrane-targeted Ca2+ indicator lck-GCaMP6f (Extended Data Fig. 1h–l), but there was minimal (cyto-GCaMP6f) CS-related activity during E-Ext when either the US or CS was omitted during F-Con, and no trial-wise decrease in Ca2+ activity in non-extinguished mice (Extended Data Fig. 2a–k). Lastly, simultaneous photometry performed in BLA astrocytes and neurons in different hemispheres of the same conditioned and extinguished animals indicated generally similar test-stage-related changes in Ca2+ activity, but with comparatively slower-to-peak and longer-lasting US- and CS-related Ca2+ responses in astrocytes than in neurons (Fig. 1j,k and Extended Data Fig. 3a–n).To gain insight into memory-related astrocyte Ca2+ dynamics at the level of individual astrocytic events, we next performed in vivo multiphoton Ca2+ imaging of BLA astrocytes of Glt1-G-CaMP7 mice40 through a chronically implanted gradient index (GRIN) lens (Fig. 1l–q and Extended Data Fig. 4a–c). Imaging data that were processed to isolate discrete activity-related astrocytic Ca2+ events38 in head-fixed animals undergoing fear and extinction testing indicated an overall pattern of changes in Ca2+ activity (AUC and transient frequency) across test stages that resembled our photometry data, including elevated CS-related activity during E-Ext (relative to pre-conditioning, Pre-Con) that diminished during L-Ext and E-Ret (Fig. 1o, Extended Data Fig. 4d–h and Supplementary Video 2). Moreover, linear classifiers trained on Ca2+ activity from individual astrocyte events during E-Ext could successfully deco