Mice prenatally exposed to CORT out-of-phase re-adapt slower to a shift in the LD cycle

Previous evidence from our lab showed that the maternal exposure to prenatal GCs at the “wrong” time of day (i.e., out-of-phase, misaligned to the maternal physiological rhythms) altered the offsprings’ GCs homeostasis [10]. Interestingly, if the mother is exposed to the same CORT dose but at the beginning of the active phase (i.e., in-phase, aligned to the maternal physiological rhythms), the offspring behavior is comparable to that of naïve mice, whose mothers were undisturbed during pregnancy [10]. Since the out-of-phase offspring produced overall higher levels of GCs [10] and the resynchronization after jet lag depends on the levels and circadian phase of adrenal GCs [18,39], we asked whether the re-entrainment capacity to a 6-hour phase advance in the LD cycle (jet lag) would be affected in our model (Fig 1a). The representative actograms show the re-adaption of the locomotor activity onset during jet lag in adult mice exposed prenatally to CORT in- and out-of-phase and age-matched naive mice (Figs 1b and S1). To quantify the speed of re-entrainment, we calculated phase-shift 50 (PS₅₀), defined as the time at which half of the 6-hour phase shift is completed. We observed that the kinetics of the re-adaptation to the new LD cycle was slower in out-of-phase exposed mice compared to in-phase and naive mice (Fig 1c and 1d). This slower re-adaptation did not seem to depend on a weaker entrainment to the LD prior to the shift (S1a–S1c Fig) or on the sex of the mice (S1d Fig). Interestingly, in mice prenatally exposed to CORT out-of-phase, we found a reduced expression of GR in the SCN. Astrocytes seem to be the main contributor of this difference (Figs 1e, 1f, and S1e).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 1. Mice prenatally exposed to CORT out-of-phase re-adapt slower to a shift in the LD cycle.

a) Scheme of the gestational intervention, pregnant mice received subcutaneous (s.c.) injections of corticosterone (CORT) at Zeitgeber time 0 (ZT0, out-of-phase) or at ZT12 (in-phase) on gestational days (GD) 15 and 16. A naive group of adult mice whose mothers were left undisturbed during their whole pregnancy was included and the effects on the offspring’s re-adaptation after a 6-hour phase advance in the LD cycle (jet lag) were assessed from postnatal day 60–80. b) Representative double-plotted running wheel actograms before and after jet lag, dark phase is shaded in gray. c) Average of the activity onset time before and after the phase-shift, dotted line indicate phase-shift 50 (PS₅₀) defined as the time at which half of the phase shift is completed (naïve offspring (n = 14, 8 males and 6 females), out-of-phase offspring (n = 15, 7 males and 8 females) and in-phase group (n = 16, 8 males and 8 females). d) For PS₅₀ comparisons between groups, data are expressed as mean ± SEM, passed normality test and statistical difference was assessed by one-way ANOVA, F(2,42) = 5.382, p = 0.0083, followed by Holm–Sidak’s multiple-comparison test (naïve vs. out-of-phase, *p = 0.0291, out-of-phase vs. in-phase, *p = 0.0114). e) Representative double immunohistochemistry (GR and GFAP) confocal pictures. The dotted white regions demarcate the SCN and the magnified area is indicated with yellow squares, scalebars correspond to 100 µm, except for the zoomed images, where it is 25 µm. f) Quantification of the expression of GR specifically in SCN astrocytes. Data are expressed as mean ± SEM (n = 8–12/group), passed normality test and statistical difference was assessed by one-way ANOVA, F(2,27) = 9.648, p = 0.0007, followed by Holm–Sidak’s multiple comparison test (naïve vs. out-of-phase, ***p = 0.0005, out-of-phase vs. in-phase, *p = 0.0104). Numerical data can be found in S1 Data file.

https://doi.org/10.1371/journal.pbio.3003870.g001

These data suggest that the exposure to CORT out-of-phase during a critical window of development has long-term consequences on the expression of GR in SCN astrocytes and on the resynchronization during jet lag. While the SCN sensitivity to GCs has been already shown during the perinatal period in rodents [10–12], the adult SCN has been considered resistant to GCs [12–14]. Importantly, the activity/rest patterns are not affected by the injection/manipulation at different times of day suggesting that CORT is the main contributor to the offspring’s phenotype (S2 Fig and S1 Data). To understand our observation in this context, we decided to follow the expression of GR along SCN development at single-cell level.

Single-nuclei transcriptional profiling of the SCN along maturation

We performed single-nuclei RNA sequencing (snRNA-seq) of micro-dissected mouse SCN at three key developmental stages and at maturity. The developmental stages were selected to match milestone events during the maturation of the mouse SCN that were reported in the literature [25,40]: the neurogenic-gliogenic switch (gestational day (GD) 17.5) [41], the peak of astrocyte proliferation (PND2) [42], the maturation of the main photic input pathway, the retinohypothalamic tract (PND10) [26,43,44] and the adult-like mature circuit (PND30) [27,28]. To generate a comprehensive transcriptomic data set of SCN cells, we dissected brains from male and female mice between ZT 3 and 5 (i.e., Zeitgeber time, 3–5 hours after “lights on”) and prepared 250-μm thick coronal sections to precisely identify the SCN under the microscope (Fig 2a, S3a, and S3b). To visualize the transcriptional signature of the SCN during development, we generated an UMAP embedding colored by timepoints (Fig 2b), which shows that the SCN gains transcriptional heterogeneity with maturation. We annotated 13 main clusters based on known molecular markers in the developing anterior hypothalamus and the mature SCN [45–48] (Fig 2c and 2d). These clusters were named according to their association with cell lineages: neuronal (Avp⁺, Vip⁺, Cck⁺, and Nms⁺ mature neurons, early SCN neurons, and extra-SCN neurons), astroglial (immature/ependymal cells and mature astrocytes), oligodendroglial (NG2 cells/oligodendrocyte precursor cells (OPCs) and mature oligodendrocytes), microglial, and endothelial cells. We were unable to identify a small cluster of neuronal cells (79 cells) mainly present at GD17.5, PND2 and PND10 (Figs 2c, 2d, and S5a). Even though it is possible that certain clusters are under represented due to the gene expression levels at the sample collection time point (ZT 3–5), the total number of cells per cluster that we identified at PND30 fits with the expected proportion of cell types in the adult SCN (i.e., a neuron:astrocyte ratio of 3:1, around 5% of NG2 cells/oligodendrocytes and below 2%–3% of endothelial cells and microglia) [27–29,45]. We further validated the presence of cell types corresponding to the main clusters by in situ hybridization on age-matched mouse brain sections. Representative confocal images and snRNA-seq gene expression plots (Fig 2e and 2f) show the main cell clusters: oligodendroglial lineage (Pdgfra, green), neuronal lineage (Syt1, cyan) and astroglial lineage (Aldh1l1, red). These marker genes were chosen to validate the main clusters because they are expressed across all developmental stages.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. Single-nuclei transcriptional profiling of the SCN along maturation.

a) Overview of the experimental workflow. SCN was dissected from 5 male and 5 female fetuses/newborns at three key developmental stages (GD17.5, PND2, PND10) and from adults at PND30 when the SCN reach adult-like maturity at ZT3-5 (3–5 hours after “lights on”). Single-nuclei RNA-sequencing was performed on the pooled extracted nuclei suspensions. b) UMAP embedding colored based on developmental stage. c) Dot plot showing expression of the markers used to identify the main clusters. d) Same UMAP embedding as shown in b but colored based on main cluster annotations using markers in c. e) Representative confocal images of in situ hybridization to detect markers of the main cell clusters (Pdgfra, NG2 cells/OPCs/oligodendrocytes, green; Syt1, Neurons/neuronal progenitors, cyan; Aldh1l1, mature/immature astrocytes and ependymal cells, red). The dotted white regions demarcate the SCN and the yellow squares the four magnified areas; scale bars on top and bottom pictures correspond to 100 µm and 20 µm, respectively. OC, optic chiasm, 3V, third ventricle. f) Same UMAP embedding as in b and d, but colored for expression of Pdgfra (green), Syt1 (cyan), and Aldh1l1 (red). Raw transcriptomic data can be found under GEO accession number GSE240803.

https://doi.org/10.1371/journal.pbio.3003870.g002

Overall, our data show that during maturation, the SCN gradually gains cellular and transcriptional heterogeneity and that distinct transcriptional signatures define developmental stages, corresponding to the previously described temporal window of SCN circuit maturation. Therefore, we used this data set to follow cell type-specific changes of glucocorticoid receptor Gr (Nr3c1) expression along the development of the SCN.

Cell-type-specific changes in Gr expression along SCN maturation

Interestingly, the data displayed in Fig 3a and quantified in Fig 3b, show that Gr is expressed in most clusters at GD17.5 and PND2, but later, at PND10 and PND30, the expression becomes more pronounced in astrocytes. Although, microglia and oligodendrocyte clusters also express the receptor, they make up a small percentage of all SCN cells. In Fig 3b, the second peak corresponding to cells with non-zero expression increases in magnitude (indicating higher number of cells with non-zero Gr expression) and shifts to the right (indicating increased Gr expression) in the case of astrocytes, while it displays an opposite trend in SCN neurons. Interestingly, at PND10 the absolute number of astrocytes expressing Gr exceeds the number of neurons expressing the gene, despite there being roughly three times as many neurons as there are astrocytes in the SCN from this developmental time point on. In comparison, at PND30, while the absolute number of astrocytes expressing Gr is lower than that of neurons, astrocytes express it at higher levels. Density plots also show the expression of Gr in microglia and mature oligodendrocytes clusters (S6a and S6b Fig). Although the expression of Gr during perinatal development was already reported [10–12], here we show that Gr expression is not simply down-regulated as the SCN matures, but rather modulated in a cell type-specific manner. To quantify this observation, we performed a trajectory analysis on our snRNA-seq data, which also indicated an up-regulation of Gr expression in astrocytes along developmental pseudo-time (S6c and S6d Fig). To assess this developmental trend at the protein level, we performed a co-staining of GR with markers of both, mature and immature astrocytes (expressing GFAP and vimentin, VIM) and neurons (expressing the vesicular GABA transporter, VGAT) (S7 Fig). Quantification shows that the number of cells expressing GR co-stained with astrocytic markers tend to increase (Figs 3c and S7a) while those co-stained with VGAT tend to become less frequent with age (Figs 3c and S7b). Although we cannot compare the transcriptional and immunohistochemical data in a quantitative way, the observed developmental changes follow a similar trend with both approaches (Figs 3c and S6e).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Cell-type-specific changes in Gr expression along SCN maturation.

a) Expression of Gr in the snRNA-seq dataset in the same UMAP embedding as Fig 2b across different developmental stages including all cell clusters. Cells across all time points including those that do not expressed are shown in the background in light gray. b) Density plots of the Gr expression in astrocytes and SCN neurons segregated by developmental stage. The percentage (and absolute numbers) of cells with non-zero expression of Gr are indicated for each cell cluster. Note that since the total number of cells was different across the time points, the cell numbers have been normalized to a total of 5,000 cells per developmental time point. c) The percentage of GR⁺ cells co-expressing astrocytic markers (GFAP/VIM⁺ in red) or neuronal marker (VGAT⁺ in cyan) are plotted for each developmental stage as mean ± SEM (n = 4–6 independent samples). Statistical differences were assessed by multi-comparison Mann–Whitney test, * p < 0.05. d) Representative in situ hybridization confocal images showing expression of Syt1 (cyan), Gr (green), and Aldh1l1 (red) of the mouse SCN at PND30. White arrows highlight the co-expression of Gr and Aldh1l1. e) Representative immunohistochemistry confocal images of the SCN (DAPI), for GR (green) in astrocytes (GFAP/VIM⁺; red) of the mouse SCN at PND30. White arrows highlight the co-expression of GR and GFAP/VIM. f) Representative immunohistochemistry confocal images of the SCN (DAPI), for GR (green) in neurons (VGAT; red) of the mouse SCN at PND30. In d–f, the dotted white regions demarcate the SCN and the two yellow squares, the magnified areas. Scalebars correspond to 200 µm, except for the zoomed images, where it is 20 µm. OC, optic chiasm, 3V, third ventricle. Raw transcriptomic data can be found under GEO accession number GSE240803, numerical data can be found in S1 Data file.

https://doi.org/10.1371/journal.pbio.3003870.g003

We then assessed whether the up-regulation of Gr expression in astrocytes was aligned with a greater activation of GR-dependent downstream pathways. Transcripts associated with the GO biological process term Cellular response to glucocorticoid stimulus were maintained at low levels in neurons, while there was a tendency for increased enrichment in astrocytes along SCN maturation (S6f Fig). Overall, our data suggest that throughout development the expression of Gr in the SCN is modulated depending on the cell type. This finding suggests that the SCN could remain responsive to GCs in the adult stage.

Adult SCN astrocytes respond to circulating GCs in vivo

To follow-up on a potential responsiveness to GCs by the adult SCN, we further confirmed by in situ hybridization at PND30 the presence of Gr transcript in astrocytes labeled with Aldh1l1 probes in red (showing co-expression, white arrows). However, we did not observe co-expression with Syt1 probes, a gene exclusively expressed in neurons (Fig 3d). At the protein level, we found a significantly higher percentage of GR⁺ cells co-expressing the astrocytic markers GFAP and VIM (Fig 3e) than co-expressing the neuronal marker VGAT (Fig 3f). To evaluate whether GR was responsive to circadian GC changes in the SCN, we used a proximity ligation assay (PLA) to visualize GR activation in vivo in PND30 mice [49]. HSP90 is a well-known chaperone protein that binds inactive GR in the cytosol, and this interaction can be visualized with PLA (Fig 4a) [49,50]. We reasoned that at ZT0 (Zeitgeber time 0, time of day when the lights were switched on) the GC levels are low (Fig 4b) and the inactive GR would be bound to HSP90 (i.e., maximal interaction generating maximal signal). In contrast, at ZT12 (Zeitgeber time 12, time of day when the lights were switched off) murine GC levels are at their physiological peak (Fig 4b) and active GR would not be bound to HSP90 (i.e., minimal interaction generating minimal signal). We coupled immunofluorescence against GFAP with PLA to quantify GR-HSP90 interaction in GFAP+ astrocytes on brain sections from mice at ZT0 and ZT12. We observed a significant difference in the PLA signal (red puncta, white arrow) detected in SCN GFAP+ astrocytes between ZT0 and ZT12 (Fig 4c and 4d) indicating that GR in SCN astrocytes is activated by circadian GC levels. As a positive control and to validate this approach, we compared PLA signals in the PVN, a component of the HPA axis that strongly responds to GC feedback, finding similar effects (S9a and S9c Fig). To further confirm that GR in the SCN responds to circulating GC changes, we performed the same experiment in brain sections from mice obtained at ZT0, but one hour after subcutaneous injection with either vehicle (VEH, PEG-400 50% in PBS) or corticosterone (CORT, 5 mg/kg of b.w.). We monitored plasma corticosterone levels and found low and high GC levels in the VEH and CORT groups, respectively (Fig 4e). We detected a significantly higher PLA signal in SCN GFAP+ astrocytes at ZT0-VEH when compared to ZT0-CORT (Fig 4f and 4g).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Adult SCN astrocytes respond to circulating GCs in vivo.

a) Schematic showing proximity ligation assay (PLA) setup. b) Corticosterone levels (in ng/mL) in plasma of mice kept in 12-hour:12-hour light:dark (LD) conditions at ZT0 (lights on) and ZT12 (lights off). Data are expressed as mean ± SEM (n = 6 mice/group), passed normality test and statistical difference was assessed by two-tailed T test, t = 3.801, df = 10, **p = 0.0035. c) Proximity ligation assay in the SCN showing the interaction between HSP90 and GR (PLA puncta), in astrocytes (GFAP+) with nuclei counterstained with DAPI. Top: reconstruction of the SCN from pictures taken at 40× magnification (scale bar 50 µm). Bottom: zoom in all channels separately and merged (scale bar 10 µm). The dotted white regions demarcate the SCN and the magnified areas are indicated with yellow squares, arrows highlight PLA puncta. d) Quantification of the number of GR-HSP90 signal (puncta)/GFAP⁺ area at both timepoints. Data are expressed as mean ± SEM (1 section/animal, 11–12 mice/time point), passed normality test and statistical difference was assessed by two-tailed T test, t = 2.969, df = 21, **p = 0.0073. e) Corticosterone levels (in ng/mL) in plasma of mice kept in 12-hour:12-hour light:dark (LD) conditions at ZT0, one hour after subcutaneous injection either with VEH or CORT (5 mg/Kg b.w. at ZT23). Data are expressed as mean ± SEM (n = 6 mice/group), passed normality test and statistical difference was assessed by two-tailed T test, t = 12.35, df = 10, ***p < 0.0001. f) Proximity ligation assay in the SCN showing the interaction between HSP90 and GR (PLA puncta), in astrocytes (GFAP⁺) with nuclei counterstained with DAPI. Top: reconstruction of the SCN from pictures taken at 40× magnification (scale bar 50 µm). Bottom: zoom in all channels separately and merged (scale bar 10 µm). The dotted white regions demarcate the SCN and the magnified areas are indicated with yellow squares, arrows highlight PLA puncta. g) Quantification of the number of GR-HSP90 signal (puncta)/GFAP⁺ area at both timepoints. Data are expressed as mean ± SEM (1 section/animal, 5 mice/group), passed normality test and statistical difference was assessed by two-tailed T test, t = 3.036, df = 8, *p = 0.0162. Numerical data can be found in S1 Data file.

https://doi.org/10.1371/journal.pbio.3003870.g004

Since we also observed the above-described changes in GFAP− areas (S9d and S9e Fig), we could not confirm astrocytic-specific responses with this approach. This could be due to other SCN cells expressing GR that respond to peripheral GCs or, alternatively, to the fact that not all SCN astrocytes express GFAP. We discarded the first option because in our transcriptional data (at PND30), we found that only around 9% of SCN cells express Gr and no astrocytic markers. On the other hand, we also found in our transcriptional data that only around 37% of the cells clustered as SCN astrocytes expressed Gfap (S9g Fig). Therefore, we decided to assess the activation of GR in the SCN by a second approach that allows us to capture a higher proportion of astrocytes. We assessed the percentage of nuclear GR (active) by triple immunohistochemistry combining two antibodies to label SCN astrocytes. The combination of GFAP and SRY-Box transcription factor 9 (SOX9) allowed us to detect more of these cells (S9g Fig). With this approach we found a significantly higher percentage of GR+ nuclei at ZT12 (Fig 5a and 5b) and at ZT0, 1 hour after CORT injection (Fig 5c and 5d) compared to respective controls. On average, around 85% of the detected GR⁺ nuclei co-stained with either GFAP or SOX9 or both astrocytic makers (S9f Fig). These data align well with the PLA suggesting that GR is activated by circulating GCs in the SCN with a major contribution, although not exclusively, from astrocytes.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. SCN astrocytes express GR and respond to circulating GCs in vivo.

a) Quantification of active GR in the nuclei at both ZT0 (low circulating GCs) and ZT12 (high GCs). Data are expressed as mean ± SEM (1 section/animal, n = 6 mice/time point), passed normality test and statistical difference was assessed by two-tailed T test, t = 2.43, df = 10, *p = 0.035. b) Representative confocal pictures of the SCN with a triple immunostaining for GR (green), GFAP (red), and SOX9 (gray). Overview of the SCN from pictures taken at 20× magnification (scale bar 100 µm) and zoomed in (scale bar 25 µm). The dotted white regions demarcate the SCN, and the white squares indicate the zoomed areas. c) Quantification of active GR (in the nuclei) at ZT0 1 hour after subcutaneous injection with either vehicle (VEH, PEG-400 50% in PBS) or corticosterone (CORT 5 mg/kg b.w.) at ZT23. Data are expressed as mean ± SEM (1 section/animal, n = 6 mice/group), passed normality test and statistical difference was assessed by two-tailed T test, t = 3.07, df = 10, *p = 0.012. d) Representative confocal pictures of the SCN with a triple immunostaining for GR (green), GFAP (red), and SOX9 (gray). Overview of the SCN from pictures taken at 20× magnification (scale bar 100 µm) and zoomed in (scale bar 25 µm). The dotted white regions demarcate the SCN, and the white squares indicate the zoomed areas. Numerical data can be found in S1 Data file.

https://doi.org/10.1371/journal.pbio.3003870.g005

Finally, we sought to assess in previously published snRNA-seq data sets, whether the expression of GR related transcripts shows circadian changes in a cell type-specific manner in the adult SCN. We selected two data sets: Morris and colleagues, 2021 [45], generated from PND10 to 12 mouse SCN organotypic explants in culture and collected at two timepoints, circadian time (CT) 7.5 (i.e., when low physiological GC levels are expected) and CT15.5 (i.e., when high GC levels are expected) [45]. Wen and colleagues, 2020 [28], generated from SCN samples obtained from PND56 mouse at different circadian timepoints, at which we analyzed (CT26 and CT50; i.e., at low circulating GC levels, and CT14 and CT38; i.e., at high circulating GC levels). First, we annotated the clusters based on known cell types in the mature SCN (S10a and S10d Fig), we then ran a single-sample gene set enrichment analysis to compare the enrichment of transcripts in each cell type at low- and high-GC conditions (both, expected and present). We could statistically confirm an enrichment of genes associated with the GO molecular function term GR binding in SCN astrocytes, ependymal cells, and in extra-SCN neurons when high GCs levels are expected (S10b Fig) in the Morris and colleagues data set. We also found an enrichment of genes associated with the GO molecular function term GR pathway in SCN astrocytes, ependymal cells and in extra-SCN neurons when circulating GCs are high in the Wen and colleagues data set (S10e Fig). Moreover, we observed a significant enrichment of ligand-receptor (LR) pairs involved in extracellular matrix remodeling (e.g., Angptl4 -> Sdc2-4) and cholesterol uptake (e.g., Apoe -> Scarb1) in SCN astrocytes, when we restricted LRs to GR-regulon member genes (S10c Fig). These data could potentially explain previous observations suggesting that the structural plasticity of SCN astrocytes changes in a circadian manner and it is influenced by circulating GCs [51–55]. Interestingly, the comparative analysis of both data sets suggest that even in the absence of circulating GCs the SCN still expects to receive the signal. This could be a feature of a daily program of the SCN that requires further investigation. Overall, our data show that adult SCN astrocytes can sense circulating GCs through GR activation. Thus, we further explored the effect of GCs on SCN astrocytes.

Calcium signaling is activated by GCs in SCN astrocytes

We monitored calcium transients during CORT treatment in acute SCN slices loaded with the cell-permeable calcium sensor Fluo-4-AM. We imaged fluorescence emission, as a readout of intracellular calcium release, using two-photon imaging and then performed image analysis to evaluate astrocytic calcium events (Fig 6a). All pharmacological treatments were done sequentially by perfusing the solutions through the recording chamber. Baseline (Ctrl) and vehicle (VEH, aCSF with 0.1% DMSO) recordings were done first, followed by 5-min recordings during treatment with: 1 μM CORT, 400 μM GAP26 and a mixture of the two. GAP26 blocks hemichannels as well as intercellular communication through gap junctions specifically containing connexin 43 (Cn43) [56,57]. GAP26 was the inhibitor of choice to prevent astrocyte–astrocyte communication, because Cn43 is expressed in astrocytes in the SCN (S11a Fig) [32,58–60]. Since we found that circulating GCs activate GR in the SCN with an important contribution of astrocytes, we used step-by-step selection criteria to quantify the calcium transients from astrocyte-like regions of interest (ROIs) [61–65]. In a first step, we distinguished Fluo-4-loaded astrocyte-like cells based on their morphology (i.e., cells with small soma and complex radial arborizations) (S11b Fig) [63]. In a second step, we filtered only those Ca²⁺ events longer than 10 s that are typically observed in the astrocytic soma and main processes (Fig 6c and S1 Data) [66]. In a third step, we further filtered events with signal decay longer than the signal rise, a well-described feature of Ca²⁺ events in astrocytes (Fig 6c and S1 video) [63,66–68]. Once all regions of interest were selected, we assessed whether these Ca²⁺ events occurred in synchrony, a unique feature of astrocytic activation. Therefore, faster and sharper events, with a lower decay:rise duration ratio were not included in the analysis and were classified as non-astrocytic events (S11c–S11e Fig and S1 Data). To block a possible indirect astrocytic activation through the connections between the SCN and the PVN, we performed a scalpel cut between both regions, perpendicular to the third ventricle (Figs 6d and S11f–S11h). In Fig 6e, we show a representative image of the localization of the SCN and the SCN–PVN incision in the acute slice imaged in brightfield with infrared illumination.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. Calcium signaling is activated by GCs in SCN astrocytes.

a) Overview of the experimental workflow. Created in BioRender. Administrator, S. (2026) https://BioRender.com/ql4wtok. b) Imaging protocol depicting the sequence of treatments performed to the same slice. Ca²⁺ signal was quantified in Fluo-4-stained SCN slices perfused at 0.5 mL/min with oxygenated aCSF at 37 °C. Pharmacological treatments were done perfusing the solutions to the recording chamber at the following working concentrations (in aCSF): 1 μM CORT, 400 μM GAP26, or the mixture of the two, a 5-min wash period was done in between. Baseline (control, Ctrl) and vehicle (VEH) recordings were done at the beginning and at the end (recovery) and the Z-plane was maintained between recordings. c) Representative image of a typical astrocytic Ca²⁺ event, identified according to the following criteria: a typical astrocytic morphology, a duration of at least 10 s and a signal decay longer than the signal rise. d) Schematic of how the connection between the SCN and the PVN was interrupted with a scalpel cut between both regions and perpendicular to the third ventricle (https://biorender.com/ql4wtok). e) Representative image of the localization of the SCN and the SCN–PVN incision in the acute slice imaged in brightfield with infrared illumination. f) Representative traces of the three conditions (Control, CORT, and CORT + GAP26) with examples of calcium transients. g) Integrated Ca²⁺ response represented as the area under the curve (AUC) (independent experiments, n = 4 mice), data passed normality test and statistical differences were assessed by repeated-measures one-way ANOVA (F(3,9) = 3.941, p = 0.0476) followed by Holm–Sidak’s multiple comparison test (CORT vs. GAP26, *p = 0.052). h) Number of astrocytic Ca²⁺ events is represented (independent experiments, n = 4 mice, in gray) and as mean (in black), data passed normality test and statistical differences were assessed by repeated measures one-way ANOVA (F(3,9) = 9.847, p = 0.0033) followed by Holm–Sidak’s multiple comparison test (Ctrl vs. CORT, *p = 0.022; CORT vs. CORT + GAP26, *p = 0.0451; CORT vs. GAP26, **p = 0.0031). SCN slices from g and h were prepared from mice at ZT0-2 and incubated before and during the recording with regular aCSF. i) Number of astrocytic Ca²⁺ events is represented (independent experiments, n = 4 mice, in gray) and as mean (in black). SCN slices were prepared from mice at ZT0-2 and incubated before and during the recording in aCSF containing actinomycin D (a transcription inhibitor) and cycloheximide (a protein synthesis inhibitor). Data were analyzed by a negative binomial generalized linear model to assess statistical differences between the treatments accounting for ROIs nested within slices (animals), VEH vs. CORT, **p = 0.0027. j) Number of astrocytic Ca²⁺ events is represented (independent experiments, n = 4 mice, in gray) and as mean (in black). SCN slices were prepared from mice at ZT12-14 and incubated before and during the recording in aCSF containing actinomycin D and cycloheximide. Data were analyzed by a negative binomial generalized linear model, VEH vs. CORT, **p = 0.0028, CORT vs. CORT + GAP26, **p = 0.0001. k) Number of astrocytic Ca²⁺ events is represented (independent experiments, n = 3 mice, in gray) and as mean (in black). SCN slices were prepared from mice at ZT0-2 and with CORT-BSA followed by the GR antagonist RU486 in regular aCSF. Data were analyzed by a negative binomial generalized linear model, VEH vs. CORT-BSA, *p = 0.028, CORT-BSA vs. CORT-BSA + RU, ***p = 0.00042 and VEH vs. CORT-BSA + RU, **p = 0.021. l) Representation of all Ca²⁺ transients from astrocytic-like ROIs were followed during the 200 s of recording while perfusing VEH, CORT and CORT + GAP26 to the SCN slice. Heatmaps show a temporal sequence of activation with color indicating the number of ROIs that fired in synchrony. m) Number of synchronous Ca²⁺ events across all treatments. Each gray line represents one independent experiment showing synchronyc events (n = 7) and the mean is shown in black. All treatments were performed on the same slice. Poisson generalized linear mixed-effects model detects a significant increase in synchronic events during CORT treatment (*p = 0.037), the offset model revealed no significant effect of the treatment on the synchronization rate (Wald χ² = 1.00, df = 2, p = 0.608). Numerical data can be found in S1 Data file.

https://doi.org/10.1371/journal.pbio.3003870.g006

Representative traces of the three conditions (Ctrl, CORT and CORT + GAP26) showed that CORT treatment increases the number of astrocyte-like Ca²⁺ events in the SCN depending, at least in part, on astrocyte–astrocyte communication through gap junctions (Fig 6f and S1 video). Applying our analysis pipeline, we quantified the integrated Ca²⁺ response to CORT by calculating the area under the curve (AUC) of all detected events. We did not found a significant effect of CORT compared to baseline control (Fig 6g). The total AUC depends on the number of events as well as the duration and amplitude of each of them. While we did not find a significant difference on the duration and the amplitude (S11i and S11j Fig), we found that CORT significantly increased the number of astrocyte-like Ca²⁺ events (Fig 6h). In four independent experiments (in gray lines experiments and in black the mean), we found that the CORT-induced Ca²⁺ events were significantly reduced when astrocyte–astrocyte communication was blocked by addition of GAP26 (Fig 6h). We also found that the total number of active ROIs tended to decrease when CORT binding to the GR was antagonized by co-incubation with the GR antagonist mifepristone (RU486) and that the SCN–PVN incision in the acute slice tended to reduce the duration and the amplitude of all Ca²⁺ events, suggesting that we have reduced the influence from the PVN with our approach (S11f–S11h Fig).

While genomic effects and slow responses were expected, we observed Ca²⁺ events shortly after CORT was perfused into the recording chamber. Then, we sought to understand better how CORT might act on the SCN astrocytes. Therefore, we performed the same sequence of treatments in slices that were incubated before and during the recording with an aCSF containing actinomycin D (a transcription inhibitor which intercalates into DNA, preventing the unwinding of the DNA double-helix and inhibiting the RNA polymerase activity) and cycloheximide (a translation inhibitor which interferes with the translocation step of translation elongation). Applying the same analysis pipeline we quantified the number of active astrocytes in four independent experiments (in gray lines experiments and in black the mean) and found a similar result: an increase in the number of astrocytic Ca²⁺ events with CORT treatment which was significantly reduced when astrocytic communication was blocked with GAP26 (Figs 6i and S11l). These data suggest that transcription and translation are not necessary for the rapid effects of CORT. Moreover, we performed the same sequence of recordings in slices obtained at ZT12–14 to assess CORT effects at a more physiologically relevant time of day. Again, applying the same analysis pipeline we found a similar result (Figs 6j and S11m). Interestingly, we observed that at ZT12–14 the effect of CORT was stronger (i.e., yielding a higher frequency of Ca²⁺ events/min) and more susceptible to the treatment with GAP26. These findings are in line with previous data suggesting that, in mice, astrocytes might be more active during the dark phase [32], when circulating GCs are higher [9] and that the structural plasticity of SCN astrocytes including the expression of Cn43 changes in a circadian manner [51–53,55,58]. These data are also in line with our comparative analysis of published RNA-seq data sets (S10 Fig), strenghthening the idea that adult SCN astrocytes can sense circadian GCs through GR activation. Finally, to check whether CORT might be acting through GR on the cell membrane, we treated the SCN slice with CORT-BSA (corticosterone conjugated with bovine serum albumin) which is unable to enter the cell and found again a similar result: a significant increase in the number of astrocytic Ca²⁺ events with CORT treatment, prevented by the GR antagonist RU486 (Figs 6k and S11n).

In Fig 6l, all Ca²⁺ events from astrocyte-like ROIs were followed during 200 s while perfusing VEH, CORT, and CORT + GAP26 to the SCN slice. Heatmaps show a temporal sequence of activation with the color indicating the number of ROIs that fired in synchrony from a representative experiment. Synchronic events were identified when ≥2 astrocytes exhibited Ca²⁺ events within a 3-second window. When additional active astrocytes were detected in a subsequent 3 s interval, the window expanded iteratively until no new participants were found. Fig 6m shows that from the 8 slices (experiments shown in Fig 6i and 6j), synchronic events were detected in 7 of them. Slices treated with CORT showed significantly more synchronic events (mean ± SEM = 2.71 ± 1.14) than VEH (0.86 ± 0.55, p = 0.037). Co-perfusion of GAP26 alongside CORT attenuated this effect (1.14 ± 0.70). However, a Poisson generalized linear mixed-effects model (GLMM) analysis revealed that the rate of synchronized events per total events did not differ significantly across conditions (χ² = 1.00, df = 2, p = 0.608), suggesting that the increase in synchronized events likely reflects the overall CORT-driven rise in astrocytic calcium activity rather than a specific enhancement of network synchronization. Considering that a key component of astrocytic signaling is the propagation of Ca²⁺ waves within the astroglial syncytia expanding through neighboring cells, the above-mentioned observation further validates the effect of GAP26 as a blocker of astrocyte–astrocyte communication.

Overall, our data suggest that CORT activates SCN astrocytes, increasing the number of Ca²⁺ events in these cells independently of the PVN input. Interestingly, when we specifically blocked a key feature of astrocytic activity, the propagation of Ca²⁺ waves, with GAP26, CORT induced less and more isolated events. Our data also suggest that the effect of CORT is stronger at ZT12, providing a physiologicaly relevant information to the observed effect. Additionally, CORT induced rapid events in the SCN that are independent of transcriptional and translational programs and seem to be due to the activation of GR located at the cell membrane.

Ethics statement

All experiments in mice were ethically approved by the Committee on Animal Health and Care of the Government of Schleswig-Holstein, Germany (protocol approval number 4_2017_08_30_Oster and 86-10/22_Oster) and the Ethical Committee of the Basque Country University, Spain (protocol approval number M20_2023_064_Astiz). Experiments were performed according to international guidelines (2010/63/EU) on the ethical use of animals.

Animals

All mice used in this study were kept in light (L):dark (D) 12:12 conditions in which light intensity was fixed at ca. 300 lux with abrupt LD/DL transitions. Temperature was maintained at 22 ± 2 °C and at a relative humidity of 60 ± 5% with ad libitum access to standard chow food and water.

For the prenatal manipulation, C57BL6/J males and females were purchased from Janvier labs, France. Females of 2–4 months of age were group-housed, estrous cycles were synchronized by adding male bedding to the females’ cage. After 4 days (when most of them reach proestrus) females were individually housed overnight (ON) in presence of a fertile and experienced male. The next day was considered GD0.5 if a vaginal plug was present. Females were separated, single-housed and left undisturbed until gestational day (GD) 12.5 when successful pregnancy was confirmed by weight increase. On GD 15.5 and 16.5, mice were subcutaneously injected with corticosterone (CORT, S1 Table) at a concentration of 5 mg/kg body weight (dissolved in polyethyleneglycol-400 (PEG-400) 50% in PBS, S1 Table). These time points during mouse pregnancy were selected to mimic the time window in which human fetuses receive antenatal GCs (i.e., when an elevated risk of premature birth is detected between gestational week 24–34). The dose was selected to override the capacity of the placenta to inactivate maternal CORT. One hour after injection, the levels of CORT in the fetal blood increased to a concentration comparable to the maternal GCs circadian peak [10]. Both days the injections were performed at zeitgeber (ZT) 0, when the lights were switched on in the animal facility (i.e.,: 6 a.m., out of phase with the maternal physiological peak) or at ZT12, when the lights were switched off (i.e.,: 6 p.m, in phase with the maternal physiological peak). The naïve group was generated from females that were undisturbed during pregnancy with the only intention of comparing the effect of the manipulation timing to an undisturbed condition. Between postnatal day (PND) 60–80, a jet-lag experiment (i.e.,: 6-hour phase advance in the LD cycle) was performed in CORT exposed and naive offspring to assess the long-term effect of the prenatal intervention in the response of the circadian system. Brain samples of these mice were collected at ZT0 and ZT12 by the end of the behavioral assessment for immunohistochemistry.

For the snRNA-seq experiments and all validation experiments, the same protocol was followed to generate the samples from pregnant mice. After mating, females were separated, single-housed and left undisturbed until the appropriate GD/PND to obtain the samples between ZT3 and 5 (i.e., 3–5 hours after lights were switched on) to prevent an influence of the collection time on the expression of clock or clock-controlled genes.

For the proximity ligation assay and the characterization of GR activation by immunohistochemistry, male and female C57BL/6J mice at PND30 were used. Brains and plasma were collected at ZT0 and ZT12 and at ZT0 one hour after injecting them subcutaneously with either vehicle (PEG-400 50% in PBS) or corticosterone (CORT 5 mg/kg b.w. PEG-400 50% in PBS) at ZT23.

For the calcium imaging experiment, male and female C57BL/6J mice at PND 20–27 were used for the preparation of acute SCN slices between ZT0–2 and between ZT12–14.

Jet lag experiments

Circadian locomotor activity was assessed using running wheels in LD conditions (300 Lux) with an abrupt LD/DL transition. Mice were housed in individual cages, wheel revolutions were counted every minute (n = 14–16/group). On day 8, the dark phase was advanced for 6 hs while recording the activity. Data collection and analysis were conducted using ClockLab software. The average of the activity onset time before and after the phase shift was calculated to quantify the time at which half of the phase shift was completed (PS₅₀). After the behavioral analysis mice were euthanized at ZT0 and ZT12. Brain sections from ZT12 mice were used to assess the expression of GR in SCN astrocytes by immunohistochemistry (see below). In all the groups, both males and females were included (S1 Fig).

Samples for snRNA-seq and validation

To reduce sex bias and litter effects, SCNs were collected from only 1 male and 1 female fetus/pup per mother of at least 5 different mothers. Then, 10 samples were pooled per developmental stage. Brains were dissected and sliced in 250 μm-thick coronal sections with a vibratome in ice-cold HBSS 1× (S1 Table). GD17.5 and PND2 brains were placed in a block of low-melting agarose 4% (S1 Table) and PND10 and 30 brains were glued directly on the vibratome platform. At least 3 sections were placed on RNase-free glass slides, incubated for 2 min with a nuclei fluorescent marker (DAPI 300 nM, S1 Table) and observed under the microscope. The SCN from the medial section was chosen, dissected with scalpel under the microscope and snap frozen in liquid nitrogen (S3 Fig). For validation experiments (in situ hybridization and immunohistochemistry), brains from fetuses (GD17.5), pups (PND2 and PND10) and adult mice (PND30) from the same litters were collected and immediately frozen in OCT embedding matrix (S1 Table) on dry ice. Fetal sex was determined by PCR as previously described to pool samples from 5 males and 5 females [88]. Briefly, DNA was isolated from the tail by incubating 1 hour at 55 °C with shaking in 20 µL of 50 mM Tris (pH 8), 2 mM NaCl, 10 mM EDTA, 1% SDS containing 0.5 mg/mL of proteinase K. After 1:10 dilution with water, proteinase K was inhibited by incubation for 10 min at 95 °C. One µL of DNA was used in 20 µL of PCR reaction containing 1× ammonium buffer, dNTPs (S1 Table), MgCl₂, and Taq polymerase (S1 Table). PCR was run as follows 10 min at 94 °C, 33 cycles of (40 s at 94 °C, 60 s at 50 °C, 60 s at 72 °C), and 5 min 72 °C. The presence of IL-3 indicated females (544 bp) and the presence of IL-3 (544 bp) and SRY (402 bp) indicated males. The following primers were used at a concentration of 20 µM IL-3 (5′-GGGACTCCAAGCTTCAAT- 3′ and 5′-TGGAGGAGGAAGAAAAGCAA- 3′) and SRY (5′-TGGGACTGGTGACAATTGTC- 3′ and 5′-GAGTACAGGTGTGCAGCTCT- 3′). In newborns and adults, sex was determined by visually inspecting the ano-genital distance.

snRNA-seq

Nuclei were isolated from fresh frozen SCN tissue in a total of 6 batches corresponding to the respective developmental timepoints (10 fetuses/pups per developmental stage) in ice cold lysis buffer (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl₂, 0.1% IGEPAL CA-630 (S1 Table), 1% Recombinant RNase Inhibitor (20 U/µL, S1 Table), 2% BSA). For the two later timepoints (PND10 and PND30) library preparation and sequencing was repeated to increase the number of cells recovered in the first run and to confirm the absence of batch effects (S4 Fig). After 10 min incubation on ice with intermediate mixing by pipette, the suspension was strained through a 20 µm cell strainer, pelleted (4 °C, 500g, 5 min), resuspended in ice-cold nuclei suspension buffer (10 mM Tris-HCl, pH7.4, 10 mM NaCl, 3 mM MgCl₂, 1% Recombinant RNase Inhibitor (20 U/µL, S1 Table), 2% BSA) and quantified in a Neubauer chamber with trypan blue staining followed by snRNA-seq library preparation using Chromium Next GEM Single Cell 3′ Kit v3.1 chemistry (S1 Table) on a 10× Genomics device according to manufacturer’s recommendations. Libraries were quantified, and library fragment size distribution was determined using a Qubit 1× dsDNA HS assay and an Agilent High Sensitivity DNA Kit (S1 Table), respectively. Libraries were multiplexed and sequenced using P2 and P3 (100 Cycles) reagents on a NextSeq 2000 device (Illumina, S1 Table). Sequencing data was demultiplexed and converted to FASTQ format using bcl2fastq2 v2.20 (Illumina) and count matrices were generated with Cellranger v5.0.1 (10× Genomics) using the mouse reference transcriptome mm10-2020-A (10× Genomics). The rest of the analysis was performed using Seurat v4 [89] and standard R packages. Custom filtering at the cell level was performed with a minimum of 750 UMI counts and between 500 and 10,000 distinct features (genes), a maximum of 2.5% mitochondrial and 3% ribosomal reads, as well as a doublet score (scrublet-0.2.3 [90]) below 0.15. Cluster annotation was performed based on differentially expressed marker genes from published mouse anterior hypothalamus development and SCN data (S5a Fig) [27,28,46–48,91,92]. Single-nuclei suspensions resulted in a total of around 8,000 nuclei per snRNA-seq library preparation as determined by nuclei counting. The final sequencing output contained a median of 4,550 transcripts (unique molecular identifiers) and 2,326 distinct genes per nucleus after filtering out poorly sequenced nuclei, nuclei with abnormally high mitochondrial and ribosomal gene counts, and nuclei predicted to be doublets. The final filtered and merged dataset comprised a total of 22,796 high-quality single nuclei across the four stages defining a developmental trajectory. To assess whether the dissection protocol introduced a contamination with extra-SCN cells in our dataset, we implemented a bioinformatic strategy similar to others [45,46]. To classify nuclei as extra-SCN, we focused on marker genes expressed in the preoptic area (rostral to SCN), arcuate nucleus (caudal to SCN) and paraventricular nucleus (dorsal to SCN) at each developmental stage. Since the SCN neurons express particularly high levels of genes from the molecular clock machinery, we further confirmed that cells classified as extra-SCN neurons in our data set express low levels of clock genes (S5b Fig). It is interesting to note that the cluster we annotated as early SCN neurons (mainly present at GD17.5 and PND2, Fig 2c and 2d), showed expression of more than one marker defining different adult SCN neuronal populations (i.e., these are Vip⁺, Avp⁺, Nms⁺, Cck⁺, Grp⁺). These cells likely share the developmental origin but might not be fully mature yet. In support of this interpretation, the cluster we annotated as early SCN neurons showed low levels of clock gene expression (except for Rora and Rorb) as it has been described during the perinatal period (S5b Fig) [40,91,93]. The pseudotime trajectory along developmental time in the astroglial subset was manually selected to follow along the astrocyte and not the ependymal developmental trajectory (S6c and S6d Fig). To monitor cellular response to GC stimulus over developmental stages, single-sample Gene Set Enrichment Analysis (ssGSEA) was applied to snRNA-seq data using the escape R-package (https://doi.org/10.18129/B9.bioc.escape). The msigdbr and getGeneSets functions were used to fetch and filter the corresponding Ontology gene sets (C5) of Mus musculus from the MSigDB [94,95]. EnrichIt with default parameters, except for using 10,000 groups and variable number of cores, was performed on the seurat-object (S6f Fig). To follow Gr expression dynamics during development, trajectory analyses were performed for neuronal (excluding extra-SCN and unidentified neuron clusters) and astroglial subsets across developmental stages using monocle3 [96,97].

snRNA-seq validation

To validate snRNA-seq data and assess regional distribution we performed in situ hybridization and immunohistochemistry. Each reaction was performed in three sections per mouse and a total of 4–6 different mice per time point of both sexes. Both validations were performed on 12 µm-thick cryosections, fixed with PFA 4% in PBS 1× for 20 min at RT.

In situ hybridization was performed using RNAscope Multiplex Fluorescent Assay (S1 Table) on brain tissue sections of mice at the same developmental stages (GD17.5, PND2, PND10 and PND30). The targets, negative and positive controls probes (representative pictures in S8a and S8b Fig) were hybridized following manufacturer’s instructions. This method allowed us to visualize three different RNAs simultaneously to validate the identity of the main cell clusters identified in the SCN throughout development. We used Pdgfra-C1 to label oligodendrocytes and NG2 cells, Syt1-C2 to label neurons and Aldh1L1-C3 to label astrocytes and ependymal cells (Fig 2e). To validate the presence of Gr we used Gr-C1 to label the receptor, Syt1-C2 to label neurons and Aldh1L1-C3 to label astrocytes (Fig 3d) in sections of PND30 mice. Slices were counterstained with a DAPI solution provided with the kit and mounted with Prolong Gold mounting media (S1 Table).

Immunohistochemistry was performed after blocking for 1 hour at RT with normal goat serum 4% and Triton X-100,0.4% followed by incubation with rabbit anti-GR (1:200) (S1 Table) for 2 or 1 days at 4 °C, respectively. In parallel, the negative control was incubated with blocking solution (representative picture in S8c Fig). After 5 washes of 5 min with TBS 1×, we incubated with mouse anti-GFAP (1:200), chicken anti-Vimentin (1:500), or mouse anti-Vgat (1:200) (S1 Table) or blocking solution in case of the negative control ON at 4 °C. Next day, after 5 washes of 5 min with TBS 1×, all slices were incubated with a mix of the three secondary antibodies (all in 1:500 dilution), a-rabbit Alexa 555, a-mouse Alexa 488 and a-chicken FITC (S1 Table) in TBS 1× solution for 2 hours at RT in a dark chamber. Slices were washed, counterstained with DAPI (300 nM in PBS) for 5 min and mounted with Prolong Gold media (S1 Table). Quantification of the percentage of cells expressing GR within the astrocytes (GFAP/VIM+) and GR within neurons (VGAT+) was assessed manually by an experimenter who was blinded to the identity of the pictures using the counting tool of ImageJ software (NIH, Bethesda, Maryland, USA). In 40× zoomed pictures, total nuclei were counted in the DAPI picture and total GR+ cells in the green channel. In the red and green merged image, the total GR+ cells that were also GFAP/VIM+ or VGAT+ were counted. For each time point, pictures from 4 to 6 mice were analyzed.

Confocal imaging

Images at different magnifications were acquired using a Leica Stellaris5 confocal microscope (Leica Microsystems CMS GmbH, Germany). A HC PL APO 20×/0.75 Air CS2 and PL APO 40×/1.30 Oil CS2 objectives were used, with the latter also being used for images taken with optical zoom. The microscope features an AOBS device for fluorescence emission detection allowing free tuning and high-speed switching. The LAS X software, version 4.5.0.25531 as well as the open version (3.3.0.16799-Leica Microsystems CMS GmbH, Germany) were used for the analysis of the images. ImageJ software (NIH, Bethesda, Maryland, USA) was used to create the Z-stack and the merge images. Raw images and metadata are available upon request.

Corticosterone in plasma

Corticosterone was measured in plasma by ELISA (S1 Table) according to the manufacturer’s instructions. Samples were collected in EDTA-coated tubes and plasma isolated by centrifugation at 240g, 20 min, 4 °C. Corticosterone concentration was assessed in duplicates from mouse plasma obtained at ZT0 and ZT12 and at ZT0 one hour after injection with VEH or CORT.

Proximity ligation assay (PLA)

PLA was performed and quantified as previously described [49]. Five µm paraffin-embedded coronal sections containing the SCN and the PVN as a positive control (S9a and S9c Fig) were gradually rehydrated through incubation in graded alcohols and antigen retrieved in citrate buffer pH 6 in a 700W microwave. Alternatively, 12 µm fresh frozen sections were fixed in 4% PFA for 20 min at RT, and antigen retrieved in citrate buffer pH 6 in a 700W microwave. Thereafter, samples were blocked in PLA blocking buffer (S1 Table) for 1 hour at 37 °C, and subsequently incubated overnight with the primary antibodies anti-GR, anti-HSP90, and anti-GFAP (S1 Table) at 4 °C in Tris buffer containing 0.05% Tween-20 (TBST). The next day the samples were washed and incubated with secondary conjugates (S1 Table) ON at 4 °C. Samples were then washed and incubated with ligation solution (S1 Table) for 2 hours at 37 °C and washed again in TBST. The negative controls were run omitting either ligase, PLUS secondary probe, or MINUS secondary probe as appropriate (representative picture in S9b Fig). Finally, sections were incubated at 37 °C in polymerization solution (S1 Table) for 4 hours and then washed with TBS. Samples were counterstained by incubation with DAPI in TBS and secondary antibodies (S1 Table) for 30 min at RT. Sections were mounted with Fluorsave (S1 Table) and stored in the dark at 4 °C until imaging on a Leica Stellaris5 confocal microscope (Leica Microsystems CMS GmbH, Germany). Images were quantified automatically with ImageJ, PLA channel images were background-subtracted and thresholded. Puncta within the GFAP+ and GFAP− masked area of the SCN that were between 3 and 25 pixels in size were automatically counted as positive and averaged per animal and time point. We assessed n = 11–12 mice at each time point in control conditions and n = 5 mice in VEH or CORT injected conditions.

Immunohistochemistry to assess GR activation by circulating GCs

Five µm paraffin-embedded coronal sections containing the SCN were gradually rehydrated through incubation in graded alcohols and antigen retrieved in citrate buffer pH 6 in a 700W microwave. Slides were blocked for 1 hour at RT with normal goat serum 4% and Triton X-100,0.4% followed by incubation with rabbit anti-GR (1:200) and mouse anti-GFAP (1:200) (S1 Table) for 2 or 1 days at 4 °C, respectively. In parallel, the negative control was incubated with blocking solution (representative picture in S8c Fig). After 5 washes of 5 min with TBS 1× all slices were incubated with a mix of the two secondary antibodies (in 1:500 dilution), a-rabbit Alexa 555 and a-mouse Alexa 488 (S1 Table) in TBS 1× solution for 2 hours at RT in a dark chamber. Slices were washed (5 times, 5 min) and incubated with normal rabbit serum 5% (S1 Table) and Triton X-100,0.4% for 1 hour at RT in a dark chamber and followed by incubation with rabbit anti-SOX9 conjugated with Alexa 647 (1:100) (S1 Table) for 2 days at 4 °C. After 5 washes of 5 min with TBS 1× slices were counterstained with DAPI (300 nM in PBS) for 5 min and mounted with Prolong Gold media (S1 Table). Quantification of the percentage of GR+ nuclei (Fig 5a and 5c) and GR+ nuclei co-stained either with GFAP, with SOX9 or with both (S9f Fig) was assessed by an experimenter who was blinded to the identity of the pictures using the counting tool of ImageJ software (NIH, Bethesda, Maryland, USA). In 40× zoomed pictures, total nuclei were counted in the DAPI picture and total GR+ cells in the green channel. In the red and green merged image, the total GR+ cells that were also GFAP+ or in the red and gray merged image total GR+ cells that were also SOX9+ were counted. For each time point, pictures from 6 mice were analyzed.

In silico validation of GR binding and activation in SCN astrocytes

Published data (Morris and colleagues, 2021 [45]) obtained from PND10 to 12 mouse SCN organotypic explants was used to compare gene expression profiles under conditions of expected high (CT15.5) and expected low (CT7.5) GCs. We also used another data set (Wen and colleagues, 2020 [28]) generated from PND56 mouse SCN obtained at different circadian timepoints, and we analyzed CT26 and CT50 (when circulating GCs are low) and CT14 and CT38 (when circulating GCs are high) [28]. Fastq files were downloaded and processed in the same fashion as described in Materials and methods, including cluster annotation by published identity markers. The dataset was filtered to only contain neurons (SCN and extra-SCN) and glial cell clusters and ssGSEA was performed as described above. Single-sided Wilcoxon test (rstatix::wilcox_test) with default parameters was applied to statistically compare scores between cells with high and low expected/circulating GCs (S10a–S10e Fig). Additionally, differential ligand receptor analysis was performed using the Liana+ v1.0.4 package in python 3.11.7 [98]. Briefly, differentially expressed genes were identified between samples collected at CTs when GC were expected to be high and CTs when GC were expected to be low in a pseudo-bulk manner per cell type using get_pseudobulk function in the python version of decoupler v1.5.0, followed by DeseqStats function in pydeseq2 v0.4.4. Ligand-receptor pairs containing these differentially expressed genes were identified using the multi.df_to_lr function, with “mouseconsensus” as the ligand-receptor database. These interactions were further filtered to only keep the ones where either the ligand or the receptor was a regulatory target of GR and the expression changed in the direction expected based on the mode of regulation (i.e., activation or repression). The mouse regulon data for GR was obtained using the CollecTRI database [98] using the decoupleR::get_collectri function with split complexes parameter set to false (S10c Fig).

Two-photon calcium imaging

Since the time-of-day can affect the magnitude of the calcium signals Ca²⁺, we prepared all the acute slices from animals euthanized during the early light phase (ZT0–2) or the early dark phase (ZT12–14). Brains were rapidly removed from the skull and coronal sections (250 µm thick) containing the SCN were prepared in a Leica VT1200S vibratome. Slicing was performed in ice-cold low-Na⁺ low-Ca²⁺ high Mg²⁺ artificial cerebrospinal fluid (aCSF) solution gassed with carbogen (95% O₂/5% CO₂) and containing (in mM): 189 sucrose, 3 KCl, 1.25 KH₂PO₄, 5 MgSO₄·7H₂O, 26 NaHCO₃, 10 glucose and 0.1 CaCl₂·2H₂O (pH 7.3–7.4). To interrupt the connection between the SCN and the PVN, a cut between both regions perpendicular to the third ventricle was performed with a scalpel under the magnifying glass (Fig 6d and 6e). Slices were then transferred to recovery aCSF containing (in mM): 124 NaCl, 2.7 KCl, 1.25 KH₂PO₄, 2 MgSO₄·7H₂O, 26 NaHCO₃, 10 glucose, 2 CaCl₂·2H₂O and 0.4 ascorbic acid (pH 7.3–7.4) gassed with carbogen and containing the cell-permeant calcium fluorescent dye Fluo-4-AM (S1 Table) at 2 μM in 0.01% pluronic (S1 Table). Fluo-4 staining was performed in the dark for 1 hour at RT. Slices were transferred to oxygenated recovery aCSF and allowed to wash for at least 1 hour. For those experiments assessing CORT effects independent of transcriptional or post-transcriptional events, we supplemented the aCSF with actinomycin D 5 μM (S1 Table) and cycloheximide 10 μM (S1 Table). Slices were then transferred to an ultra-quiet imaging chamber (RC-27D, Warner Instruments), perfused at 0.5 mL/min with oxygenated aCSF at 37 °C for imaging in a Femto-2D two-photon microscope (Femtonics) equipped with a Ti:Sapphire MaiTai DeepSee laser (Spectra Physics) tuned at 820 nm. To locate the SCN and check the overall quality of the slice, slices were first imaged in brightfield with infrared illumination and a standard camera. Only one Fluo-4-stained SCN was imaged per animal with 2P-excitation between 30 and 60 µm from the surface as single plane (1 μm). The number of slices per experiment are indicated in the figure legend. Images were acquired with the MES software (Femtonics) at a speed of 1 Hz, 480 × 480 pixels (444 nm/pixel), a 520/60 filter and a water immersion 20× objective (NA 1.0, Olympus). Pharmacological treatments were perfused to the recording chamber at the following working concentrations (in aCSF): 1 μM CORT and CORT-BSA (S1 Table), 1 μM a glucocorticoid receptor antagonist (RU486, S1 Table), 400 μM GAP26 (S1 Table), a Connexin 43 mimetic gap junction inhibitor. Baseline (aCSF) and vehicle (aCSF + 0.1% DMSO) recordings were performed before the sequence of pharmacological treatments. Each recording lasted for 5 min, the Z-plane was maintained between recordings (Fig 6a and 6b).

Calcium imaging processing and analysis

Images were cropped in FIJI [99], drift corrected with EZCalcium in MATLAB [100] and denoised with the deep learning-based software SUPPORT [101]. Active regions of interest (ROIs) were either computed in Aqua2 [102], or outlined using the dF/F0 images computed in ImageJ software. The resulting ROIs were analyzed with the program Calsee in MATLAB [103] to quantify calcium transients. Event detection was determined when the fluorescence signal (dF/F0) was at least three times the standard deviation of the baseline, calculated with a moving threshold of 10 frames. To assess calcium transients from astrocytes we used a step-by-step criterion: In a first step, we detected Fluo-4-loaded astrocyte-like ROIs (i.e., small soma with complex radial arborizations) based on their morphology (S11b Fig). In a second step, we filtered those with Ca²⁺ events longer than 10 s (Figs 6c and S11c). In a third step, we further filter for those events with signal decay longer than the signal rise (Fig 6c), a feature of astrocytic events in the soma and the main processes. All events with less than 0.1 dF/F0 were considered false positives, except in those experiments assessing CORT effects independent of transcriptional or post-transcriptional where the amplitude threshold was 0.05. These filtered ROIs outlined astrocytes with visible events, and thus represented the number of active astrocytes within the region during the recording period. In addition, we also computed for the area under the curve as an integrated metric of the response of each treatment. To analyze the presence of synchronic Ca²⁺ events, we created binary raster tables of individual astrocyte Ca²⁺ events in MATLAB. To this aim, we used an expanding temporal window: an initial 3-s window identified timepoints where ≥2 astrocytes exhibited Ca²⁺ events; when additional active astrocytes were detected in a subsequent 3 s interval, the window expanded iteratively until no new participants were found. The “level” of the synchronic event (number of ROIs in sync) was defined as the total number of co-active ROIs within the expanded window and was assigned to the time point of the first detected event. This approach captures the full temporal extent and participation level of coordinated astrocytic Ca²⁺ episodes while preventing fragmentation of sustained synchronized events into multiple lower-level detections.

Statistics

Immunohistochemistry and in situ hybridization to validate snRNA-seq data were performed on three SCN sections per mouse and at least three different mice per time point of both sexes; only representative pictures are included. To assess statistical differences between two groups we used two-tailed T test and between three or more groups, we used one-way ANOVA followed Holm–Sidak multi-comparison test, after confirming normality by Shapiro–Wilk test. To assess statistical differences between three or more groups and two variables, we used two-way ANOVA followed Holm–Sidak multi-comparison test, after confirming normality by Shapiro–Wilk test. If normality was not confirmed, we used the corresponding non-parametric test. A negative binomial generalized linear model was used to assess statistical differences between treatments in Ca²⁺ events/min, this analysis accounts for ROIs nested within slices (animals) avoiding pseudoreplication. To assess synchronized Ca²⁺ events we used a Poisson GLMM with experiment (slice) as a random intercept, accounting for the repeated-measures design in which the three treatment conditions were applied sequentially to the same slice. To assess whether changes in synchronized events were independent of changes in total event frequency, we fitted an additional Poisson GLMM with log(total events) as an offset and experiment as a random intercept, effectively modeling the rate of synchronized events per total event across conditions. In all cases, the tests used and the results of the test as well as the size of the experimental groups are indicated in the figure captions. P values <0.05 were considered statistically significant. FDR adjusted p values <0.05 were considered statistically significant.