Amygdala astrocyte primary cilium mechanisms contribute to stress behaviours

Unravelling the mechanisms that link adverse life events to stress-related behavioural changes and disorders such as depression and anxiety represent an unmet goal of neuroscience. Important advancements have identified several brain regions associated with stress-related disturbances of emotion, reward and cognitive function, including the basolateral amygdala (BLA)^1,10,11,12. The BLA has a crucial role in associative learning, particularly for fear and reward, by encoding emotional valence of stimuli and by influencing motivated behaviours. The BLA is activated by stress and is altered in conditions of pathological stress^1,10,11,12. However, the molecular and cellular basis of stress-related changes in the amygdala are not well understood. Moreover, because many past studies have focused on neurons, the contributions of non-neuronal cells such as astrocytes to stress-related behaviours remain incompletely explored¹³.

Astrocytes, a type of glial cell, are crucial components of neural circuits and have essential roles in physiology and disease¹⁴. Recent studies have shown that astrocytes are altered in psychiatric disorders and in animal models of psychiatric disorders^{15,16,17,18,19,20,21,22,23,24,25}. These findings underscore the importance of exploring whether astrocytes contribute to stress-related pathophysiological processes in the central nervous system (CNS). Amygdala astrocytes are altered in rodent models of stress²⁶, and both astrocytes and neurons contribute to amygdala function^{25,27,28,29,30,31}. Stress-related behavioural disorders represent a major unmet clinical need³², but the possibility that amygdala astrocytes can contribute to and be used for new therapeutic strategies remains largely unexplored.

The primary cilium is an antenna-like membrane protrusion adorning cells that integrates extracellular cues and serves essential roles in signalling^2,8,9,33. Relatively little is known about astrocyte primary cilia^5,34,35; however, their deletion causes molecular and developmental changes⁵. In this study, we use various mouse models of stress and multiple approaches to elucidate how amygdala astrocytes change molecularly and cellularly. We then identify astrocyte-related signalling mechanisms that may enhance positive valence during aberrant stress-related behaviours. Unexpectedly, our evaluations reveal astrocyte primary cilium mechanisms associated with stress-related behavioural disturbances. Moreover, S1PR1 GPCRs, which are shared between humans and mice, are potential targets to restore primary cilia and astrocytic contributions. Complementary evidence from mice and humans indicate that amygdala astrocyte primary cilia are disrupted during stress and that their restoration is associated with stress-related improvements in mice. We therefore identify previously unknown mechanisms of amygdala astrocytes that can potentially be used to treat stress-related disorders.

We studied amygdala astrocytes in control unstressed (US) mice and in mice subjected to chronic restraint stress (CRS), water avoidance stress (WAS) or spared nerve injury (SNI), which model chronic, psychological and pain-mediated stress, respectively^{36,37,38,39,40,41,42} (Fig. 1a). After stress induction, we performed stress-related physiological and behavioural evaluations alongside comprehensive molecular and cellular profiling of astrocytes. In contrast to previous neuron-centric studies, our experimental design deliberately prioritized the identification of astrocyte mechanisms. We note that synaptic and neuronal circuit mechanisms of stress will require dedicated circuit-mapping approaches for full definition. However, the astrocyte-focused findings we report here provide molecular entry points to meaningfully assess the contributions of astrocytes to these circuits. Accordingly, on the basis of our findings, we formulated and tested hypotheses centred on astrocytes and their primary cilia (Fig. 1a).

We used CRS, WAS and SNI as stressors to evaluate subsequent physiological and behavioural changes in comparison to control US mice (Fig. 1b,c). For CRS and WAS, the mice underwent 6 h of restraint or 2 h of water avoidance on a small platform, respectively, for 14 days. The mice then performed open-field tests (OFTs), forced-swim tests (FSTs) and sucrose-preference tests (SPTs), which capture behaviours related to anxiety, depression and anhedonia, respectively (Fig. 1b). After the behavioural tests, brains were collected for analyses. For models of SNI stress, the common peroneal and tibial sciatic nerves were ligated and transected, whereas the sural nerve was spared (Fig. 1c). Mice with SNI displayed significantly increased sensitivity in the Von Frey test, which indicated hyperalgesia in the innervated paw (Extended Data Fig. 1a,b; n = 9–10 mice per group). However, this did not progress to the development of stress-related physiological changes or behaviours (Extended Data Fig. 1a,b; n = 9–10 mice per group).

Marked phenotypes were observed following CRS and WAS (Fig. 1d and Supplementary Table 1), as evidenced by decreased body weight, decreased thymus weight, increased adrenal gland weight and elevated corticosterone levels (Fig. 1e; n = 10–20 mice per group). These results reflect established changes^{43,44,45,46,47} caused by successful stress induction in the mice. Furthermore, CRS mice displayed significant changes in OFTs, and both CRS and WAS mice displayed significant behavioural alterations in SPTs and FSTs, which indicated stress-related behaviours related to anxiety, depression and anhedonia, respectively (Fig. 1d,f; n = 10–20 mice per group). Although CRS and WAS mice exhibited similar physiological readouts of stress, at the behavioural level, CRS mice were most robust (Fig. 1d–f). To broadly assess whether CRS and WAS lead to alterations in astrocytes and neurons in the amygdala, we performed immunohistochemistry (IHC) to evaluate protein expression levels of the immediate early gene Fos. FOS expression was increased in amygdala neurons (Fig. 1g,i) but not in astrocytes, as 94 ± 2, 99 ± 0.4 and 97 ± 1% of the FOS-positive cells were also NeuN-positive in US, CRS and WAS groups, respectively (n = 10, 8 and 9 mice, respectively, per group). GFAP levels generally decrease in amygdala astrocytes in models of early life stress²⁸, depression and in postmortem human tissue from patients with major depressive disorder (MDD)⁴⁸. Accordingly, the number of GFAP-positive astrocytes after CRS or WAS was decreased (Fig. 1h,k), with no changes in the number of SOX9-positive astrocytes²⁶ (Fig. 1j–l). GFAP levels increase in other disease settings⁴⁹.

Stress-related behaviours such as depression are often treated with antidepressants⁵⁰. Therefore, we used two such drugs to determine which of the above-reported metrics are altered^50,51,52: 30 mg kg^–1 ketamine and 10 mg kg^–1 fluoxetine (n = 10 mice per group, and n = 9–10 vehicle controls). We used the CRS model because it produced the most reliable behavioural changes (Fig. 1). Although both fluoxetine and ketamine improved some CRS-induced behaviours, ketamine was more effective (Extended Data Fig. 1c–g). However, the antidepressants were incompletely effective; complete rescue of the stress-related behavioural changes was not expected and did not occur.

We performed RNA sequencing (RNA-seq) and proteomics of amygdala samples to explore molecular mechanisms that may contribute to stress-related behavioural changes (Fig. 1). Astrocyte-specific RNA-seq of CRS and WAS mice (relative to US mice) revealed 24 downregulated and 29 upregulated differentially expressed genes (DEGs) in CRS mice, and 86 downregulated and 41 upregulated DEGs in WAS mice (absolute log₂[stress/US] > 1 and adjusted P < 0.05; Fig. 2a, Extended Data Fig. 2a–c and Supplementary Table 2). Astrocyte-enriched DEGs were mostly downregulated and, unexpectedly, many downregulated DEGs were related to primary cilia (for example, Nme8, Myoc, Cfap53i and Iqca; Fig. 2b,c; n = 4 mice per sample, 3–4 samples per group). The primary cilium is a GPCR-laden antenna-like membrane protrusion that adorns cells, including astrocytes^5,35, and integrates extracellular signals. Pathway analyses of downregulated DEGs in astrocytes (adjusted P < 0.05) revealed enrichment of pathways related to cilium assembly, movement and organization for both CRS and WAS mice (Fig. 2c and Extended Data Fig. 2d). This result underscores recent findings on astrocyte primary cilium biology^5,35. Heatmaps for cilium-related DEGs in CRS and WAS mice are shown in Extended Data Fig. 2e.

We next performed near-plasma membrane proteomics for amygdala astrocytes using LCK–BioID2 (Fig. 2d and Extended Data Fig. 3a–e; n = 8 mice per sample, 4 samples per group). We identified about 1,500 proteins in relation to endogenously biotinylated proteins with parallel controls for US, CRS and WAS groups (log₂[BioID2/GFP] > 1 and adjusted P < 0.05; Fig. 2e, Extended Data Fig. 3c–e and Supplementary Table 3). These proteins represent those unique to the experimental groups and those shared between them, which can be differentially expressed proteins (DEPs) (Fig. 2e–g). In CRS mice, 620 proteins were unique, whereas 315 proteins were unique to WAS mice. Among the shared proteins, 18 were DEPs in CRS mice compared with US mice, and 271 were DEPs for WAS mice compared with US mice (Fig. 2f; absolute log₂[stress/US] > 0.5 and adjusted P < 0.1). Consistent with the RNA-seq findings (Fig. 2a–c), analyses of DEPs revealed cilium-related proteins unique to each group or differentially expressed compared with the US group (Extended Data Fig. 3f,g). In contrast to the cilium-related genes and proteins, changes in corticosteroid Nr3c1 and Nr3c2 receptors, which are often implicated in stress responses, were not observed (Extended Data Fig. 3h). Downregulation of cilium-related protein SEPTIN2 (encoded by Septin2; also known as SEPT2) in CRS and WAS mice was confirmed by IHC (Extended Data Fig. 3i).

IHC analyses revealed about 80% of amygdala astrocytes possessed ARL13B-positive cilia in US, CRS and WAS groups (Extended Data Fig. 4g; n = 4 mice per group). However, the lengths and volumes of astrocyte primary cilia were significantly reduced after CRS or WAS (Fig. 2p–s and Extended Data Fig. 4h; n = 4 mice per group). By contrast, ARL13B-positive cilia of neurons did not change in length (Fig. 2t; n = 4 mice per group). Furthermore, in a separate cohort of mice, CRS and WAS again resulted in shorter primary cilia on amygdala astrocytes, whereas primary cilia of cortical astrocytes were unaffected (Extended Data Fig. 5a–d; n = 4 mice per group with rabbit or rat ARL13B antibody IHC). Neuronal primary cilia in the amygdala or cortex were also unaffected (Extended Data Fig. 6a,b; n = 4 mice per group). As amygdala astrocyte primary cilia were reduced in length after stress, we determined whether astrocyte morphology⁵⁸ was also altered after CRS; however, this was not so (Fig. 2u and Extended Data Fig. 4i–l; n = 4 mice per group). We also did not find marked changes in the expression of genes related to astrocyte territory size⁵⁸ after stress (Extended Data Fig. 4l). However, of the 770 primary cilium-related genes (Supplementary Table 4), 121 and 124 were altered in amygdala astrocyte DEGs (adjusted P < 0.05) after CRS and WAS, respectively, when assessed by RNA-seq (Fig. 2v). Our evaluations provide evidence that amygdala astrocyte primary cilia are significantly smaller following stress in mice. Moreover, these changes are accompanied by molecular alterations of primary cilium genes and proteins (Fig. 2a–t).

We also performed a specific set of experiments to assess the effects of ketamine, which reduced some of the stress-related behavioural changes (Extended Data Fig. 1f,g). The effects of ketamine were accompanied by significant restoration of amygdala astrocyte primary cilium length. Conversely, no effects were observed for neuronal primary cilia (Extended Data Fig. 7). We do not interpret this result to suggest that ketamine works via the astrocyte primary cilium, but instead that the astrocyte primary cilium is recovered during the beneficial response to ketamine.

As we observed reduced astrocyte primary cilium lengths, we next considered signalling pathways that are associated with this organelle. Primary cilia are well-established hubs for GPCR-mediated signalling⁸, and GPCR pathways have key roles in astrocyte function⁷, including in the amygdala³¹. On the basis of this overlap, we analysed our proteomic datasets (Fig. 2d–f) from US, CRS and WAS groups to identify GPCRs expressed in amygdala astrocytes and to determine their predicted G protein coupling^17,60 (Fig. 3a–c). On the basis of these data, we speculated that activation of GPCR signalling pathways might influence primary cilium morphology. Among the GPCRs detected in astrocytes by proteomics, most were predicted to couple to G_i and/or G_q proteins^61,62,63,64 (Fig. 3a–c; for example, G_i-coupled receptors S1PR1, FZD2, GRM2 and GRM3; G_q-coupled receptors GRM5, GRM1 and SSTR2, CHRM3). These data led us to test whether astrocyte-specific activation of GPCR signalling pathways could influence stress-induced behaviours.

We expressed hM4Di (G_i-coupled) and hM3Dq (G_q-coupled) designer receptors exclusively activated by designer drugs (DREADDs)⁶⁵ in amygdala astrocytes via AAV microinjections (Fig. 3d; n = 7–10 mice per group). Two weeks later, the mice were subjected to CRS (Fig. 3e–g) or WAS (Fig. 3e,h,i), followed by physiological and behavioural assessments (as in Fig. 1). Control mice received tdTomato in place of DREADDs, and all the mice received deschloroclozapine (10 μg kg^–1) in vivo as the DREADD ligand (Fig. 3e). Neither hM4Di nor hM3Dq DREADD activation rescued the decrease in body weight, increase in adrenal gland weight or decrease in thymus weight induced by CRS (Fig. 3f). This result implies that induction of stress and its related physiological deficits are unaffected by amygdala astrocyte GPCR signalling. However, hM4Di activation reduced corticosterone levels, which indicated that astrocytes contribute to this stress-related outcome (Fig. 3f). We also observed significantly improved measures of behaviours of CRS mice in OFTs, SPTs and FSTs after both hM4Di and hM3Dq activation (Fig. 3g and Extended Data Fig. 10a). Similarly, activation of hM4Di and hM3Dq improved stress-related outcomes following WAS (Fig. 3h,I and Extended Data Fig. 10a). To assess hM4Di and hM3Dq GPCR activation across metrics in CRS and WAS, we calculated z scores and percentage recovery (Fig. 3j,k) and found that hM4Di activation was the most effective¹⁷. In terms of controls, hM4Di activation did not affect stress-related physiological and behavioural readouts in US mice (Extended Data Fig. 10b). Moreover, consistent with the lack of stress-related behaviours after SNI (Extended Data Fig. 1b), neither hM4Di nor hM3Dq activation affected any physiological or behavioural readouts after SNI (Extended Data Fig. 10a,c). Therefore, amygdala astrocyte-specific G_i–GPCR signalling activation significantly improves behavioural responses caused by stress.

As G_i–GPCR activation markedly improved stress-related molecular, cellular and behavioural phenotypes (Fig. 3) and because amygdala astrocytes express GPCRs (Fig. 3a–c), we mined amygdala data from US mice (Fig. 2) to identify natively expressed GPCRs^17,60. We identified multiple GPCRs (about 25) that were highly expressed and enriched in astrocytes (Fig. 4a). Of the top ten, three were predicted to be G_i–GPCRs that may produce effects akin to G_i–DREADDS (Fig. 3). We chose to explore S1PR1 (encoded by S1pr1, also known as Edg1), which was both enriched in astrocytes and exhibited the highest expression levels at the RNA and protein levels (Fig. 4a,b). Existing S1PR1 agonists and functional antagonists for this G_i–GPCR are clinically used for multiple sclerosis⁶⁶, and it is the only member of the sphingosine-1-phosphate receptor family expressed in astrocytes⁵⁸. The importance of S1PR1 for astrocytes has been noted^17,60,67. We confirmed that S1PR1 was highly and specifically expressed in astrocytes using RNA-seq and near-plasma membrane proteomics of samples from US mice (Fig. 4c). Moreover, its expression was not altered by stress at the RNA or protein level (Fig. 4c). Accordingly, S1PR1 was detected in amygdala astrocytes from a mouse single-nucleus RNA-seq (snRNA-seq) database²⁹ (Fig. 4d and Extended Data Fig. 13a,b). Using RNAscope, all S100β-positive amygdala astrocytes expressed S1pr1 transcripts (Fig. 4e; n = 3 mice). Expression was dense in astrocyte somata but was also found in end feet and in end foot-adjacent endothelial cells⁶⁸ (Extended Data Fig. 14a). S1PR1 protein was not found in astrocyte primary cilia (Extended Data Fig. 14b). In accord with the mouse data, S1PR1 was expressed in human astrocytes from snRNA-seq databases⁵⁷ (Fig. 4f and Extended Data Fig. 13c). RNAscope experiments revealed that about 50% of ALDH1L1-positive amygdala astrocytes from neurotypical human individuals expressed S1PR1 (Fig. 4g,h; n = 4 human donors). Furthermore, S1PR1 activation with the selective agonist ponesimod (3 μM) increased Ca²⁺ signals in astrocytes in brain slices from control mice and those treated with ponesimod ex vivo and in vivo (Extended Data Fig. 13d–h; n = 6 mice ex vivo, and 3 mice per group in vivo), thereby demonstrating functional expression of the receptor. On the basis of these findings, we explored whether S1PR1 activation with ponesimod (30 mg kg^–1) improves CRS-related physiological and behavioural changes and astrocyte primary cilia that are disrupted following stress (Figs. 2 and 3). As expected, the stress-related physiological changes were unaffected by ponesimod (Fig. 4i; n = 10 mice per group). However, ponesimod significantly improved performance in SPTs and FSTs across groups to levels equivalent to US + vehicle controls (Fig. 4j and Extended Data Fig. 13i; n = 10 mice per group). Furthermore, the reduction in astrocyte primary cilium length observed following CRS was recovered after ponesimod treatment (Fig. 4k,l and Extended Data Fig. 13j; n = 4 mice per group). Therefore, targeting native astrocyte S1PR1 G_i–GPCRs (Fig. 4a–g) produced similar restorative effects to those observed with astrocyte G_i–GPCR chemogenetics (Fig. 3).

Relatively little is known about amygdala astrocytes, which prompted us to use astrocyte-selective Aldh1l1^cre/ERT2;RiboTag mice to evaluate their molecular properties in relation to other CNS regions⁵⁸ (Fig. 5a, Extended Data Fig. 15a and Supplementary Table 6; n = 4 mice per sample, 4 samples per region). Multidimensional scaling showed separation between major brain areas, and the astrocyte-specific immunoprecipitation (IP) fraction contained known astrocyte markers (Extended Data Fig. 15b–d). We identified the top 50 astrocyte enriched genes shared across CNS regions, including the amygdala (for example, Kcnj10, Sox9, Gja1, Slc1a2 and Apoe), representing core features of amygdala astrocytes⁵⁸ (Extended Data Fig. 15e–h). Amygdala astrocytes also showed 2,669 astrocyte-enriched genes relative to input (log₂[IP/input] > 1, adjusted P < 0.05, IP fragments per kilobase million (FPKM) > 1) and 495 amygdala-unique DEGs relative to other CNS areas (log₂[IP/average IP of other regions] > 1, adjusted P < 0.05, IP FPKM > 1). Furthermore, 69 genes were enriched in astrocytes and unique to the amygdala (log₂[IP/input] > 1, adjusted P < 0.05, IP FPKM > 1), which enabled us to impute amygdala-enriched functions (Extended Data Fig. 15e,i). We also referenced amygdala astrocyte-enriched genes against mouse amygdala astrocyte scRNA-seq data²⁹ and found about 90% overlap (Extended Data Fig. 15j). Overall, amygdala astrocytes clustered with those from the hippocampus, striatum and cortex (Extended Data Fig. 15d,k) and had shared and amygdala-specific features. For example, the corticosteroid receptors Nr3c1 and Nr3c2 were enriched in astrocytes for most regions of the CNS without dominance in the amygdala (Extended Data Fig. 15l). However, consistent with our findings related to the primary cilium (Figs. 1–4), amygdala astrocyte expression of cilium-related genes⁵ was higher than most of the other regions (Fig. 5b,c and Extended Data Fig. 16a), which suggests that astrocyte primary cilium biology may be region-specific. This finding is consistent with the field of astrocyte diversity and the functions of astrocyte subtypes. In this regard, the molecular profiles we report here enable mechanistic study of amygdala astrocyte responses in vivo^25,31.

To address the limited knowledge of the molecular composition of astrocyte primary cilia, we profiled the astrocyte primary cilium proteome using proximity-dependent biotinylation with ARL13B–BioID2, relative to ARL13B–GFP, delivered with astrocyte-selective AAVs (Fig. 6a; n = 4 mice per group). Owing to its small size, we could not obtain sufficient protein from the amygdala. Instead, we targeted cortical astrocytes, which clustered with amygdala astrocytes in multidimensional scaling (Extended Data Fig. 15d,k), have primary cilia (Extended Data Fig. 5) and provided larger tissue yields. Streptavidin staining revealed strong biotinylation along the primary cilia of S100β-positive astrocytes that colocalized with ARL13B–BioID2 immunoreactivity, whereas little biotinylation was observed in the ARL13B–GFP group (Fig. 6b,c). Mass spectrometry analyses identified significantly higher intensity and more proteins in the ARL13B–BioID2 group than in the GFP control, and principal component analysis (PCA) clearly separated the two conditions, which enabled detection of 278 proteins as hits in the ARL13B–BioID2 samples (log₂[BioID2/GFP] > 1), adjusted P < 0.05), including known primary cilium proteins (for example, ARL13B, TOGARAM1 and TULP3; Fig. 6d–f and Supplementary Table 11). We scored these 278 proteins using CilioGenics, which revealed that 206 out of 278 (74%) were classified as high or medium probability (Fig. 6g,h). Many were also enriched in cortex astrocyte RNA-seq data (Fig. 6h, and top 15 shown in Fig. 6i). Notably, S1PR1 was absent from the 278 hits comprising this putative astrocyte primary cilium proteome (Fig. 6f,j) despite its strong detection in the astrocyte plasma membrane proteome and in RNA-seq and RNAscope analyses (Fig. 4). These findings suggest that S1PR1 exerts its beneficial effects on stress-related behaviours from the plasma membrane, where its activation is associated with increased primary cilium length (Fig. 4), rather than by localizing to the primary cilium itself. Consistent with bona fide ciliary identification, the 206 CilioGenics high and medium probability hits were enriched for biological processes related to primary cilium function (for example, cilium organization, cilium assembly, centrosome assembly and actin filament organization; Fig. 6k). In Fig. 6l, we illustrate the putative subciliary localization of selected proteins based on previous work; spatial proteomics⁶⁹ of astrocyte primary cilia will be an important goal for future work. Among the 206 CilioGenics high and medium probability proteins, 18% and 26% were altered by CRS and WAS, respectively (Fig. 6m), and of the 37 proteins significantly disrupted by CRS, approximately 50% were restored by G_i–GPCR activation (Fig. 6n). These data provide a putative astrocyte primary cilium proteome and bolster findings (Figs. 1–5) that stress disrupts astrocyte primary cilia and that G_i–GPCR activation can partially restore defects.

Elucidating the mechanisms by which stressful life events precipitate enduring behavioural changes that result in psychiatric disease remains a fundamental challenge. Astrocytes, as integral components of neural circuits, engage in developmental, structural, signalling and homeostatic interactions with neurons throughout the brain, including in the amygdala^{25,29,30,31,38}, a region implicated in the integration of valence and the orchestration of emotional and stress responses^1,10,11,12. However, astrocytic mechanisms of relevance to stress-induced behavioural changes have remained largely unexplored, although there has been progress with neuroimmune interactions and psychedelics³⁸.

In this study, we identified an important role for amygdala astrocytes in the modulation of stress-related behaviours mediated via signalling pathways associated with their primary cilia². Exposure to stress produced marked alterations in amygdala astrocytes at both the proteomic and transcriptomic levels, assessed using astrocyte-selective reagents. These changes reflected dysregulation of genes and proteins associated with the structure and function of the primary cilium. Consistent with this molecular signature, morphological analyses revealed a reduction in primary cilium length on astrocytes following stress, which indicates that ciliary dysfunction has a role in response to stress-induced changes in the amygdala.

Although a precise threshold for functionally meaningful changes in primary cilium length has not been established, the primary cilium is recognized as a morphologically dynamic organelle in which even modest alterations in length, either increases or decreases, can influence signalling and cellular function. In addition to altering the amount, spatial distribution or effective density of signalling components in the cilium, changes in ciliary length may influence cellular signalling through several other mechanisms^2,33. Shortening of the primary cilium may reduce the available membrane surface for receptor localization, which may affect signal sensitivity to extracellular ligands. It may also influence intraflagellar transport dynamics, which in turn may alter the trafficking, delivery or turnover of signalling proteins in the ciliary compartment. Changes in length could also modify the spatial organization of subciliary domains (for example, base, shaft and tip)^2,33, potentially affecting interactions among receptors and scaffolding proteins. Because the primary cilium can function as a mechanosensor, shorter cilia may also alter the mechanical properties of the organelle and its ability to transduce physical stimuli. Finally, cilium length may influence cellular states linked to ciliary dynamics, such as quiescence, which could contribute to altered downstream signalling outcomes. These possibilities may contribute to the changes reported in this study, but we are not aware of any cases in which these mechanisms have been teased apart in vivo and are not aware of genetic tools to do so systematically. Consistent with a role for primary cilia in neurobiological responses, genetic ciliopathies frequently include neurological manifestations². Together, our findings indicate that stress induces coordinated changes in cilium-associated genes, proteins and morphology in amygdala astrocytes, consistent with a role for astrocytic ciliary alterations in stress-related pathology. Further mechanistic studies are now needed to unlock the signalling and biophysics of the primary cilium.

Given the established roles of primary cilia as signalling hubs², particularly for GPCR pathways, the consequential alterations of cilial genes in disease⁷⁰ and the importance of GPCRs in astrocyte biology⁷, we explored amygdala astrocyte GPCRs and their responses after stress. Through chemogenetic manipulation or targeted modulation of endogenous S1PR1, we were able to restore primary cilium length, normalize stress-induced molecular perturbations and produce substantial improvements in stress-related behavioural phenotypes. S1PR1 expression was strongly detected in amygdala astrocytes in both mouse and human tissue, which highlights the translational potential of these findings related to S1PR1 drugs approved for use in humans⁶⁶. The high level of S1PR1 in the plasma membrane proteome of amygdala astrocytes suggests that it exerts its actions from that location rather than from the primary cilium. However, further work on the relationship between astrocytic S1PR1 and cilia is needed. Blood sphingosine-1-phosphate levels have been reported to increase when stress-related behaviours were treated⁷¹, but the relevance of such clinical data will require further study. We also identified multiple other astrocyte GPCRs as potential targets, including several pharmacologically tractable orphan receptors expressed on astrocytes (for example, GPR37L1, GPR162, GPR158 and GPR179).

Genetic disruption of primary cilia through the deletion of key genes (Arl13b and Ift88) in amygdala astrocytes resulted in astrocytic and additional effects, such as perturbing the molecular properties of neighbouring parenchymal cells, including neurons. This finding highlights the multicellular regulatory capacity of astrocytes in the amygdala microenvironment and suggests that astrocytic cilia serve as modulators of local circuit functions during stress. Primary cilia are observed in most astrocytes³⁵ and have developmental roles⁵; thus, we found that in humans, cilia-related genes were abundant in amygdala astrocytes and were altered in several psychiatric disorders. Although studies of how astrocyte cilia influence neural circuits, to our knowledge, do not yet exist, our findings and tools now enable such circuit-level dissection.

In summary, our findings establish that stress disrupts primary cilia in amygdala astrocytes and that GPCR-mediated restoration of cilium integrity can reverse several molecular and behavioural sequelae of stress. These results advance our understanding of the fundamental mechanisms of glial cells that underlie stress adaptation in vertebrates. By defining a link between astrocyte primary cilia dysfunction and stress, our work opens new avenues for the development of targeted strategies to ameliorate maladaptive behavioural responses due to stress. As stress and depression-related phenotypes are common in neurodegenerative diseases, future studies could determine whether primary cilium mechanisms contribute to neurological disease.

Mouse models

All animal experiments at the University of California Los Angeles (UCLA) were conducted in accordance with the US National Institutes of Health (NIH) Guide for the Care and Use of Laboratory Animals and were approved by the Chancellor’s Animal Research Committee at UCLA. The UCLA ARC protocol numbers for the experiments were 2009-043, 2020-002, 2017-039 and 2022-016. All mice were housed with food and water available ad libitum in a 12-h light–dark environment at temperatures of 20–22 °C with 40–60% humidity. All mice were 2–5 months old when used for experiments, were healthy with no obvious behavioural phenotype, were not involved in previous studies and were euthanized during the light cycle. Wild-type C57BL/6N mice were maintained in an in-house breeding colony or purchased from The Jackson Laboratory (JAX) or Taconic Biosciences. Approximately equal numbers of male and female mice were used and no differences were observed between sexes. RiboTag mice (B6N.129-Rpl22tm1.1Psam/J, JAX stock 011029) were acquired from JAX and crossed with Aldh1l1^cre/ERT2 BAC mice (B6N.FVB-Tg(Aldh1l1-Cre/ERT2)1Khakh/J, JAX stock 029655) fully backcrossed to C57BL/6NTac background in an in-house breeding colony. GCaMP6f mice (Ai95D (C57BL/6J), JAX stock 028865) were acquired from JAX and crossed with Aldh1l1^cre/ERT2 BAC mice (B6N.FVB-Tg(Aldh1l1-Cre/ERT2)1Khakh/J, JAX stock 029655) in an in-house breeding colony. Ponesimod behavioural experiments were performed at ONO Pharmaceutical and were approved by the Institutional Animal Care and Use Committee. Arl13b^f/f;Aldh1l1^cre/ERT2 mice were group housed (4–5 mice of the same sex per cage) in a controlled environment (12-h light–12-h dark cycle at 21 °C) with unrestricted access to water and a standard chow diet at the University of Calgary, and experiments were approved by the University of Calgary Health Sciences Animal Care Committee, Health Sciences Animal Care Committee, and all experiments were completed in accordance with the Canadian Council of Animal Care guidelines. Arl13b^f/f and Ift88^f/f mice^72,73 were a gift from J. Guo and were maintained in an in-house breeding colony at UCLA. All mice were assigned to experimental groups at random. Because of the experimental design and availability of resources, blinding was not always feasible. For behavioral experiments, we minimized subjectivity as much as possible by using automated analysis software whenever possible.

Ethics statement

Human postmortem brain tissue samples were obtained from the Harvard Brain Tissue Resource Center, a NIH NeuroBioBank repository. All procedures were approved by the Mass General Brigham Institutional Review Board and conducted in accordance with institutional guidelines, NIH regulations and the Health Insurance Portability and Accountability Act. Written informed consent for brain donation was obtained from the legal next-of-kin or legally authorized representative. All samples were de-identified before distribution.

Tamoxifen injections for inducible gene expression

To selectively express RPL22–HA in astrocytes, hemizygous Aldh1l1^cre/ERT2;RiboTag mice^74,75 were injected with 75 mg kg^–1 tamoxifen (Sigma-Aldrich, T5648) dissolved in corn oil once per day for 5 consecutive days at 6–8 weeks of age and were euthanized 2–3 weeks after tamoxifen injections for astrocyte RNA-seq experiments. Arl13b^f/f;Aldh1l1^cre/ERT2 mice were administered tamoxifen (200 μg per mouse per day) from postnatal days 7–9 and were euthanized 4 weeks later for RNA-seq experiments.

CRS

Mice were weighed and placed in restrainers (diameter of 2.5 cm × 9.5 cm, 551-BSRR, Plas-Labs) for 6 h each day (from 10:00 to 16:00) for 14 consecutive days. US mice were weighed and kept in their home cages. Light intensity during restraint stress was maintained at approximately 50 lux. All mice were returned to their home cages at the end of the stress period.

WAS

Mice were weighed and placed on a platform (diameter of 5 cm × 10 cm) attached to the bottom of a plastic tank (45-cm long × 25-cm wide × 25-cm high) for 2 h each day (from 10:00 to 12:00) for 14 consecutive days. The plastic tank was filled with water (25 °C) up to 1 cm below the top of the platform. US mice were weighed and kept in their home cages. Light intensity during WAS was maintained at approximately 50 lux. All mice were returned to their home cages at the end of the stress period.

SNI model

All surgical procedures were conducted under general anaesthesia using continuous isoflurane (induction at 5%, maintenance at 1–2% v/v). The depth of anaesthesia was monitored continuously and adjusted as necessary. After the induction of anaesthesia, mice were administered with 0.1 mg kg^–1 buprenorphine (Buprenex) subcutaneously before surgery. The surgical incision site was then cleaned 3 times with 10% povidone iodine and 70% ethanol (v/v). The skin and muscles were incised to expose the sciatic nerve. The common peroneal nerve and tibial nerve, two of the three terminal branches of the sciatic nerve, were tightly ligated with 6-0 silk sutures and then cut distally, whereas the sural nerve was left intact. The muscles and skin were then sutured with 4-0 silk sutures. In the sham surgery, only the sciatic nerve was exposed, and afterwards, the muscles and skin were sutured. After surgery, mice were allowed to recover overnight in cages placed partially on a low-voltage heating pad. Buprenorphine was administered twice a day for up to 2 days after surgery. The mice were allowed to recover for 2 weeks after SNI before performing physiological and behavioural evaluations comprising OFTs, conditioned-thermal place-aversion tests and Von Frey tests. As expected⁷⁶, SNI resulted in the development of mechanical and thermal allodynia, as well as hyperalgesia, in the area innervated by the sural nerve⁷⁶.

OFT

The open-field chamber consisted of a square arena (40 cm × 40 cm) enclosed by plastic walls (30 cm in height). The centre of the arena was defined as the area 10 cm away from the chamber walls. The light intensity in the centre of the open field was maintained at 50 lux. The locomotor activity of the mice was recorded for 15 min using a camera located above the open-field chamber. The time spent, centre entries and speed in the centre, as well as the total distance travelled, were analysed using the automated video tracking software ANY-maze (v.6.3, Stoelting).

SPT

The mice were switched to individual housing and given two water bottles to acclimate them to the bottles. On day 2, the mice were given two bottles: one containing water and the other containing 1% sucrose. On day 3, the mice were fasted and again given two bottles, one with water and the other with 1% sucrose, both of which were pre-weighed. The mice had access to these bottles for 16 h (from 18:00 to 10:00). After the test period, the weight of each bottle was measured, and sucrose preference (%) was calculated using the following formula: (1% sucrose intake/(water intake + 1% sucrose intake)) × 100. The mice were euthanized after the SPTs.

FST

Mice were placed in a cylinder (diameter of 12 cm ×30 cm in height) filled with water to a depth of 20 cm, and their swimming behaviour was recorded with a video camera for 6 min. The light intensity of the test area was maintained at 50 lux. The time the mice remained immobile was recorded.

Von Frey test

The pain threshold was measured on day 21 after SNI surgery using von Frey filaments. The following 7 different strengths of von Frey filaments were used: 0.04, 0.07, 0.16, 0.4, 0.6, 1 and 2 g. Mice were placed in individual small cages on a wire mesh floor and allowed to acclimate for up to 1 h. The von Frey filaments were applied vertically to the plantar surface of the hind paw on the SNI-treated side from underneath the wire mesh floor. A quick withdrawal response or paw flicking was considered a positive pain response, and pain response was measured using a previously described up–down method⁷⁷.

Conditioned-place temperature-aversion test

The conditioned-place temperature-aversion test was conducted using an apparatus (33 × 16.5 ×30 cm, Thermal Place Preference-2 (BIO-T2CT, Bioseb)) equipped with two plates for which temperatures can be individually controlled and enclosed by Plexiglass. On the first day, mice were placed in the centre of the apparatus with both plates set to 25 °C, and the mice were allowed to freely explore the plates for 10 min to acclimate to the apparatus. Subsequently, the test plate was set to 19 °C whereas the reference plate remained at 25 °C, and visual cues were placed on the outside of the apparatus on the test plate side. The mice were placed in the centre of the apparatus and allowed to freely explore the plates for 10 min during two conditioning trials, with an interval of approximately 3 h between trials. On the second day, both the test plate and the reference plate were set to 25 °C, with visual cues still placed on the outside of the apparatus on the test plate side. The mice were placed in the centre of the apparatus and allowed to explore the plates freely for 10 min. The locomotor activity of the mice was recorded using a camera located above the apparatus, and the time spent on each plate was automatically analysed using Bioseb software. Place preference (%) was calculated using the following formula: (time spent on test plate/(time spent on test plate + time spent on reference plate)) × 100.

Blood and tissue collection

After anaesthetizing the mice with isoflurane, and following euthanasia, an incision was made in the abdomen to collect 0.5 ml blood from the abdominal vena cava, which was then placed into a 1.5 ml tube containing heparin. Subsequently, the adrenal glands from both sides and the thymus were excised and their wet weights were measured. The collected blood was centrifuged at 1,500g for 15 min to obtain plasma for corticosterone measurement. Plasma corticosterone levels were measured using an ELISA kit (Enzo, ADI-900-097), according to the manufacturer’s instructions.

Ponesimod treatment

Ponesimod (30 mg kg^–1) or vehicle (0.25% (w/v) methylcellulose with 0.05% (v/v) Tween 80) was intraperitoneally administered once daily, 1 h before the start of CRS. Moreover, on the days of the behavioural tests, a single dose was given 1 h before the start of each test. After the tests were completed, blood and brain tissue samples were collected for plasma corticosterone measurements and IHC, respectively. US and CRS mice received identical ponesimod doses, whereas the vehicle controls received just the vehicle. For [Ca²⁺]_i imaging experiments, the mice received a single injection of ponesimod or vehicle. We comment briefly on the known off-target effects of ponesimod. The FDA drug approval package for ponesimod (application number: 213498; www.accessdata.fda.gov/drugsatfda_docs/nda/2021/213498Orig1s000TOC.html) reports known off-target binding activity. Ponesimod showed greater than 50% inhibition of binding to three targets: endothelin ETA receptor, around 74% inhibition; monoamine oxidase MAO-B, about 62% inhibition; and protein serine/threonine kinase Ca²⁺/calmodulin-dependent II (CaMKII), about 63% inhibition. However, several factors were suggested to show that these off-target effects are negligible. Principally, the approximately 1,750-fold difference between the inhibition constant of S1PR1 (5.7 nM) and the screening concentration (10 μM) suggests that at clinically relevant doses, the effects of ponesimod will be dominated by S1PR1 activation. Nonetheless, off-target effects should be considered in future studies that develop S1PR1 or the other GPCRs we report here.

Neuroinflammation induction with LPS

Neuroinflammation was induced with 5 mg kg^–1 intraperitoneal injections of LPS, exactly as previously reported⁷⁸. After 24 h, the effects of neuroinflammation were confirmed by reduced ambulation metrics in OFTs, and the mice were euthanized for tissue collection.

Stereotaxic microinjections of AAVs

All surgical procedures were conducted under general anaesthesia using continuous isoflurane (induction at 5%, maintenance at 1–2% v/v). The depth of anaesthesia was monitored continuously and adjusted as necessary. After the induction of anaesthesia, the mice were fitted into a stereotaxic frame with their heads secured by blunt ear bars and their noses placed into a veterinary-grade anaesthesia and ventilation system (VetEquip). Mice were administered with 0.1 mg kg^–1 buprenorphine (Buprenex) subcutaneously before surgery. The surgical incision site was then cleaned 3 times with 10% povidone iodine and 70% ethanol (v/v). Skin incisions were made, followed by two craniotomies of 2–3 mm in diameter using a small steel burr (Fine Science Tools) powered by a high-speed drill (K.1070, Foredom). Saline (0.9%) was applied onto the skull to reduce heating caused by drilling. Bilateral viral injections were performed using a stereotaxic apparatus (David Kopf Instruments) to guide the placement of bevelled glass pipettes (1B100-4, World Precision Instruments). For the BLA, the following coordinates were used: 1.45 mm posterior to bregma; 3.20 mm lateral to the midline; and 3.75 mm from the pial surface. AAVs were injected by using a syringe pump (Pump11 PicoPlus Elite, Harvard Apparatus). After AAV microinjections, glass pipettes were left in place for at least 10 min before being slowly withdrawn. Surgical wounds were closed with external 4-0 silk sutures. After surgery, mice were allowed to recover overnight in cages placed partially on a low-voltage heating pad. Buprenorphine was administered twice a day for up to 2 days after surgery. AAV-injected mice were used for experiments at least 2 weeks after surgery. All AAV titres were adjusted to 1.0 × 10¹³ genome copies per ml with sterile 0.1 M PBS. The following viruses were used: 0.15 μl AAV2/5 GfaABC1D-tdTomato (Addgene, 44332); 0.15 μl AAV2/5 GfaABC1D-hM4Di–mCherry (Addgene, 92286); 0.15 μl AAV2/5 GfaABC1D-hM3Dq–mCherry (Addgene, 92284); 0.15 μl AAV2/5 GfaABC1D-cyto–GCaMP6f (Addgene, 52925); 0.15 μl AAV2/5 GfaABC1D-LCK–GFP (Addgene, 105598); 0.15 μl AAV2/5 GfaABC1D-LCK–BioID2-BioID2–HA (Addgene, 176741); 0.15 μl AAV2/5 GfaABC1D-eGFP (Addgene, 176861); and 0.15 μl AAV2/5 GfaABC1D-Cre-4x6T (Addgene, 196410).

IHC

For transcardial perfusion, mice were euthanized with isoflurane and perfused with 0.1 M PBS followed by 10% buffered formalin (Fisher SF100-20). After gentle removal from the skull, the brain was post-fixed in 10% buffered formalin overnight at 4 °C. The brains were then cryoprotected in 30% sucrose with 0.1 M PBS solution for at least 48 h at 4 °C until use. Serial coronal sections (40 μm) containing the amygdala were prepared using a cryostat microtome (Leica) at −20 °C and processed for IHC. Sections were washed 3 times in 0.1 M PBS for 5 min each and then incubated in a blocking solution consisting of 10% normal goat serum in 0.1 M PBS with 0.5% Triton-X100 for 1.5 h at room temperature with agitation. Sections were then incubated with agitation in primary antibodies diluted in blocking solution overnight at 4 °C. The following primary antibodies were used: guinea pig anti-NeuN (1:1,000; Synaptic Systems, 266004); rabbit anti-FOS (1:1,000; Synaptic Systems, 226008); chicken anti-mCherry (1:1,000; Abcam, ab205402); chicken anti-GFAP (1:1,000; Abcam, ab4674); rabbit anti-S100β (1:1,000; Abcam, ab41548); guinea pig anti-S100β (1:500; Nittobo, MSFR105350); chicken anti-S100β (1:500; Synaptic Systems, 287006), rabbit anti-HA tag (1:1,000; Abcam, ab9110); rabbit anti-septin 2 (1:100; Thermo Scientific, 11397-1-AP); rabbit anti-ARL13B (1:500; Proteintech, 17711-1-AP); rat anti-ARL13B (1:500; BiCell Scientific, 90413); rabbit anti-S1PR1 (1:50; Invitrogen, PA1-1040); rabbit anti-AC3 (1:4,000; Invitrogen, PA5-35382); rabbit anti-SOX9 (1:500; Millipore, AB5535); goat anti-collagen IV (1:500; Novus Bio, NBP1-26549); and biotinylated WFA (1:1,000; Vector labs, B-1355-2). For IHC of collagen IV, the sections were pre-treated with pepsin (Sigma, R2283) for 5 min at 37 °C and washed with PBS at 27 °C for 8 min. Sections were washed 4 times in PBS-T for 5 min and then incubated at room temperature for 1 h with secondary antibodies diluted in blocking solution. The following Alexa-conjugated secondary antibodies (1:1,000; Molecular Probes) were used: Alexa Fluor 488 goat anti-guinea pig (A11073), Alexa Fluor 546 goat anti-chicken (A11040), Alexa Fluor 647 goat anti-rabbit (A21244) and streptavidin Alexa Fluor 647 conjugate (S21374). The sections were rinsed 3 times in PBS-T for 5 min and then 3 times in 0.1 M PBS for 5 min before being mounted on microscope slides and sealed in ProLong Glass Antifade (P36984, Invitrogen). Fluorescent images were taken using UPlanXApo ×4 (NA 0.16), UPlanXApo ×10 (NA 0.40), UPlanFL N ×40 (NA 1.30) and UPlanXApo ×60 (NA 1.42) objective lenses on a confocal laser-scanning microscope (FV3000; Olympus) using Fluoview software (Olympus). Laser settings for imaging were kept the same in each experiment. Images represent maximum-intensity projections of optical sections with a step size of 0.5–1.0 μm. Images were processed usinf ImageJ (v1.53u–1.54p). Non-ciliary ARL13B rabbit antibody background staining was sometimes thresholded out of the figure images to improve identification of the primary cilium compared with the background. Example raw images from flattened z stacks are provided in Fig. 2n and Extended Data Fig. 4h. No thresholding was applied to IHC images showing primary cilium staining with ARL13B rat and AC3 antibody, or to any other images.

3D rendering of IHC images was performed using Imaris software (Bitplane, Oxford Instruments). Raw image stacks (0.5 µm) were imported into Imaris and visualized using the Surpass mode to generate 3D representations of the labelled structures. Then, using Imaris software, the following steps were performed to reconstruct the surface and to calculate the volume: selection of a region of interest containing the cilium for segmentation, smoothing with a surface grain size detail of 0.3–0.4 µm (most often 0.4 µm), threshold using the background subtraction method with diameter of largest sphere of 1–1.6 µm (most often 1 µm). This was followed by manual value threshold selection guided by (1) maintaining maximum resemblance to IHC cilium staining captured and (2) ensuring the cilium was rendered as one volume. For rendering of the astrocyte soma surface, the same protocol was followed except that thresholding was performed with the absolute intensity method instead of the background subtraction method. 3D renderings were adjusted for transparency, colour and lighting to facilitate visualization. Volume analysis data were extracted from the 3D reconstructions using the Imaris MeasurementPro module.

RNAscope of mice and human tissue

Fixed-frozen tissue was processed as described above. Serial coronal sections (20 μm) containing amygdala were prepared using a cryostat microtome (Leica) at −20 °C and mounted immediately onto glass slides. Dual ISH–IHC was performed using a Multiplex RNAscope (v.2) with integrated co-detection workflow (ACDBio, 323180 and 323110). Sections were baked for 30 min at 60 °C. Sections were washed for at least 15 min in 0.1 M PBS and then incubated in target retrieval reagents for 5 min at 95 °C. After washing with ddH₂O twice, the sections were dehydrated with 100% ethanol and dried at room temperature. Sections were then incubated with primary antibody guinea pig anti-S100β (1:500; Nittobo, MSFR105350) or chicken anti-S100β (1:500; Synaptic Systems, 287006) overnight at 4 °C. Sections were then incubated with protease (ACDBio, 322330) for 30 min at 40 °C. The sections were washed with ddH₂O twice for 1 min each and then incubated with the corresponding probe for 2 h at 40 °C: Mm-Arl6-C2 (ACDBio, 492301-C2), a custom probe for mouse ARL3-C1 (ACDBio) or Mm-S1PR1-C2 (ACDBio, 426001-C2).

The sections were incubated in AMP 1-FL for 30 min, AMP 2-FL for 30 min and AMP 3-FL for 15 min at 40 °C while washing in wash buffer (ACDBio, 310091) between incubations. The HRP-C1 signal was developed with Opal 520 fluorophore (Akoya Biosciences, FP1487001KT). The HRP-C2 signal was developed with Opal 570 fluorophore (Akoya Biosciences, FP1488001KT). All incubations at 40 °C or 60 °C were performed in a HybEZ hybridization system (ACDBio). Last, sections were incubated with Alexa Fluor goat secondary antibodies described in the IHC section for 30 min at room temperature. Images were obtained in the same way as for IHC (described above) with a step size of 1 μm. Images were processed using ImageJ (v.1.53u–1.54p). Astrocyte somata were labelled with S100β, and the presence of puncta in each soma was quantified.

For RNAscope of human brain tissue, a RNAscope Multiplex Fluorescent Reagent kit v.2 (all reagents from ACDBio) was used with the astrocyte-specific C1 probe ALDH1L1 and the S1PR1 target C2 probe according to the user manual with modifications. In brief, 18 μm coronal sections of fresh-frozen human amygdala samples from four neurotypical individuals (n = 2 male and n = 2 female individuals) were fixed with 4% paraformaldehyde in PBS for 45 min at 4 °C, rinsed with PBS and dehydrated with gradient ethanol solutions. Endogenous peroxidase activity was quenched with hydrogen peroxide solution for 10 min, and protease IV reagent was used for 30 min for proteolytic tissue digestion. After hybridization steps, the signal was developed with TSA Vivid Fluorophores 570 and 650, both at 1:1,000 dilution in TSA buffer. Slides were then counterstained with DAPI, and lipofuscin autofluorescence was quenched by incubating the slides for 1 min with 1× TrueBlack solution (Biotum, 23007). After autofluorescence quenching, slides were rinsed 3 times with PBS and mounted with ProLong Gold Antifade reagent (Thermo Fisher). Four z stack images under ×20 magnification with a step size of 1 μm were acquired per sample from the basolateral region of the amygdala with a Leica SP8 confocal microscope. A total of 632 ALDH1L1-positive nuclei were manually counted using Fiji. The data are expressed as the percentage of S1PR1-positive astrocytes per each case.

Calcium imaging of astrocytes in brain slices

In brief, mice were anaesthetized with isoflurane and decapitated with sharp shears. The brains were placed and sliced in ice-cold modified artificial cerebrospinal fluid (aCSF) containing the following components: 194 mM sucrose, 30 mM NaCl, 4.5 mM KCl, 1 mM MgCl₂, 26 mM NaHCO₃, 1.2 mM NaH₂PO₄ and 10 mM d-glucose, saturated with 95% O₂ and 5% CO₂. A vibratome (DSK Microslicer; Ted Pella) was used to cut 300 μm brain sections. The slices were allowed to equilibrate for 30 min at 33 °C in normal aCSF containing the following components: 124 mM NaCl, 4.5 mM KCl, 2 mM CaCl₂, 1 mM MgCl₂, 26 mM NaHCO₃, 1.2 mM NaH₂PO₄ and 10 mM d-glucose, continuously bubbled with 95% O₂ and 5% CO₂. Slices were then stored at 21–23 °C in the same buffer until use. All slices were used within 6 h of slicing. Brain-slice imaging was performed at room temperature (21–23 °C). Cells for all the experiments were imaged using a confocal microscope (Fluoview 1000; Olympus) with a ×40 water-immersion objective lens with 0.8 NA and a digital zoom of 2–3. We used the 488 nm line of an Argon laser, with the intensity adjusted to 5–10% of the maximum output of 10 mW. The emitted light pathway consisted of an emission high-pass filter (505–525 nm) before the photomultiplier tube. Astrocytes were typically around 25 μm below the slice surface and were scanned at 1 frame per s for imaging sessions. For pharmacological activation of endogenous GPCRs, agonists were dissolved in water. The DREADD agonist DCZ and the S1PR1 agonist ponesimod were dissolved in DMSO. Stock solutions were diluted in aCSF immediately before use. Analyses of time-lapse image series were performed using ImageJ (v.2.1, NIH). The xy drift was corrected using ImageJ, and cells with z drift were excluded from analyses. Time traces of fluorescence intensity were extracted from the regions of interest and converted to change in fluorescent (ΔF/F) values. Traces of the regions of interest were extracted using GECIquant software (v.1.0). Extracted Ca²⁺ signals were analysed using custom R (v.4.4.1) scripts. Events were identified on the basis of amplitudes that were at least threefold above the baseline noise of the ΔF/F trace and analysed using the findpeaks function (pracma v.1.9.9) for event identification and quantification, the AUC function (DescTools v.0.99.55) for area under the curve calculation and tidyverse (v.2.0.0) for data manipulation.

Extraction of astrocytic RNA from Aldh1l1 ^cre/ERT2;RiboTag mice

Extraction of astrocyte-derived RNA from Aldh1l1^cre/ERT2;RiboTag mice was done as previously described⁷⁵. In brief, tissue from the amygdala were collected from 16 animals of 2–3-month-old Aldh1l1^cre/ERT2;RiboTag mice (8 males and 8 females, 4 mice pooled per sample, 4 samples per group) and homogenized. RNA was extracted from 10–20% of cleared lysate as input containing RNA from all cell types using a RNA extraction kit (Qiagen, 74034). The remaining lysate was incubated with 5 µl mouse anti-HA antibody (about 1:160, BioLegend, 901514) for 4 h at 4 °C followed by addition of magnetic beads (Thermo Scientific, 88817) and overnight incubation at 4 °C. RNA was extracted from the IP samples containing astrocyte-enriched RNA using an RNA extraction kit (Qiagen, 74034). RNA concentration and quality were assessed using an Agilent 2100 Bioanalyzer.

Astrocyte RNA-seq and analysis

RNA samples with RNA integrity number greater than 7.8 were used for subsequent multiplexed library preparation with TruSeq Stranded Total RNA with Ribozero Gold, except the MACS and the Ift88^f/f bulk samples, for which library preparation was done with SMART-Seq mRNA and NexteraXT. For each experiment, all samples were multiplexed into a single pool to avoid batch effects. Sequencing was performed on NovaSeq 6000 or NovaSeq × Plus, which produced at least 80 million reads per sample. Demultiplexing was performed using Illumina Bcl2fastq2 (v.17). Reads were aligned to the mouse mm10 reference genome using the STAR spliced read aligner (v.2.7), and 75 ± 16% of the reads were uniquely mapped. Previously reported datasets of the 13 regions of the CNS⁵⁸ were combined with the amygdala dataset by batch correction using Removing Unwanted Variation (RUVr). All DEGs were obtained using limma-voom. Differential gene expression analysis was performed with the Bioconductor package limmaVoom (v.3.60) with the adjusted P value threshold set at <0.05. Astrocyte-enriched genes were defined as those with log₂[IP/input] > 1, adjusted P < 0.05 and IP FPKM > 1. For analysis of astrocyte region-specific genes, genes were analysed by comparing IP samples with the average of IP samples from the other 13 regions using limma-voom. GO pathway analysis was performed using Enrichr (https://maayanlab.cloud/Enrichr/). Unwanted variation in the US, CRS and WAS datasets was removed using RUVr. S1pr1 RNA-seq expression in cortex astrocytes for Fig. 6 was calculated by averaging that of the SCX, the MCX and the VCX⁵⁸.

Cilium-related genes

The list of 770 cilium-related genes to compute region enrichment was obtained by combining genes from the GO terms ‘cilium’ (GO:0005929), ‘cilium assembly’ (GO:0060271), ‘cilium organization’ (GO:0044782) and ‘axoneme’ (GO:0005930). The final list of 770 genes shared 60% overlap with Syscilia Gold Standard (SCGv1) of known ciliary components⁷⁹ and 42% overlap with a longer list of cilia-related genes from CiliaCarta⁸⁰. The full list of genes is provided in Source Data Fig. 4.

In vivo BioID2 protein biotinylation and pull-downs

Three weeks after AAV microinjection of BioID2 or GFP (control), mice were treated with a subcutaneous injection of biotin (24 mg kg^–1; Millipore Sigma, RES1052B-B7) dissolved in sterile 0.1 M PBS once per day for 7 consecutive days. Mice were used 16 h after the last biotin injection.

Purification of biotinylated proteins was conducted as previously described^81,82. Eight mice were used for each biotinylated protein purification. Amygdala tissue (8 mice pooled per sample, 4 samples per group, equal numbers of male and female mice) was dounce-homogenized in lysis buffer A (1 mM EDTA, 150 mM NaCl and 50 mM HEPES pH 7.5 supplemented with Halt protease inhibitor (Thermo Scientific, 78429)). Immediately after homogenization, lysis buffer B (2% sodium deoxycholate, 2% Triton-X, 0.5% SDS, 1 mM EDTA, 150 mM NaCl and 50 mM HEPES pH 7.5) was added. The lysed samples were sonicated for 5 min at 60% power and then centrifuged at 15,000g for 15 min at 4 °C. The resulting supernatant was then ultracentrifuged at 100,000g for 30 min at 4 °C. SDS was added to the supernatant to obtain a final concentration of 1%. The sample was then boiled at 95 °C for 5 min. The sample was cooled on ice and incubated with 35 μl equilibrated anti-pyruvate carboxylase (5 μg; Abcam, 110314) conjugated agarose beads (Pierce, 20398) for 4 h at 4 °C while rotating. Subsequently, the sample was centrifuged at 500g for 5 min at 4 °C and the supernatant was incubated with Pierce streptavidin magnetic beads (Thermo Scientific, 88817) at 4 °C overnight while rotating. The magnetic beads were then washed twice with 0.2% SDS, twice with wash buffer (1% sodium deoxycholate, 1% Triton-X and 25 mM LiCl), twice with 1 M NaCl and 5 times with 50 mM ammonium bicarbonate. Proteins bound to the beads were then eluted in elution buffer (5 mM biotin, 0.1% Rapigest SF surfactant and 50 mM ammonium bicarbonate) at 60 °C for a minimum of 2 h.

Analysis of biotinylated proteins by mass spectrometry

Eluates obtained after IP were subjected to reduction and alkylation using 5 mM Tris(2-carboxyethyl)phosphine and 10 mM iodoacetamide, respectively, followed by treatment with single-pot, solid-phase-enhanced sample preparation protocol (SP3) for clean-up of protein. The eluates obtained after SP3 protein clean-up were subjected to overnight digestion with Lys-C and trypsin at 37 °C. The digested peptides were treated with SP3-based peptide clean-up protocol, dried and subsequently analysed by LC–MS/MS. In brief, reversed-phase chromatography was used to separate peptides on Bruker PepSep C18 columns (150 μm i.d., 15 cm in length and 1.5 μm particle size) using a gradient of increasing acetonitrile delivered by a Vanquish Neo UHPLC-system (Thermo Scientific) operating at a flow rate of 500nl min^–1. Data acquisition was carried out on a Bruker timsTOF HT mass spectrometer using a diaPASEF mode of data acquisition. The dia-PASEF windows spanned a precursor range from 300 to 1,200 m/z using the following ion mobility parameters: 1/K0 start, 0.60 Vs cm⁻²; 1/K0 end, 1.60 Vs cm⁻²; ramp time, 100 ms; accumulation time, 50 ms.

LC–MS/MS analysis for ARL13B–BioID experiments was carried out on a Vanquish Neo UHPLC system coupled to an Astral mass spectrometer (Thermo Fisher Scientific). Peptide separation was performed using a trap-and-elute workflow on an ionopticks C18 reverse-phase column (8 cm × 150 µm, 1.7 µm particle size) maintained at 52 °C. Mobile phase A consisted of water containing 0.1% formic acid, whereas mobile phase B consisted of acetonitrile with 0.1% formic acid. A 15-min gradient was applied as follows: 5% B from 0 to 1 min (2.45 µl min^–1); 5–15% B from 1 to 5 min (1.75 µl min^–1); 15–25% B from 5 to 12.6 min (1.75 µl min^–1); 25–38% B from 12.6 to 13.6 min (1.75 µl min^–1); 38–80% B from 13.6 to 13.7 min (2.45 µl min^–1); followed by a hold at 80% B for 15 min (2.45 µl min^–1). Data-independent acquisition (DIA) was performed on an Orbitrap Astral mass spectrometer operating in positive electrospray ionization mode. MS1 spectra were acquired at 240,000 resolution across an m/z range of 380–980, with a normalized AGC target of 500% and a maximum injection time of 3 ms. DIA used sequential 4 m/z isolation windows covering a precursor range of 380–980 m/z. MS2 scans were acquired at 80,000 resolution using a normalized HCD collision energy of 25%, a normalized AGC target of 500% and a maximum injection time of 7 ms. For ARL13B proteomics, we had to use the cortex instead of the amygdala because the number of mice needed to perform the proteomic experiment (about 64) was prohibitive because of the small size of the amygdala. However, as there is limited information on the astrocyte primary cilium proteome, the cortex was a good place to start because astrocytes have ARL13B-positive primary cilia in the cortex and multidimensional scaling has shown that amygdala and cortical astrocytes cluster together at a molecular level.

The dia-PASEF files and the Thermo RAW files obtained after LC–MS/MS analysis were searched using DIA-NN (v.1.8.1) with an in silico library generated from a UniProt database containing all mouse proteins⁸³. Statistical analysis of hits and DEPs was done using the Bioconductor package limma (v.3.60). A protein was declared a hit when it satisfied the following filters: log₂[BioID2/GFP] > 1 and adjusted P < 0.05. GO pathway analysis was performed using Enrichr (https://maayanlab.cloud/Enrichr/). Putative localization in the primary cilium of high probability hits from proteomics was based on the following published work: TULP3 (ref. ⁸⁴), ARL3 (ref. ⁸⁵), IFT74 (ref. ⁸⁶), TOGARAM1 (ref. ⁸⁷), MAP4 (ref. ⁸⁸), SNAP29 (ref. ⁸⁹), RAB34 (ref. ⁹⁰), EHD1 (ref. ⁸⁹), DLG1 (ref. ⁹¹), EZR⁹², INPP5E⁹³ and IFT43 (ref. ⁹⁴).

Molecular cloning and PHP.eB AAV generation

To generate the ARL13B–BioID2 and GFP plasmids, the Astro-BioID2 (Addgene plasmid 176740) or Astro-GFP (Addgene plasmid 176861) plasmid was cut by restriction digestion at the XhoI site. cDNA for ARL13B was amplified from Addgene plasmid 232880 (ref. ⁹⁵) and cloned into the XhoI site using In-fusion cloning (Takara Bio). The plasmids were packaged into PHP.eB AAVs in-house following a published protocol⁹⁶.

Retroorbital delivery of PHP.eB AAVs

Brain-wide expression of ARL13B–BioID2 and ARL13B–GFP in astrocytes in 8-week-old mice was achieved by retro-orbital injection into the eye sinus with 10¹² genome copies per mouse of AAV-PHP.eB-GfaABC1D-ARL13B-BioID2-HA or AAV-PHP.eB-GfaABC1D-ARL13B-GFP⁹⁷. The biotin injections were performed as indicated above. The entire cortex was dissected and used for streptavidin-mediated pull-down of biotinylated proteins as indicated above. The plasmids are available from Addgene (identifiers 253927 and 253928).

CilioGenics probability

Data were downloaded⁹⁸ from https://ciliogenics.com/?page=Home. Proteins were assigned a probability on the basis of the assigned CilioGenics score as per the website: high = 1, medium = 1–0.5, low < 0.5.

MACS-based astrocyte isolation and qPCR

The protocol was adapted from a previous study⁹⁹. Amygdala tissue collected from 3–4 adult mice was homogenized together for each sample. The tissue was incubated in papain at 37 °C for 30 min with a constant supply of 95% O₂ and 5% CO₂ on the liquid surface, then centrifuged and resuspended in cold PBS, filtered through a 70 µm cell strainer and debris was removed with Debris Removal solution (130-109-398, Miltenyi Biotec) following the manufacturer’s instructions. The resulting supernatant was resuspended in 90 µl 0.5% BSA (A7030, Millipore Sigma) containing 1 mM EDTA (15575-020, Invitrogen) and incubated with 10 µl FcR blocking beads (130-097-678, Miltenyi Biotec) for 10 min at 4 °C, followed by the addition of 10 µl ACSA-2 beads and incubation for 15 min at 4 °C. Then, 2 ml BSA–EDTA was added, the solution was centrifuged, resuspended in BSA–EDTA and filtered through a MS column (130-042-201, Miltenyi Biotec) placed on a MACS magnet. The column was eluted with BSA–EDTA and RNA was isolated with a RNA extraction kit (Qiagen, 74034). cDNA was obtained from RNA with SuperScript IV (18090010, Invitrogen) following the manufacturer’s instructions. qPCR was performed with the following primers: Aldh1l1, Rbfox3 and Aif1 sequences have been previously described¹⁰⁰; Slc1a3 fw 5′-GCGATTGGTCGCGGTGATAATG-3′; Slc1a3 rv 5′-CGACAATGACTGTCACGGTGTAC-3′; Mbp fw 5′-ATTCACCGAGAGGCTGGAA-3′; Mbp rv 5′-TGTGTGCTTGGAGTCTGTCACC-3′; Ppia fw 5′-CATACAGGTCCTGGCATCTTGTC-3′; and Ppia rv 5′-AGACCACATGCTTGCCATCCAG-3′. Gene expression was calculated as C_t values relative to the housekeeping gene Ppia in each fraction (astrocyte or input) and then plotted as the ratio of the C_t values of the two fractions.

Human amygdala astrocyte snRNA-seq analysis

Frozen amygdala tissue samples from 120 brain donors were obtained from the Harvard Brain Tissue Resource Center. Sets of 20 brain tissue specimens (each consisting of donors with PSDs and unaffected individuals) were processed at once as a single pooled sample, including nucleus extraction, generation of gel beads-in-emulsion and library preparation performed according to the 10x Chromium Single Nuclei 3′ v3.1 protocol. Raw sequencing reads were aligned to the hg38 reference genome using the standard Drop-seq (v.2.5.4) workflow. Reads were assigned to annotated genes if they mapped to exons or introns of those genes. Ambient or background RNA was removed from digital gene expression matrices using CellBender (v.0.3.0) remove-background. Dropulation (v.2.5.4) was used to assign each nucleus to its donor of origin on the basis of transcribed single-nucleotide polymorphisms. Cell-type assignment was accomplished with cell-classification models trained using scPred (v.1.9.2).

scRNA-seq data from astrocytes were pseudo-bulked by summing raw gene counts across all astrocyte cells (with ≥200 detected genes, to exclude low-quality nuclei) from each donor. Donors with fewer than 30 astrocytes were excluded, which resulted in 109 donors (71 with PSDs and 38 unaffected individuals; mean age 65.07 ± 17.21 years; n = 53 women). Differential expression analyses were performed in edgeR (v.4.2.1) using quasi-likelihood negative binomial regression with glmQLFit(), and testing was carried out with glmQLFTest() for the contrast between PSD and control donors. Normalization was conducted using the TMM method via calcNormFactors(), and lowly expressed genes were filtered with filterByExpr(). Age, sex and data collection batch were included as covariates in the design matrix, and multiple testing correction was performed using the Benjamini–Hochberg method to control the false discovery rate. We report the astrocyte data for the primary cilium-related genes here, but a follow-up study from the laboratory of S.A.M. will report all cell types of the amygdala in due course.

Statistical analysis and data presentation

Data from every experiment represent at least three replicates. Sample sizes were based on previous experiments using similar or identical experiments and the use of similar models by our laboratory. Statistical tests, unless otherwise stated, were run in OriginPro 2024. In bar plots, data are presented as the mean (bar) ± s.e.m. (error bars) along with the individual data points. The results of statistical comparisons, n numbers and significance levels are shown in the figure panels along with the average data. n is defined as the number of cells or mice on a case-by-case basis throughout the paper. We determined whether each set of data was normally distributed using OriginPro 2024. If the data were normally distributed, we used parametric tests; if they were not normally distributed, we used nonparametric tests. Unpaired Student’s two-tailed t-tests, two-tailed Mann–Whitney tests and one-way and two-way analysis of variance (ANOVA) tests were used for most statistical analyses comparing two groups. One-way ANOVA followed by Tukey’s post-hoc test or Kruskal–Wallis ANOVA followed by Dunn’s test were used for statistical analyses comparing three or more groups. Significance was declared at P < 0.05. When P values were greater than 0.05, they are stated as not significant. When the P value was less than 0.01, it is stated as <0.01. When the P value was less than 0.001, it is stated as <0.001. All transcriptomic and proteomic analyses used a statistical adjusted P < 0.05 unless otherwise stated. No data points were excluded from any experiment. Data used to generate the graphs shown in the figures are provided in the Source Data files. The results of all statistical tests are provided in Supplementary Table 1.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

For RNA-seq, differential gene expression and enrichment analysis used the R package limma-voom to process RNA counts (v.3.60; https://rdrr.io/bioc/limma/man/voom.html) and batch correction and removal of unwanted variation was done with RUVr¹⁰¹. scRNA-seq data were analysed using Seurat (https://CRAN.R-project.org/package=Seurat). Visualization of set intersections was done with UpSetR (v.1.4.0; https://github.com/hms-dbmi/UpSetR). For proteomics, label-free quantification was carried out using DIA-NN (v.1.8.1). Principal component data visualization was conducted with the R package DESeq2 (v.1.44.0; https://bioconductor.org/packages/release/bioc/html/DESeq2.html). Differential protein expression and enrichment analysis was conducted using limma (v.3.60; https://bioconductor.org/packages/release/bioc/html/limma.html). For calcium imaging, data were analysed in R using GECIquant (v.1.0), pracma (v1.9.9; https://cran.r-project.org/web/packages/pracma/index.html) and DescTools (v.0.99.55; https://cran.r-project.org/web/packages/DescTools/index.html).

1. Nestler, E. J. et al. Neurobiology of depression. Neuron 34, 13–25 (2002).

   Article  CAS  PubMed  Google Scholar 

2. Mill, P., Christensen, S. T. & Pedersen, L. B. Primary cilia as dynamic and diverse signalling hubs in development and disease. Nat. Rev. Genet. 24, 421–441 (2023).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

3. Volos, P., Fujise, K. & Rafiq, N. M. Roles for primary cilia in synapses and neurological disorders. Trends Cell Biol. 35, 6–10 (2025).

   Article  CAS  PubMed  Google Scholar 

4. Liu, S., Trupiano, M. X., Simon, J., Guo, J. & Anton, E. S. The essential role of primary cilia in cerebral cortical development and disorders. Curr. Top. Dev. Biol. 142, 99–146 (2021).

   Article  PubMed  PubMed Central  Google Scholar 

5. Wang, L. et al. Primary cilia signaling in astrocytes mediates development and regional-specific functional specification. Nat. Neurosci. 27, 1708–1720 (2024).

   Article  CAS  PubMed  Google Scholar 

6. Macarelli, V., Leventea, E. & Merkle, F. T. Regulation of the length of neuronal primary cilia and its potential effects on signalling. Trends Cell Biol. 33, 979–990 (2023).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

7. Khakh, B. S. & McCarthy, K. D. Astrocyte calcium signaling: from observations to functions and the challenges therein. Cold Spring Harb. Perspect. Biol. 7, a020404 (2015).

   Article  PubMed  PubMed Central  Google Scholar 

8. Wang, C. E. & Ogden, S. K. G protein-coupled receptor signal intersection at the primary cilium. BioEssays https://doi.org/10.1002/bies.70015 (2025).

9. Schou, K. B., Pedersen, L. B. & Christensen, S. T. Ins and outs of GPCR signaling in primary cilia. EMBO Rep. 16, 1099–1113 (2015).

   Article  CAS  PubMed  PubMed Central  Google Scholar 

10. Price, J. L. & Drevets, W. C. Neural circuits underlying the pathophysiology of mood disorders. Trends Cogn. Sci. 16, 61–71 (2012).

    Article  PubMed  Google Scholar 

11. Krishnan, V. & Nestler, E. J. The molecular neurobiology of depression. Nature 455, 894–902 (2008).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

12. Zhang, W. H., Zhang, J. Y., Holmes, A. & Pan, B. X. Amygdala circuit substrates for stress adaptation and adversity. Biol. Psychiatry 89, 847–856 (2021).

    Article  PubMed  Google Scholar 

13. Cathomas, F. et al. Beyond the neuron: role of non-neuronal cells in stress disorders. Neuron 110, 1116–1138 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

14. Nagai, J. et al. Behaviorally consequential astrocytic regulation of neural circuits. Neuron 109, 576–596 (2021).

    Article  CAS  PubMed  Google Scholar 

15. Nagai, J. et al. Hyperactivity with disrupted attention by activation of an astrocyte synaptogenic cue. Cell 177, 1280–1292 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

16. Ollivier, M. et al. Crym-positive striatal astrocytes gate perseverative behaviour. Nature 627, 358–366 (2024).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

17. Soto, J. S. et al. Astrocyte G_i–GPCR signaling corrects compulsive-like grooming and anxiety-related behaviors in Sapap3 knockout mice. Neuron 112, 3412–3423 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

18. Yu, X. et al. Reducing astrocyte calcium signaling in vivo alters striatal microcircuits and causes repetitive behavior. Neuron 99, 1170–1187 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

19. Cui, Y. et al. Astroglial Kir4.1 in the lateral habenula drives neuronal bursts in depression. Nature 554, 323–327 (2018).

    Article  ADS  CAS  PubMed  Google Scholar 

20. Xin, Q. et al. Neuron-astrocyte coupling in lateral habenula mediates depressive-like behaviors. Cell https://doi.org/10.1016/j.cell.2025.04.010 (2025).

21. Zhou, X. et al. Pharmacological inhibition of Kir4.1 evokes rapid-onset antidepressant responses. Nat. Chem. Biol. 20, 857–866 (2024).

    Article  CAS  PubMed  Google Scholar 

22. Chen, A. B. et al. Norepinephrine changes behavioral state through astroglial purinergic signaling. Science 388, 769–775 (2025).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

23. Mu, Y. et al. Glia accumulate evidence that actions are futile and suppress unsuccessful behavior. Cell 178, 27–43 (2019).

    Article  CAS  PubMed  Google Scholar 

24. Murphy-Royal, C. et al. Stress gates an astrocytic energy reservoir to impair synaptic plasticity. Nat. Commun. 11, 2014 (2020).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

25. Ghenissa, O. et al. Basolateral amygdala astrocytes encode anxiety states. Neuron https://doi.org/10.1016/j.neuron.2026.02.038 (2026).

26. Naskar, S. & Chattarji, S. Stress elicits contrasting effects on the structure and number of astrocytes in the amygdala versus hippocampus. eNeuro https://doi.org/10.1523/eneuro.0338-18.2019 (2019).

27. Nentwig, T. B. et al. Central amygdala astrocyte plasticity underlies GABAergic dysregulation in ethanol dependence. Transl. Psychiatry 15, 132 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

28. Guayasamin, M. et al. Early-life stress induces persistent astrocyte dysfunction associated with fear generalisation. eLife https://doi.org/10.7554/eLife.99988 (2025).

29. Sun, W. et al. Spatial transcriptomics reveal neuron–astrocyte synergy in long-term memory. Nature 627, 374–381 (2024).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

30. Martin-Fernandez, M. et al. Synapse-specific astrocyte gating of amygdala-related behavior. Nat. Neurosci. 20, 1540–1548 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

31. Bukalo, O. et al. Astrocytes enable amygdala neural representations supporting memory. Nature 652, 434–441 (2026).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

32. Tafet, G. E. & Nemeroff, C. B. The links between stress and depression: psychoneuroendocrinological, genetic, and environmental interactions. J. Neuropsychiatry Clin. Neurosci. 28, 77–88 (2016).

    Article  PubMed  Google Scholar 

33. Hilgendorf, K. I., Myers, B. R. & Reiter, J. F. Emerging mechanistic understanding of cilia function in cellular signalling. Nat. Rev. Mol. Cell Biol. 25, 555–573 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

34. Lahaie, S. C. et al. Characterization of astrocytic primary cilia in the adult mouse cortex and hippocampus. Heliyon 11, e44010 (2025).

    Article  CAS  Google Scholar 

35. Ott, C. M. et al. Ultrastructural differences impact cilia shape and external exposure across cell classes in the visual cortex. Curr. Biol. 34, 2418–2433 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

36. Bonaz, B. & Taché, Y. Water-avoidance stress-induced c-fos expression in the rat brain and stimulation of fecal output: role of corticotropin-releasing factor. Brain Res. 641, 21–28 (1994).

    Article  CAS  PubMed  Google Scholar 

37. Corder, G. et al. An amygdalar neural ensemble that encodes the unpleasantness of pain. Science 363, 276–281 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

38. Chung, E. N. et al. Psychedelic control of neuroimmune interactions governing fear. Nature 641, 1276–1286 (2025).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

39. Xu, N. et al. Spared nerve injury increases the expression of microglia M1 markers in the prefrontal cortex of rats and provokes depression-like behaviors. Front. Neurosci. 11, 209 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

40. Guida, F. et al. Behavioral, biochemical and electrophysiological changes in spared nerve injury model of neuropathic pain. Int. J. Mol. Sci. https://doi.org/10.3390/ijms21093396 (2020).

41. Duan, G. B. et al. Overexpression of EphB2 in the basolateral amygdala is crucial for inducing visceral pain sensitization in rats subjected to water avoidance stress. CNS Neurosci. Ther. 30, e14611 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

42. Tan, S., Wang, Y., Chen, K., Long, Z. & Zou, J. Ketamine alleviates depressive-like behaviors via down-regulating inflammatory cytokines induced by chronic restraint stress in mice. Biol. Pharm. Bull. 40, 1260–1267 (2017).

    Article  CAS  PubMed  Google Scholar 

43. Selye, H. A syndrome produced by diverse nocuous agents. Nature 138, 32–32 (1936).

    Article  ADS  Google Scholar 

44. Wang, K. X., Shi, Y. F., Ron, Y., Kazanecki, C. C. & Denhardt, D. T. Plasma osteopontin modulates chronic restraint stress-induced thymus atrophy by regulating stress hormones: inhibition by an anti-osteopontin monoclonal antibody. J. Immunol. 182, 2485–2491 (2009).

    Article  CAS  PubMed  Google Scholar 

45. Ulrich-Lai, Y. M. et al. Chronic stress induces adrenal hyperplasia and hypertrophy in a subregion-specific manner. Am. J. Physiol. Endocrinol. Metab. 291, E965–E973 (2006).

    Article  CAS  PubMed  Google Scholar 

46. Tarcic, N., Ovadia, H., Weiss, D. W. & Weidenfeld, J. Restraint stress-induced thymic involution and cell apoptosis are dependent on endogenous glucocorticoids. J. Neuroimmunol. 82, 40–46 (1998).

    Article  CAS  PubMed  Google Scholar 

47. McCullough, K. M. et al. Blood levels of T-cell receptor excision circles (TRECs) provide an index of exposure to traumatic stress in mice and humans. Transl. Psychiatry 12, 423 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

48. Rajkowska, G. & Stockmeier, C. A. Astrocyte pathology in major depressive disorder: insights from human postmortem brain tissue. Curr. Drug Targets 14, 1225–1236 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

49. Escartin, C. et al. Reactive astrocyte nomenclature, definitions, and future directions. Nat. Neurosci. 24, 312–325 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

50. Cipriani, A. et al. Comparative efficacy and acceptability of 21 antidepressant drugs for the acute treatment of adults with major depressive disorder: a systematic review and network meta-analysis. Lancet 391, 1357–1366 (2018).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

51. Walker, A. K. et al. NMDA receptor blockade by ketamine abrogates lipopolysaccharide-induced depressive-like behavior in C57BL/6J mice. Neuropsychopharmacology 38, 1609–1616 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

52. Berman, R. M. et al. Antidepressant effects of ketamine in depressed patients. Biol. Psychiatry 47, 351–354 (2000).

    Article  CAS  PubMed  Google Scholar 

53. Higginbotham, H. et al. Arl13b in primary cilia regulates the migration and placement of interneurons in the developing cerebral cortex. Dev. Cell 23, 925–938 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

54. Higginbotham, H. et al. Arl13b-regulated cilia activities are essential for polarized radial glial scaffold formation. Nat. Neurosci. 16, 1000–1007 (2013).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

55. Kasahara, K., Miyoshi, K., Murakami, S., Miyazaki, I. & Asanuma, M. Visualization of astrocytic primary cilia in the mouse brain by immunofluorescent analysis using the cilia marker Arl13b. Acta Medica Okayama 68, 317–322 (2014).

    PubMed  Google Scholar 

56. Sterpka, A. & Chen, X. Neuronal and astrocytic primary cilia in the mature brain. Pharmacol. Res. 137, 114–121 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

57. Siletti, K. et al. Transcriptomic diversity of cell types across the adult human brain. Science 382, eadd7046 (2023).

    Article  CAS  PubMed  Google Scholar 

58. Endo, F. et al. Molecular basis of astrocyte diversity and morphology across the CNS in health and disease. Science 378, eadc9020 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

59. Shigetomi, E., Patel, S. & Khakh, B. S. Probing the complexities of astrocyte calcium signaling. Trends Cell Biol. 26, 300–312 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

60. Yu, X. et al. Context-specific striatal astrocyte molecular responses are phenotypically exploitable. Neuron 108, 1146–1162 (2020).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

61. Flock, T. et al. Selectivity determinants of GPCR–G-protein binding. Nature 545, 317–322 (2017).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

62. Harding, S. D. et al. The IUPHAR/BPS guide to pharmacology in 2024. Nucleic Acids Res. 52, D1438–D1449 (2024).

    Article  PubMed  PubMed Central  Google Scholar 

63. Inoue, A. et al. Illuminating G-protein-coupling selectivity of GPCRs. Cell 177, 1933–1947 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

64. Pándy-Szekeres, G. et al. GPCRdb in 2018: adding GPCR structure models and ligands. Nucleic Acids Res. 46, D440–D446 (2018).

    Article  PubMed  PubMed Central  Google Scholar 

65. Roth, B. L. DREADDs for neuroscientists. Neuron 89, 683–694 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

66. Pérez-Jeldres, T., Alvarez-Lobos, M. & Rivera-Nieves, J. Targeting sphingosine-1-phosphate signaling in immune-mediated diseases: beyond multiple sclerosis. Drugs 81, 985–1002 (2021).

    Article  PubMed  PubMed Central  Google Scholar 

67. Chen, J. et al. Astrocyte growth is driven by the Tre1/S1pr1 phospholipid-binding G protein-coupled receptor. Neuron 112, 93–112 (2024).

    Article  CAS  PubMed  Google Scholar 

68. Weigel, C., Bellaci, J. & Spiegel, S. Sphingosine-1-phosphate and its receptors in vascular endothelial and lymphatic barrier function. J. Biol. Chem. 299, 104775 (2023).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

69. Hansen, J. N. et al. Intrinsic heterogeneity of primary cilia revealed through spatial proteomics. Cell 188, 6804–6824 (2025).

    Article  CAS  PubMed  Google Scholar 

70. Alhassen, W. et al. Patterns of cilia gene dysregulations in major psychiatric disorders. Prog. Neuropsychopharmacol. Biol. psychiatry 109, 110255 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

71. Werner, F. et al. Psychosomatic–psychotherapeutic treatment of stress-related disorders impacts the sphingolipid metabolism towards increased sphingosine and sphingosine-1-phosphate levels. Eur. Arch. Psychiatry Clin. Neurosci. 275, 2049–2058 (2025).

    Article  PubMed  PubMed Central  Google Scholar 

72. Su, C. Y., Bay, S. N., Mariani, L. E., Hillman, M. J. & Caspary, T. Temporal deletion of Arl13b reveals that a mispatterned neural tube corrects cell fate over time. Development 139, 4062–4071 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

73. Haycraft, C. J. et al. Intraflagellar transport is essential for endochondral bone formation. Development 134, 307–316 (2007).

    Article  CAS  PubMed  Google Scholar 

74. Chai, H. et al. Neural circuit-specialized astrocytes: transcriptomic, proteomic, morphological, and functional evidence. Neuron 95, 531–549 (2017).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

75. Srinivasan, R. et al. New transgenic mouse lines for selectively targeting astrocytes and studying calcium signals in astrocyte processes in situ and in vivo. Neuron 92, 1181–1195 (2016).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

76. Decosterd, I. & Woolf, C. J. Spared nerve injury: an animal model of persistent peripheral neuropathic pain. Pain 87, 149–158 (2000).

    Article  PubMed  Google Scholar 

77. Dixon, W. J. Efficient analysis of experimental observations. Annu. Rev. Pharmacol. Toxicol. 20, 441–462 (1980).

    Article  CAS  PubMed  Google Scholar 

78. Diaz-Castro, B., Bernstein, A. M., Coppola, G., Sofroniew, M. V. & Khakh, B. S. Molecular and functional properties of cortical astrocytes during peripherally induced neuroinflammation. Cell Rep. 36, 109508 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

79. van Dam, T. J., Wheway, G., Slaats, G. G., Huynen, M. A. & Giles, R. H. The SYSCILIA gold standard (SCGSv1) of known ciliary components and its applications within a systems biology consortium. Cilia 2, 7 (2013).

    Article  PubMed  PubMed Central  Google Scholar 

80. van Dam, T. J. P. et al. CiliaCarta: an integrated and validated compendium of ciliary genes. PLoS ONE 14, e0216705 (2019).

    Article  PubMed  PubMed Central  Google Scholar 

81. Soto, J. S. et al. Astrocyte–neuron subproteomes and obsessive–compulsive disorder mechanisms. Nature 616, 764–773 (2023).

    Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

82. Soto, J. S., Jami-Alahmadi, Y., Wohlschlegel, J. A. & Khakh, B. S. In vivo identification of astrocyte and neuron subproteomes by proximity-dependent biotinylation. Nat. Protoc. 19, 896–927 (2024).

    Article  CAS  PubMed  Google Scholar 

83. Demichev, V., Messner, C. B., Vernardis, S. I., Lilley, K. S. & Ralser, M. DIA-NN: neural networks and interference correction enable deep proteome coverage in high throughput. Nat. Methods 17, 41–44 (2020).

    Article  CAS  PubMed  Google Scholar 

84. Mukhopadhyay, S. et al. TULP3 bridges the IFT-A complex and membrane phosphoinositides to promote trafficking of G protein-coupled receptors into primary cilia. Genes Dev. 24, 2180–2193 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

85. Powell, L., Samarakoon, Y. H., Ismail, S. & Sayer, J. A. ARL3, a small GTPase with a functionally conserved role in primary cilia and immune synapses. Small GTPases 12, 167–176 (2021).

    Article  CAS  PubMed  Google Scholar 

86. Brown, J. M., Cochran, D. A., Craige, B., Kubo, T. & Witman, G. B. Assembly of IFT trains at the ciliary base depends on IFT74. Curr. Biol. 25, 1583–1593 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

87. Latour, B. L. et al. Dysfunction of the ciliary ARMC9/TOGARAM1 protein module causes Joubert syndrome. J. Clin. Invest. 130, 4423–4439 (2020).

    CAS  PubMed  PubMed Central  Google Scholar 

88. Ghossoub, R. et al. Septins 2, 7 and 9 and MAP4 colocalize along the axoneme in the primary cilium and control ciliary length. J. Cell Sci. 126, 2583–2594 (2013).

    CAS  PubMed  PubMed Central  Google Scholar 

89. Lu, Q. et al. Early steps in primary cilium assembly require EHD1/EHD3-dependent ciliary vesicle formation. Nat. Cell Biol. 17, 228–240 (2015).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

90. Stuck, M. W., Chong, W. M., Liao, J. C. & Pazour, G. J. Rab34 is necessary for early stages of intracellular ciliogenesis. Curr. Biol. 31, 2887–2894 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

91. Manneville, J. B., Jehanno, M. & Etienne-Manneville, S. Dlg1 binds GKAP to control dynein association with microtubules, centrosome positioning, and cell polarity. J. Cell Biol. 191, 585–598 (2010).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

92. Hebert, A. M., DuBoff, B., Casaletto, J. B., Gladden, A. B. & McClatchey, A. I. Merlin/ERM proteins establish cortical asymmetry and centrosome position. Genes Dev. 26, 2709–2723 (2012).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

93. Zhang, R., Tang, J., Li, T., Zhou, J. & Pan, W. INPP5E and coordination of signaling networks in cilia. Front. Mol. Biosci. 9, 885592 (2022).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

94. Zhu, B. et al. Functional exploration of the IFT-A complex in intraflagellar transport and ciliogenesis. PLoS Genet. 13, e1006627 (2017).

    Article  PubMed  PubMed Central  Google Scholar 

95. DeCaen, P. G. & Kimura, L. F. Methods to assess neuronal primary cilia electrochemical signaling. J. Cell. Physiol. 240, e70034 (2025).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

96. Challis, R. C. et al. Systemic AAV vectors for widespread and targeted gene delivery in rodents. Nat. Protoc. 14, 379–414 (2019).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

97. Nagai, J. et al. Specific and behaviorally consequential astrocyte G_q GPCR signaling attenuation in vivo with iβARK. Neuron 109, 2256–2274 (2021).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

98. Pir, M. S. et al. CilioGenics: an integrated method and database for predicting novel ciliary genes. Nucleic Acids Res. 52, 8127–8145 (2024).

    Article  CAS  PubMed  PubMed Central  Google Scholar 

99. Holt, L. M., Stoyanof, S. T. & Olsen, M. L. Magnetic cell sorting for in vivo and in vitro astrocyte, neuron, and microglia analysis. Curr. Protoc. Neurosci. 88, e71 (2019).

    Article  PubMed  PubMed Central  Google Scholar 

100. Bravo-Ferrer, I., Khakh, B. S. & Díaz-Castro, B. Cell-specific RNA purification to study translatomes of mouse central nervous system. STAR Protoc. 3, 101397 (2022).

     Article  CAS  PubMed  PubMed Central  Google Scholar 

101. Risso, D., Ngai, J., Speed, T. P. & Dudoit, S. Normalization of RNA-seq data using factor analysis of control genes or samples. Nat. Biotechnol. 32, 896–902 (2014).

     Article  CAS  PubMed  PubMed Central  Google Scholar 

102. Zhang, Y. et al. An RNA-sequencing transcriptome and splicing database of glia, neurons, and vascular cells of the cerebral cortex. J. Neurosci. 34, 11929–11947 (2014).

     Article  CAS  PubMed  PubMed Central  Google Scholar 

103. Liddelow, S. A. et al. Neurotoxic reactive astrocytes are induced by activated microglia. Nature 541, 481–487 (2017).

     Article  ADS  CAS  PubMed  PubMed Central  Google Scholar 

104. O’Shea, T. M. et al. Lesion environments direct transplanted neural progenitors towards a wound repair astroglial phenotype in mice. Nat. Commun. 13, 5702 (2022).

     Article  ADS  PubMed  PubMed Central  Google Scholar 

105. Cahoy, J. D. et al. A transcriptome database for astrocytes, neurons, and oligodendrocytes: a new resource for understanding brain development and function. J. Neurosci. 28, 264–278 (2008).

     Article  CAS  PubMed  PubMed Central  Google Scholar 

Download references

Author notes

1. These authors contributed equally: Sara G. Pelaz, Katsukuni Mitsui

Authors and Affiliations

1. Department of Physiology, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA, USA

   Sara G. Pelaz, Katsukuni Mitsui, Chiranjivi Neupane, Vanessa H. Casha, Marta Alonso-Gardón & Baljit S. Khakh

2. Research Center of Neurology, ONO Pharmaceutical, Osaka, Japan

   Katsukuni Mitsui

3. McLean Hospital, Belmont, MA, USA

   Natalia Kolosowska & Sabina Berretta

4. Department of Psychiatry, Program in Neuroscience, Harvard Medical School, Boston, MA, USA

   Natalia Kolosowska & Sabina Berretta

5. Broad Institute of MIT and Harvard, Cambridge, MA, USA

   Haley Fritch, Steven A. McCarroll & Sabina Berretta

6. Department of Biological Chemistry, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA, USA

   Vijaya Pandey & James A. Wohlschlegel

7. Department of Cell Biology and Anatomy, Alberta Children’s Hospital Research Institute, Hotchkiss Brain Institute, Cumming School of Medicine, University of Calgary, Calgary, Alberta, Canada

   Lizheng Wang & Jiami Guo

8. Program in Neurogenetics, Department of Neurology, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA, USA

   Riki Kawaguchi

9. Center for Neurobehavioral Genetics, Semel Institute for Neuroscience and Human Behavior, University of California Los Angeles, Los Angeles, CA, USA

   Riki Kawaguchi

10. Howard Hughes Medical Institute and Harvard Medical School Department of Genetics, Boston, MA, USA

    Steven A. McCarroll

11. Department of Neurobiology, David Geffen School of Medicine, University of California Los Angeles, Los Angeles, CA, USA

    Baljit S. Khakh

12. UK Dementia Research Institute, Cardiff University, Cardiff, UK

    Baljit S. Khakh

Contributions

K.M. initiated the project and performed most of the in vivo work, and S.G.P. performed most of the in vitro work and multiomic data analyses. K.M. and S.G.P. therefore performed most of the experiments and are listed as co-first authors for that reason. K.M. and S.G.P. also analysed data with guidance from B.S.K. while K.M. was at UCLA. N.K. performed human tissue validations under the guidance of S.B. C.N. performed essential mouse surgeries. V.H.C. generated some of the PHP.eB AAVs. M.A.-G. performed PHP.eB and biotin injections. V.P. operated the mass spectrometry machine, performed quality control and all initial proteomic data analyses under the guidance of J.A.W. L.W. bred Arl13b^f/f;Aldh1l1^cre/ERT2 mice and helped with tissue dissections under the guidance of J.G. R.K. performed all initial mouse RNA-seq data analyses. H.F. and S.A.M. performed and analysed human astrocyte snRNA-seq data from PSD samples. B.S.K. conceived the study with K.M., planned and directed the experiments, gathered funding, space, collaborators and resources, and planned the final figures jointly with S.G.P. and K.M. S.G.P., K.M. and B.S.K. discussed and addressed the reviewers’ comments with help from other authors. B.S.K. wrote the final version with help from S.G.P. and K.M. All authors contributed to and edited the manuscript.

Corresponding authors

Correspondence to Katsukuni Mitsui or Baljit S. Khakh.

Competing interests

K.M. is employed by ONO Pharmaceutical. The other authors declare no competing interests.

Peer review information

Nature thanks Michelle Olsen and the other, anonymous, reviewer(s) for their contribution to the peer review of this work. Peer reviewer reports are available.

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/.

Reprints and permissions