Animals<\/p>\n
Animal care and surgical procedures were approved and overseen by the Harvard University Institutional Animal Care and Use Committee and the Harvard Center for Comparative Medicine or the Institutional Animal Care and Use Committee at Boston Children\u2019s Hospital. The following mouse lines were used: C57BL\/6 (Jackson Laboratory, 000664), GRfl\/fl (Jackson Laboratory, 021021), Vglut1cre (Jackson Laboratory, 023527), Vgatcre (Jackson Laboratory, 028862), Olig2cre (Jackson Laboratory, 025567), Sun1-GFPfl\/fl (Jackson Laboratory, 021039), and PvalbFlpO (Jackson Laboratory, 022730). Mice were maintained in a reverse 12\u2009h light:12\u2009h dark cycle (22:00\u201310:00) in a temperature- and humidity-controlled environment and provided food and water ad libitum. For DR conditions, mice were maintained in chronic darkness until collection. To control for circadian rhythm and to prevent light stimulation of DR mice, mice were collected under red light between 10:00\u201312:00 (Zeitgeiber time 12\u201314\u2009h) for all genomic, imaging, ELISA and LC\u2013MS experiments. Male and female mice were randomly assigned to experimental groups and used in similar proportions. For the developmental genomic experiments, mice were collected between P7 and P35. Details of animal age and sex are indicated in each protocol section. For imaging experiments, mice were collected at P21, P28, P35 and 8\u201312-weeks of age (adult). For ELISA and LC\u2013MS experiments, serum samples were collected at P10, P14, P21 and P28. For electrophysiology experiments, mice were patched between P33 and P35. For CORT treatment experiments, DR C57BL\/6 mice received either 10\u2009mg\u2009kg\u22121 CORT or corn oil (vehicle) at P14 via intraperitoneal injection under red light.<\/p>\n
Cell lines<\/p>\n
NIH\/3T3 cells (ATCC, CRL-1658) and HEK293T cells (ATCC, CRL-3216) were maintained in standard DMEM supplemented with 10% CCS, sodium pyruvate, MEM non-essential amino acids, and penicillin\/streptomycin. Before GR CUT&RUN, 3T3 cells were grown in phenol-red-free DMEM containing 10% charcoal-stripped FBS for 48\u2009h and then treated with 200\u2009nM dexamethasone or vehicle (0.01% ethanol) for 2\u2009h. Before Flag immunoprecipitations, HEK293T cells were grown in phenol-red-free DMEM containing 10% charcoal-stripped FBS for 24\u2009h and then treated with 200\u2009nM dexamethasone for 2\u2009h.<\/p>\n
Serum collection<\/p>\n
Mice were anaesthetized with 10\u2009mg\u2009ml\u22121 ketamine and 1\u2009mg\u2009ml\u22121 xylazine in PBS by intraperitoneal injection for terminal blood draws. Blood was collected from the left atrium of the heart with an insulin syringe. To isolate serum, blood was incubated for 30\u2009min at room temperature and then centrifuged at 1,300g for 10\u2009min at 4\u2009\u00b0C. The supernatant (serum) was isolated and stored at \u221280\u2009\u00b0C prior to corticosterone ELISA and LC\u2013MS.<\/p>\n
Neonatal stereotactic surgery<\/p>\n
P0\u2013P1 wild-type, GRfl\/fl or PVFlp\/Flp;GRfl\/fl mice were anaesthetized on ice and positioned within a neonatal adapter (Stoelting 51625) on a stereotaxic frame (Kopf) in which the temperature of the mouse was maintained around 4\u2009\u00b0C using an ethanol bath containing dry ice. A small patch of skin above the V1 injection site (medial\u2013lateral: \u00b11.8\u2009mm; anterior\u2013posterior: +0.5\u2009mm from lambda; dorsal\u2013ventral: \u22120.1\u2009mm) was cut, and a small opening into the skull was drilled with a 30G needle. AAV (500 nl) was injected at approximately 200 nl\/min, and the pipette was left in place for 5\u2009min after infusion to allow for viral spreading. For genomic and electrophysiology assays required broad infection, mice were injected bilaterally, whereas for histology experiments, mice were injected unilaterally. Following injection, pups were recovered at 37\u2009\u00b0C before being returned to their home cage.<\/p>\n
Adult stereotactic surgery<\/p>\n
For the adult PNN staining and ODP experiments, GRfl\/fl mice were anaesthetized by isoflurane inhalation (3\u20135% induction, 1% maintenance) and positioned within a stereotaxic frame (Kopf). Animal temperature was maintained at 37\u2009\u00b0C with a heat pad. Fur around the scalp area was removed with a shaver and sterilized with three alternating washes with betadine and 70% ethanol. A burr hole was drilled through the skull above V1 (medial\u2013lateral: \u00b12.5\u2009mm; anterior\u2013posterior: \u22123.5\u2009mm from bregma; dorsal\u2013ventral: \u22120.5\u2009mm and \u22120.25\u2009mm). To broadly target superficial and deep cortical layers, 500 nl AAV was injected with a glass pipette at a depth of \u22120.5\u2009mm (~200 nl\/min), and an additional 500 nl was injected at a depth of \u22120.25\u2009mm. After each injection, the pipette was left in place for 5\u2009min to allow for viral spreading. All mice received postoperative analgesic (buprenorphine slow-release formulation, 1\u2009mg\u2009kg\u22121).<\/p>\n
Viral vectors and titre<\/p>\n
All AAVs used were prepared in the Boston Children\u2019s Hospital Viral Core or ordered directly from Addgene. For genomic, postnatal and adult immunofluorescent staining, mIPSC\/mEPSC recordings, and ODP experiments, viruses were densely injected into mice at 5\u2009\u00d7\u2009109 genome copies (gc) per V1 hemisphere. For the astrocyte morphology reconstruction experiment, 2.5\u2009\u00d7\u2009109\u2009gc of eGFP-KASH virus was co-injected with 2.5\u2009\u00d7\u2009109\u2009gc of lck-smV5 virus per V1 hemisphere. For the PV-evoked inhibitory postsynaptic current electrophysiology experiment, 2.5\u2009\u00d7\u2009109 gc of eGFP-CAAX virus was co-injected with 1\u2009\u00d7\u2009109 gc of ChrimsonR-tdTomato virus per V1 hemisphere. The viral vectors and original titres are as follows: AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-KASH-4x6T (this paper, 1.27\u2009\u00d7\u20091014\u2009gc\u2009ml\u22121), AAV2\/5-GfaABC1D-Cre-T2A-eGFP-KASH-4x6T (this paper, 1.24\u2009\u00d7\u20091014\u2009gc\u2009ml\u22121), AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-CAAX-4x6T-WPRE (this paper, 4.37\u2009\u00d7\u20091013\u2009gc\u2009ml\u22121), AAV2\/5-GfaABC1D-Cre-T2A-eGFP-CAAX-4x6T-WPRE (this paper, 6.15\u2009\u00d7\u20091013\u2009gc\u2009ml\u22121), AAV2\/5-GfaABC1D-lck-smV5-4x6T-WPRE (Addgene 196416, 9.26\u2009\u00d7\u20091013\u2009gc\u2009ml\u22121), and AAV2\/8-CAG-FLPX-rc [ChrimsonR-tdTomato] (Addgene 130909, 2.3\u2009\u00d7\u20091013\u2009gc\u2009ml\u22121).<\/p>\n
SHARE-seq<\/p>\n
SHARE-seq was performed on V1 tissue collected from P7, P14, P21, P28 and P35 C57BL\/6 mice raised in DR or NR conditions (n\u2009=\u20094 mice per condition and time point). Mice were deeply anaesthetized with isoflurane, and V1 was dissected. The tissue was dounce homogenized in 1.5\u2009ml of cold homogenization buffer (250\u2009mM sucrose, 25\u2009mM KCl, 5\u2009mM MgCl2, 20\u2009mM tricine-KOH pH 7.8, 1\u00d7 EDTA-free protease inhibitor cocktail (PIC) tablet (Sigma Aldrich 11873580001), 1\u2009mM dithiothreitol (DTT), 0.15\u2009mM spermine, 0.5\u2009mM spermidine, 0.1\u2009U\u2009\u03bcl\u22121 Enzymatics RNase inhibitor (Y9240L), 0.05\u2009U\u2009\u03bcl\u22121 SUPERase inhibitor (AM2696)) with a tight pestle for 20 strokes. The sample was then supplemented with 0.3% IGEPAL CA-630 and further dounced 5 strokes with a tight pestle. Homogenate was filtered through a 40-\u03bcm strainer into a 15\u2009ml conical tube, and 3.5\u2009ml of homogenization buffer and 5\u2009ml of 50% OptiPrep solution (50% OptiPrep (Sigma D1556), 25\u2009mM KCl, 5\u2009mM MgCl2, 20\u2009mM Tricine-KOH pH 7.8, 1\u00d7 PIC, 1\u2009mM DTT, 0.15\u2009mM spermine, 0.5\u2009mM spermidine, 0.1\u2009U\u2009\u03bcl\u22121 Enzymatics RNase inhibitor, 0.05\u2009U\u2009\u03bcl\u22121 SUPERase inhibitor) were added. Tissue lysate was underlaid with 1\u2009ml of 30% OptiPrep solution and 1\u2009ml of 40% OptiPrep solutions. Samples were then centrifuged at 10,000g for 18\u2009min at 4\u2009\u00b0C in an ultracentrifuge. Following centrifugation, approximately 250\u2009\u03bcl of nuclei were collected at the 30%\/40% OptiPrep interface and diluted in NIB-RI buffer (10\u2009mM Tris pH 7.5, 10\u2009mM NaCl, 3\u2009mM MgCl2, 0.1% IGEPAL CA-630, 0.1\u2009U\u2009\u03bcl\u22121 Enzymatics RNase inhibitor, 0.05\u2009U\u2009\u03bcl\u22121 SUPERase inhibitor). The nuclei were then centrifuged at 500g for 10\u2009min at 4\u2009\u00b0C in a swinging-bucket centrifuge and subsequently resuspended in NIB-RI buffer containing 0.2% formaldehyde. Samples were incubated at room temperature for 5\u2009min and then quenched on ice for 5\u2009min with the addition of 125\u2009mM glycine, 40\u2009mM Tris pH 8, and 0.08% bovine serum albumin (BSA). Nuclei were then washed two times in NIB-RI buffer by centrifuging at 500g for 10\u2009min at 4\u2009\u00b0C and replacing the supernatant without disturbing the pellet. After the second wash, the supernatant was removed. Nuclei were flash-frozen in liquid nitrogen and then stored at \u221280\u2009\u00b0C until all samples were collected.<\/p>\n
SHARE-seq libraries were prepared following the original protocol17<\/a>,60<\/a> on two batches of samples, with each batch consisting of 1 female and 1 male for each time point and condition (20 samples\u2009\u00d7\u20092 batches). In brief, approximately 25,000 starting nuclei were used as input for each sample. Nuclei were thawed and resuspended in NIB-RI prior to performing transposition with 6.25\u2009\u03bcl of homemade Tn5 at 37\u2009\u00b0C for 30\u2009min. After one wash with NIB-RI, samples were then reverse transcribed with 25\u2009U\u2009\u03bcl\u22121 Maxima H Minus Reverse Transcriptase. After reverse transcription, nuclei were distributed into two 96-well plates, with approximately 2,000\u20133,000 nuclei per well (9 wells per sample, 10 samples per plate), for 3 consecutive rounds of split-pool barcoding involving simultaneous hybridization of oligonucleotides to tagmented gDNA and cDNA, as previously described17<\/a>,60<\/a>. After the final barcoding round, the nuclei across all wells within each plate were pooled, washed two times with 1\u2009ml of NIB-RI buffer, and resuspended in 80\u2009\u03bcl of NIB-RI. Next, 320\u2009\u03bcl of Ligation mix (1.25\u00d7 T4 ligation buffer, 0.125% IGEPAL CA-630, 25\u2009U\u2009\u03bcl\u22121 T4 ligase, 0.4\u2009U\u2009\u03bcl\u22121 Enzymatics RNase inhibitor, 0.0625\u2009U\u2009\u03bcl\u22121 SUPERase inhibitor) was added, and the samples were shaken at 300\u2009rpm for 30\u2009min at room temperature. Nuclei were diluted with 1\u2009ml of NIB-RI and filtered through a 40-\u03bcm strainer into a new 1.5\u2009ml tube. After a final centrifugation step, nuclei were resuspended in 100\u2009\u03bcl NIB-RI and counted with a haemocytometer. Individual sub-libraries consisting of 3,000\u201310,000 nuclei per plate were prepared, and the samples were brought to a total volume of 50\u2009\u03bcl with NIB-RI. Crosslinks were reversed with the addition of 50\u2009\u03bcl of 2\u00d7 RCB (100\u2009mM Tris pH 8.0, 100\u2009mM NaCl, 0.04% SDS), 2\u2009\u03bcl of 20\u2009mg\u2009ml\u22121 proteinase K, and 1\u2009\u03bcl of SUPERase RI, followed by incubation at 55\u2009\u00b0C for 1\u2009h. At this stage, nuclear lysates were stored at \u221280\u2009\u00b0C for several days before proceeding with the final ATAC and cDNA library preparation.<\/p>\n To each thawed lysate, 5\u2009\u03bcl of 100\u2009mM PMSF (Sigma P7626) was added, and samples were incubated at room temperature for 10\u2009min. Pre-washed MyOne Streptavidin C1 Dynabeads (10\u2009\u03bcl) were added to each sample, followed by 60\u2009min incubation at room temperature on a rotator, to purify the biotinylated cDNA. ATAC and cDNA libraries were prepared from the supernatant and bead-bound fractions, as described previously17<\/a>,60<\/a>. In brief, ATAC libraries were prepared with NEBNext High-Fidelity 2\u00d7 PCR Master Mix (NEB M0541L). After an initial five cycles of amplification, a 5\u2009\u03bcl aliquot of each library was amplified in a separate qPCR reaction to determine the optimal number of PCR cycles to add (one-third of maximal SYBRgreen signal). After amplifying for additional cycles (7\u20138 added cycles per library), libraries were size-selected (0.5\u00d7\u20131.8\u00d7) with AMPure XP beads (Beckman Coulter A63880). cDNA libraries were prepared in 4 steps: template switching reverse transcription, cDNA amplification (5 PCR cycles), cDNA tagmentation, and final library amplification (7 PCR cycles). Amplified cDNA libraries were purified by 0.7\u00d7 AMPure XP beads. SHARE-seq ATAC and cDNA libraries were pooled and sequenced to a depth of approximately 40,000\u201360,000 reads per cell on an Illumina NovaSeq 6000 using a 200-cycle kit with read configuration 50\u2009bp\u2009\u00d7\u200950\u2009bp\u2009\u00d7\u200999\u2009bp\u2009\u00d7\u20098\u2009bp (read 1\u2009\u00d7\u2009read 2\u2009\u00d7\u2009index 1\u2009\u00d7\u2009index 2).<\/p>\n MERFISH<\/p>\n To collect brains at P21 and P35 for MERFISH, C57BL\/6 mice were deeply anaesthetized with isoflurane. Brains were collected and frozen in Optimal Cutting Temperature (OCT) compound using a dry ice ethanol bath and stored at \u221280\u2009\u00b0C until sectioning. A single coronal section containing posterior V1 was collected for each sample at 10\u2009\u03bcm thickness onto MERSCOPE slides (Vizgen 10500001). To minimize batch effects, 4 samples at each time point (1 NR female, 1 NR male, 1 DR female, 1 DR male) were mounted on the same slide. This was accomplished by micro-dissecting one brain hemisphere containing V1 per sample with fine forceps after mounting. Slides were then processed according to the MERSCOPE protocol (Vizgen 10400012), following manufacturer\u2019s instructions. In brief, slides with tissue sections were fixed with 4% paraformaldehyde (PFA), washed 3 times in PBS, and then stored in 70% ethanol at 4\u2009\u00b0C for 5\u20137 days to permeabilize the tissue. Ethanol was aspirated, and slides were washed once in Sample Prep Wash Buffer before incubating with Formamide Wash Buffer at 37\u2009\u00b0C for 30\u2009min. Slides were then incubated with the custom 500-Gene Panel Mix (Vizgen 10400006) for 48\u2009h at 37\u2009\u00b0C. Following probe hybridization, samples were washed with Formamide Wash Buffer for 30\u2009min at 47\u2009\u00b0C two times. After a brief wash in Sample Prep Wash Buffer, samples were embedded in a thin layer of polyacrylamide gel (Gel Embedding Solution) for 90\u2009min at room temperature. Subsequently, slides were incubated in pre-warmed Clearing Premix solution for 24\u201348\u2009h at 37\u2009\u00b0C to clear the tissue, followed by 2 washes in Sample Prep Wash Buffer. DAPI and PolyT Staining Reagent was added, and samples were incubated for 15\u2009min at room temperature on a rocker. Slides were washed once with Formamide Wash Buffer and once with Sample Prep Buffer before being loaded into the MERSCOPE instrument for imaging.<\/p>\n Isolation of nuclei from mouse brain<\/p>\n The following protocol was used to isolate nuclei from early postnatal mice (P14\u2013P28) prior to downstream CUT&RUN, ATAC-seq, bulk RNA-seq and snRNA-seq experiments. Mice were deeply anaesthetized with isoflurane. Sections of 1\u2009mm spanning V1 were collected in an adult mouse brain matrix (Alto) on ice, and V1 was microdissected. For bulk RNA-seq and snRNA-seq experiments, tissue samples were immediately flash-frozen in a dry ice ethanol bath and stored at \u221280\u2009\u00b0C prior to nuclei extraction, whereas tissue samples were processed the day of collection for CUT&RUN and ATAC-seq. After thawing or dissecting samples, V1 tissue was transferred to a 1.5\u2009ml tube containing 350\u2009\u03bcl of cold homogenization buffer (250\u2009mM sucrose, 25\u2009mM KCl, 5\u2009mM MgCl2, 20\u2009mM tricine-KOH pH 7.8 supplemented with 1\u00d7 PIC, 1\u2009mM DTT, 0.15\u2009mM spermine and 0.5\u2009mM spermidine). For RNA experiments, 0.3\u2009U\u2009\u03bcl\u22121 (bulk RNA-seq) or 1\u2009U\u2009\u03bcl\u22121 (snRNA-seq) of Protector RNase inhibitor (Sigma) was added. The tissue was dounced in a 2\u2009ml glass homogenizer (Sigma D8938) with a loose pestle for 20 strokes. Samples were supplemented with 0.3% IGEPAL CA-630 and further dounced for an additional 10 strokes with a tight pestle. Homogenate was filtered through a small 40-\u03bcm strainer (pluriSelect 43-10040-50) into a 1.5\u2009ml tube before diluting 1:1 with homogenization buffer and adding DRAQ5 (Abcam ab108410) at a 1:500 dilution. For CUT&RUN experiments, samples were additionally supplemented with 4\u2009mM EDTA. Nuclei were then sorted by GFP and\/or DRAQ5 signal using the Sony SH800S Cell Sorter (purity mode) with a 100-\u03bcm sorting chip, and gating was performed in the Sony SH800S Cell Sorter Software. For CUT&RUN, 175,000 DRAQ5+GFP+ or DRAQ5+GFP\u2212 events were collected into 1\u2009ml of cold CUT&RUN Wash Buffer (20\u2009mM HEPES pH 7.5, 150\u2009mM NaCl, 0.2% Tween-20, 0.5\u2009mM spermidine, 0.1% BSA, 1\u00d7 PIC). For ATAC-seq, 50,000 DRAQ5+GFP+ events were collected into 1\u2009ml of cold ATAC-RSB (10\u2009mM Tris-HCl pH 7.5, 10\u2009mM NaCl, 3\u2009mM MgCl2). For bulk RNA-seq, 80,000 DRAQ5+GFP+ events were collected into 350\u2009\u03bcl of cold RLT buffer (Qiagen 74004) supplemented 1:100 with \u03b2-mercaptoethanol. For snRNA-seq, 50,000 DRAQ5+GFP\u2212 events were collected into a 1.5\u2009ml tube (precoated with 30% BSA) containing 300\u2009\u03bcl of cold ATAC-RSB supplemented with 1\u2009mM DTT, 1\u2009U\u2009\u03bcl\u22121 Protector RNase inhibitor, and 1% BSA (Sigma A8577-10mL).<\/p>\n CUT&RUN<\/p>\n To perform CUT&RUN61<\/a> on 3T3 cells, cells were washed twice with cold PBS and then scraped in NE1 lysis buffer (20\u2009mM HEPES pH 7.9, 10\u2009mM KCl, 0.1% Triton X-100, 3\u2009mM MgCl2, 0.5\u2009mM spermidine, 1\u00d7 PIC, 10\u2009mM sodium butyrate). Nuclei were rotated for 10\u2009min at 4\u2009\u00b0C and centrifuged at 500g for 10\u2009min at 4\u2009\u00b0C. The nuclei pellet was resuspended in 1\u2009ml of CUT&RUN Wash Buffer. 50,000 or 250,000 nuclei were then bound to 20\u2009\u03bcl of concanavalin A beads (Bangs Laboratories, BP531) for 30\u2009min at 4\u2009\u00b0C on a rotator. Samples were then incubated for 16\u201320\u2009h on a tube nutator with primary antibody (GR: Invitrogen MA1-510 and PA1-511A, IgG: Cell Signaling Technology 2729S) diluted 1:100 in Antibody Buffer (CUT&RUN Wash Buffer containing 2\u2009mM EDTA and 0.1% Triton X-100). After primary antibody incubations, samples were washed twice in Antibody Buffer and then resuspended in Triton wash buffer (CUT&RUN Wash Buffer containing 0.1% Triton X-100). Protein A-MNase (pA-MNase, prepared in-house) was added at a concentration of 700\u2009ng\u2009ml\u22121, and samples were incubated at 4\u2009\u00b0C for 1\u2009h on a tube nutator. Two washes in Triton wash buffer were performed, after which the tubes were incubated in a metal block on ice for 5\u2009min. pA-MNase digestion was initiated with the rapid addition of 2\u2009mM CaCl2. After 60\u2009min, pA-MNase digestion was halted by adding 2\u00d7 stop buffer (340\u2009mM NaCl, 20\u2009mM EDTA, 4\u2009mM EGTA, 0.04% Triton X-100, 50\u2009\u03bcg\u2009ml\u22121 RNase A, 50\u2009\u03bcg\u2009ml\u22121 glycogen, 100\u2009pg yeast spike-in DNA) at a 1:1 dilution. Digested fragments were then released at 37\u2009\u00b0C for 10\u2009min, followed by centrifugation at 16,000g for 5\u2009min at 4\u2009\u00b0C. Released DNA was treated with 0.1% SDS and 0.13\u2009mg\u2009ml\u22121 proteinase K at 50\u2009\u00b0C for 1\u2009h with gentle shaking. DNA was then purified by phenol-chloroform extraction, using an established protocol61<\/a>. CUT&RUN libraries were prepared with the SMARTer ThruPLEX DNA-seq Kit (Takara Bio R400676), with the following PCR conditions: 72\u2009\u00b0C for 3\u2009min, 85\u2009\u00b0C for 2\u2009min, 98\u2009\u00b0C for 2\u2009min, (98\u2009\u00b0C for 20\u2009s, 67\u2009\u00b0C for 20\u2009s, 72\u2009\u00b0C for 30\u2009s) \u00d7 4 cycles, (98\u2009\u00b0C for 20\u2009s, 72\u2009\u00b0C for 15\u2009s) \u00d7 8 cycles. All samples were purified with AMPure XP beads (0.5\u00d7\u20131.7\u00d7) and sequenced paired-end on an Illumina NextSeq 500 with a 75-cycle kit.<\/p>\n For astrocyte GR and IgG CUT&RUN experiments on DR and NR mice, each biological replicate consisted of P14 V1 tissue pooled from 3-5 C57BL\/6 mice that were injected at P0\u2013P1 with AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-KASH-4x6T. For astrocyte GR and NFIA CUT&RUN experiments on GR-con and GR-KO mice, each biological replicate consisted of P14 V1 tissue pooled from 3-5 GRfl\/fl mice injected at P0\u2013P1 with AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-KASH-4x6T (GR-con) or AAV2\/5-GfaABC1D-Cre-T2A-eGFP-KASH-4x6T (GR-KO). For GR, NFIA, and IgG CUT&RUN experiments on excitatory neurons, inhibitory neurons, or oligodendrocytes, each biological replicate consisted of P14 V1 tissue pooled from 3\u20135 Vglut1cre\/+;Sun1-GFPfl\/+, Vgatcre\/+;Sun1-GFPfl\/+, Olig2cre\/+;Sun1-GFPfl\/+ mice, respectively. NFIA CUT&RUN was performed with Invitrogen PA5-52252 diluted 1:100 in Antibody Buffer. Following fluorescence-activated nuclei sorting (FANS) collection of nuclei into CUT&RUN Wash Buffer, nuclei were bound to concanavalin A beads, and all downstream CUT&RUN steps were performed, as described above.<\/p>\n ATAC-seq<\/p>\n Each biological replicate consisted of P21 V1 tissue pooled from 3-4 GRfl\/fl mice that were injected at P0\u2013P1 with AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-KASH-4x6T (GR-con) or AAV2\/5-GfaABC1D-Cre-T2A-eGFP-KASH-4x6T (GR-KO). After sorting nuclei into ATAC-RSB, 0.1% Tween-20 was added, and the nuclei were centrifuged at 500g for 10\u2009min at 4\u2009\u00b0C. The nuclei pellet was directly resuspended in the transposition reaction mix, according to the OMNI-ATAC protocol62<\/a>. A 1\u2009\u03bcl volume of homemade Tn5 enzyme was used in the transposition reaction. Libraries were prepared with NEBNext High-Fidelity 2\u00d7 PCR Master Mix (NEB M0541L), following the standard protocol. After an initial five cycles of amplification, another four cycles were added to achieve optimal library concentration, based on qPCR quantification. Libraries were then size-selected (0.5\u00d7\u20131.8\u00d7) twice with AMPure XP beads (Beckman Coulter A63880) and sequenced paired-end on an Illumina NextSeq 500 using a 75-cycle kit.<\/p>\n Bulk RNA-seq<\/p>\n Each biological replicate consisted of flash-frozen V1 tissue pooled from 1 male and 1 female GRfl\/fl mice that were injected at P0\u2013P1 with AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-KASH-4x6T (GR-con) or AAV2\/5-GfaABC1D-Cre-T2A-eGFP-KASH-4x6T (GR-KO). After sorting nuclei into RLT buffer, RNA was extracted using the RNeasy Micro Kit (Qiagen 74004). RNA-seq libraries were prepared from 10\u2009ng RNA per sample using the SMARTer Stranded Total RNA-Seq Kit v2 (Takara 634413), according to the manufacturer\u2019s instructions. Libraries were sequenced paired-end on an Illumina NextSeq 500 using a 75-cycle kit.<\/p>\n snRNA-seq<\/p>\n snRNA-seq was performed on V1 tissue collected from P21 GRfl\/fl mice injected at birth with AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-CAAX-4x6T-WPRE (GR-con) or AAV2\/5-GfaABC1D-Cre-T2A-eGFP-CAAX-4x6T-WPRE (GR-KO) and then reared in DR or NR conditions. V1 was flash-frozen and stored at \u221280\u2009\u00b0C until n\u2009=\u20092 mice were collected per group, as described above. Nuclei were extracted and sorted into cold ATAC-RSB, as described above. After collection, nuclei were centrifuged two times at 200g for 2\u2009min in a swinging-bucket centrifuge at 4\u2009\u00b0C. The supernatant was then carefully aspirated from the surface of the liquid until approximately 40\u2009\u03bcl remained. Nuclei were gently resuspended in the residual volume and immediately processed with the 10X Genomics Single Cell 3\u2019 Gene Expression Kit v4 (PN-1000686), following the manufacturer\u2019s instructions. The 8 libraries were pooled and sequenced on an Illumina NovaSeq X to a depth of approximately 40,000\u201390,000 reads per cell.<\/p>\n Corticosterone ELISA<\/p>\n Total corticosterone concentration in 5\u2009\u03bcl of mouse serum was measured with a commercial corticosterone ELISA kit (Arbor Assays K014-H1), following manufacturer\u2019s instructions.<\/p>\n LC\u2013MS<\/p>\n Mouse serum samples were processed for LC\u2013MS detection of aldosterone, corticosterone, cortisol, and thyroxine (T4) by the Harvard Center for Mass Spectrometry. In brief, 20\u2009\u03bcl of mouse serum or hormone standards (a 15 point, 1\/3 dilution series with 1\u2009\u03bcM for the highest concentration of aldosterone, cortisol, corticosterone, and T4) were mixed with 100\u2009\u03bcl of extraction solution (methanol containing internal standards cortisol-D4, corticosterone-D8, and T4-13C6 each at 0.15\u2009nM, Cayman Chemicals). Samples were centrifuged at 18,000g at \u221211\u2009\u00b0C to remove proteins. The supernatant was transferred to silanized glass microinserts and evaporated to dryness under nitrogen flow. Samples were then resuspended in 15\u2009\u03bcl of 50% methanol solution in water. Samples (5\u2009\u03bcl) were run on a Kinetex C18 column (150 \u00d7 2.1\u2009mm, Phenomenex) at a column temperature of 35\u2009\u00b0C. Mobile phase A was water with 0.1% formic acid, and mobile phase B was acetonitrile with 0.1% formic acid. The gradient was as follows: 5\u2009min at 20% B, then 10\u2009min at 100% B, followed by 5\u2009min at 100% B. The column was then re-equilibrated to 20% B for 5\u2009min. The data were acquired in positive mode on a Sciex 7500 Triple Quad Mass Spectrometer (AB Sciex). The source parameters were: gas 1: 60\u2009psi, gas 2: 70\u2009psi, curtain gas: 40\u2009psi, CAD gas: 8, source temperature: 550\u2009\u00b0C, capillary at 4,000\u2009V. The transitions used are shown in Supplementary Table 5<\/a>. Mouse serum hormone concentrations were calculated using the data from the standard curve.<\/p>\n Immunoprecipitation of Flag-tagged GR and NFIA in HEK293T cells<\/p>\n Plasmids (pCIG with CAG promoter) containing the sequence for Flag-tagged GR or HA-tagged NLS\u2013GFP, LHX2, or NFIA were co-delivered into HEK293T cells by CaCl2-BBS transfection (10\u2009\u03bcg of each plasmid, 10\u2009\u00d7\u2009106 cells). The same approach was used to co-deliver Flag-tagged NFIA with HA-tagged NLS\u2013GFP or LHX2. After 24\u2009h, the media was replaced with phenol-red-free DMEM containing charcoal-stripped FBS, as described above, and incubated for another 24\u2009h. Cells were then treated with 200\u2009nM dexamethasone for 2\u2009h. Cells were washed twice with ice-cold PBS and then scraped in Nuclear Extraction (NE) Buffer (20\u2009mM HEPES pH 7.9, 10\u2009mM KCl, 3\u2009mM MgCl2, 0.1% Triton X-100, 1\u00d7 PIC, phosphatase inhibitors 2 and 3). Cells were pelleted by gentle centrifugation (500g for 10\u2009min at 4\u2009\u00b0C). Flag\u2013GR samples were extracted with either benzonase (Sigma Aldrich E8263) or MNase (NEB M0247S), while Flag\u2013NFIA samples were extracted with benzonase. For benzonase extraction, samples were resuspended in 1\u2009ml of NE Buffer, 5\u2009\u03bcl of benzonase was added, and samples were incubated for 30\u2009min at 4\u2009\u00b0C with rotation. NaCl was added to a concentration of 300\u2009mM, and samples were mixed for an additional 30\u2009min at 4\u2009\u00b0C with rotation. Lysates were spun at 16,000g for 20\u2009min at 4\u2009\u00b0C, and the supernatant was transferred to a fresh tube. Samples were then diluted to 200\u2009mM NaCl in NE Buffer. For MNase extraction, samples were resuspended in NE Buffer, and 1\u2009mM CaCl2 and 24\u2009\u03bcl of MNase were added. Samples were incubated at 37\u2009\u00b0C for 20\u2009min and then MNase stop buffer (40\u2009mM EGTA, 20\u2009mM EDTA, 3\u2009M NaCl) was added at a 1:20 dilution. Lysates were spun at 16,000g for 20\u2009min at 4\u2009\u00b0C, and the supernatant was transferred to a fresh tube. For each sample, 5% of lysate was collected for the input fraction. Samples were pre-cleared with 50\u2009\u03bcl of washed agarose resin (Sigma Aldrich A0919). Immunoprecipitations were then performed using 100\u2009\u03bcl of anti-Flag magnetic agarose resin (Thermo Fisher A36798) for 2\u2009h at 4\u2009\u00b0C with rotation. Following immunoprecipitations, samples were washed 4 times (5\u2009min at 4\u2009\u00b0C for each wash) with NE Buffer containing 250\u2009mM NaCl. After the final wash, samples were eluted with 3\u00d7 Flag peptide (Sigma Aldrich F4799) diluted to 0.5\u2009mg\u2009ml\u22121 in NE Buffer containing 250\u2009mM NaCl for 30\u2009min at RT at 1,000\u2009rpm and then stored at \u221280\u2009\u00b0C until immunoblotting.<\/p>\n Immunoblotting<\/p>\n Input or immunoprecipitation fractions were resolved on NuPage 4\u201312% Bis-Tris gels (Thermo Fisher, NP0336BOX) and transferred to nitrocellulose membranes. Membranes were blocked for 1\u2009h with SuperBlock (Thermo Fisher 37518) and incubated overnight with the following primary antibodies (1:1,000): Flag (Sigma Aldrich F1804) and HA (Cell Signaling Technology 3724). Following washing, membranes were incubated with secondary antibodies against rabbit IgG or mouse IgG conjugated to horseradish peroxidase (Cell Signaling Technology 7074S or 7076S, respectively) at a concentration of 1:5,000, labelled with SuperSignal Pico PLUS reagents (Thermo Fisher A45915), and imaged using a Bio-Rad Gel Doc System. Full scans of western blots are shown in Supplementary Figs. 1<\/a> and 2<\/a>.<\/p>\n Immunofluorescent staining<\/p>\n Male and female mice were used in equal proportions for histology experiments. Mice were anaesthetized by intraperitoneal injection of 10\u2009mg\u2009ml\u22121 ketamine and 1\u2009mg\u2009ml\u22121 xylazine. Upon anaesthetization, mice were transcardially perfused with 15\u201320\u2009ml of cold PBS followed by 15\u201320\u2009ml of 4% PFA diluted in PBS. Brains were incubated in 4% PFA at 4\u2009\u00b0C for 16\u201320\u2009h, followed by two washes in PBS. Brains were then sectioned at 50\u2009\u03bcm or 100\u2009\u03bcm (for astrocyte reconstruction experiment) thickness on a vibratome and stored in PBS at 4\u2009\u00b0C for 48\u201372\u2009h prior to staining. Sections were blocked in PBS containing 0.1% Triton X-100 and 10% normal donkey serum (NDS, Sigma D9663) for 1\u2009h at room temperature. Staining in primary antibodies (dilutions listed below) or biotinylated WFA (Vector Laboratories B-1355-2, 1:1,000) was carried out in PBS containing 0.1% Triton X-100 and 0.5% NDS (PDT buffer) at room temperature for 16\u201320\u2009h with shaking. For labelling astrocyte morphology in 100\u2009\u03bcm-thick sections, primary antibody incubation was performed at 4\u2009\u00b0C for 48\u2009h. After labelling with primary antibody, sections were washed 3 times (10\u2009min each) in PDT buffer. Secondary antibody labelling was performed with fluorophore-conjugated secondary antibodies (rabbit Dylight 405 (Jackson ImmunoResearch 711-475-152), chicken Alexa Fluor 488 (Jackson ImmunoResearch 703-545-155), rabbit Alexa Fluor 594 (Life Technologies A21207), mouse Alexa Fluor 594 (Jackson ImmunoResearch 715-585-150), guinea pig Alexa Fluor 647 (Jackson ImmunoResearch 706-605-148), goat Alexa Fluor 647 (Life Technologies A21447), mouse Alexa Fluor 647 (Life Technologies A31571)) diluted 1:300 or streptavidin Alexa Fluor 647 (Jackson ImmunoResearch 016-600-084) diluted 1:1,000 in PDT buffer for 2\u2009h at room temperature. For experiments requiring nuclear labelling, DAPI was added at 1\u2009\u03bcg\u2009ml\u22121 during secondary antibody incubation. Sections were again washed two times (10\u2009min each) in PDT buffer, followed by an additional two washes with PBS containing 0.1% Triton X-100. Sections were then transferred to PBS and mounted with ProLong Diamond Antifade mountant (Invitrogen P36970) and imaged at 16-bit depth on a point scanning confocal microscope (Leica Stellaris 5). Primary antibodies were diluted in PDT buffer as follows: rabbit anti-GR (Cell Signaling Technology 12041S, 1:300), goat anti-SOX9 (R&D Systems AF3075, 1:300), chicken anti-GFP (Aves Labs GFP-1020, 1:1,000), rabbit anti-Acan (Sigma AB1031, 1:500), guinea pig anti-PV (Synaptic Systems 195-308, 1:1,000), rabbit anti-PV (Swant PV27a, 1:1,000), rabbit anti-HOMER1 (Synaptic Systems 160003, 1:500), guinea pig anti-Vglut1 (Sigma AB5905, 1:500), rabbit anti-NeuN (EMD Millipore ABN78, 1:500), mouse anti-SYT2 (ZIRC ZDB-ATB-081002-25, 1:500), and guinea pig anti-CB1R (Frontier Institute MSFR100630, 1:300).<\/p>\n For validating the astrocyte specificity of our viruses and assessing the degree of GR protein KO, one V1 was imaged with the 63\u00d7 objective at 1,024\u2009\u00d7\u20091,024 pixel resolution with 0.75\u00d7 optical zoom (11 z-planes, 3\u2009\u03bcm step size) and quantified per animal. For reconstructing astrocyte morphology, V1 astrocytes within 2\u20134 fields of view (FOVs) across 2\u20133 sections were imaged with the 63\u00d7 objective at 2,048\u2009\u00d7\u20092,048 pixel resolution with 0.75\u00d7 optical zoom (80\u2013140 z-planes, 0.5\u2009\u03bcm step size). For P21, P28, P35 and adult PNN and PV staining, V1 was imaged in 2 coronal sections with the 63\u00d7 objective at 1,024\u2009\u00d7\u20091,024 pixel resolution with 0.75\u00d7 optical zoom (12 z-planes, 2\u2009\u03bcm step size) and quantified at the animal level. For adult V1 L5 Acan and PV staining, 3\u20135 FOVs in 2\u20133 coronal sections were imaged with the 63\u00d7 objective at 1,024\u2009\u00d7\u20091,024 pixel resolution with 0.75\u00d7 optical zoom (37 z-planes, 1\u2009\u03bcm step size) and quantified at the animal level. For V1 L5 Vglut1\/HOMER1 staining experiments, 3\u20135 FOVs in 2\u20133 coronal sections were imaged with the 100\u00d7 objective at 2,048\u2009\u00d7\u20092,048 pixel resolution with 2.5\u00d7 optical zoom (31 z-planes, 0.5\u2009\u03bcm step size) and quantified at the animal level. For V1 L5 SYT2\/CB1R staining experiments, 3\u20135 FOVs in 2\u20133 coronal sections were imaged with the 100\u00d7 objective at 1,024\u2009\u00d7\u20091,024 pixel resolution with 0.75\u00d7 optical zoom (41 z-planes, 0.5\u2009\u03bcm step size) and quantified at the animal level. For P28 SYT2 staining (Extended Data Fig. 17e,f<\/a>), the same parameters were used, and 51 z-planes were collected.<\/p>\n ElectrophysiologyAcute slice preparation<\/p>\n Coronal cortical slices were prepared from P33-35 GRfl\/fl mice injected at birth with AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-CAAX-4x6T-WPRE (GR-con) or AAV2\/5-GfaABC1D-Cre-T2A-eGFP-CAAX-4x6T-WPRE (GR-KO) for the mIPSC\/mEPSC recordings. Slices were also prepared from P33-35 PvalbFlpO\/FlpO;GRfl\/fl mice co-injected at birth with AAV2\/8-CAG-FLPX-rc (ChrimsonR-tdTomato) and AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-CAAX-4x6T-WPRE (GR-con) or AAV2\/5-GfaABC1D-Cre-T2A-eGFP-CAAX-4x6T-WPRE (GR-KO) to measure PV-evoked IPSCs. Mice were anaesthetized with isofluorane and transcardially perfused with ice-cold choline-based artificial cerebrospinal fluid (choline ACSF, in mM: 110 choline chloride, 25 NaHCO3, 1.25 NaH2PO4, 2.5 KCl, 7 MgCl2, 0.5 CaCl2, 25 glucose, 11.6 sodium-l-ascorbate and 3.1 sodium pyruvate, 320\u2013330\u2009mOsm) equilibrated with 95% O2\/5% CO2. After perfusion, the brain was rapidly dissected and blocked in ice-cold equilibrated choline ACSF. Tissue was then transferred to a cutting chamber containing ice-cold equilibrated choline ACSF and cut on a Leica VT1200S (300\u2009\u00b5m thickness, 0.10\u2009mm\u2009s\u22121, 1\u2009mm amplitude). Slices were then collected in a holding chamber containing artificial cerebrospinal fluid (ACSF) (in mM: 127 NaCl, 25 NaHCO3, 1.25 NaH2PO4, 2.5 KCl, 1 MgCl2, 2 CaCl2 and 10 glucose, 300\u2013310\u2009mOsm). Slices were recovered at 32\u2009\u00b0C for 20\u2009min and then maintained at room temperature (20\u201322\u2009\u00b0C) for at least 20\u2009min before the start of recordings. Experiments were performed within 6\u2009h after cutting at room temperature.<\/p>\n Ex-vivo slice electrophysiology<\/p>\n For whole-cell voltage-clamp recordings of miniature excitatory and inhibitory currents, patch pipettes made with borosilicate glass with filament (Sutter BF150-86-7.5) and 2\u20135 M\u03a9 resistance were filled with a Cs+-methanesulfonate internal solution consisting of (in mM): 127 CsMeSO3, 10 CsCl, 10 HEPES, 0.5 EGTA, 2 MgCl2, 0.16 CaCl2, 2 MgATP, 0.4 NaGTP, 14 sodium phosphocreatinine, and 2 QX314-Cl (pH 7.2, 295\u2009mOsm). Recordings were made on an upright Olympus BX51 W1 microscope with an infrared CCD camera (Dage-MTI IR-1000) and 60\u00d7 water immersion objective (Olympus Lumplan FI\/IR 60\u2009\u00c5\/0.90 NA). Neuronal tissue was visualized with infrared differential interference contrast. GFP-expressing astrocytes were identified by epifluorescence driven by a light-emitting diode (Excelitas XCite LED120). Layer 5 pyramidal neurons in close proximity to GFP signal were identified based on the depth from pia (450\u2013590\u2009\u00b5m) and their large somas and apical dendrites. For isolation of excitatory and inhibitory miniature currents, cells were clamped at \u221265mV and 0\u2009mV, respectively in the presence of 0.5\u2009\u00b5M TTX (Tocris 1069). For measuring PV-mediated evoked IPSCs, a CsCl-based internal solution consisting of (in mM): 135 CsCl, 3.3 QX314-Cl, 10 Hepes, 4 MgATP, 0.5 NaGTP, 8 sodium phosphocreatine, 1.1 EGTA and 0.1 CaCl2 (pH 7.2, 290\u2009mOsm) was used, and cells were clamped at \u221270mV. NBQX (Tocris 1044, 10\u2009\u00b5M) and (R)-CPP (Tocris 0247, 10\u2009\u00b5M) were added to the ACSF perfusion to pharmacologically block AMPA and NMDA receptors, respectively.<\/p>\n Ex\u00a0vivo slice electrophysiology data acquisition and analysis<\/p>\n Synaptic miniature excitatory or inhibitory currents were collected for three minutes in whole-cell voltage-clamp (36 sweeps of 5\u2009s length). At the end of each sweep, a 200\u2009ms long \u22125\u2009pA current step was added to continuously measure series resistance, membrane resistance and capacitance. For photostimulation of ChR2-expressing boutons, orange-red light at 590\u2009nm was delivered through the 60\u00d7 objective lens using the light-emitting diode, controlled via TTL pulses synchronized with the electrophysiology acquisition system pCLAMP (10.6) Clampex module. A single light stimulation of 0.5\u2009ms was delivered at increasing light intensities at the slice surface between 1\u20134\u2009mW\u2009mm\u22122, as measured with a photodiode sensor for evoked recordings, and five pulses at 10\u2009Hz were delivered for short-term plasticity experiments. Light intensity was chosen based on prior studies that elicited PV-evoked potentials63<\/a>,64<\/a>. Cells that needed more than \u2212500 pA of current injection for clamping at desired mV, exceeded 30 mOhm of series resistance, or changed their series resistance by more than 30% before the start and end of the experiment were excluded. An Axon Multiclamp 700B (controlled by Multiclamp 700B commander) was used to perform voltage- and current-clamp experiments and low-pass filtering at 4\u2009kHz. Data were sampled at 10\u2009kHz with an Axon Digidata 1,440\u2009A and controlled using the Clampex module of pCLAMP (10.6). Miniature events were analysed using template event detection in the Clampfit module of pCLAMP (10.6). Evoked responses were analysed using peak amplitude measurements. All measurements are displayed as mean\u2009\u00b1\u2009s.e.m. Experiments and data analysis were performed blind to genotype. Statistical testing was performed in Prism 10 using the appropriate t-test based on normality distribution and equal standard deviation (unpaired t-test, Mann\u2013Whitney test, Kolmogorov\u2013Smirnov test) for mEPSC and mIPSC currents. Two-way analysis of variance (ANOVA) testing with Bonferroni correction was performed for PV-mediated evoked currents and short-term plasticity.<\/p>\n Ocular dominance plasticityMonocular deprivation<\/p>\n MD was performed while mice were under isoflurane anaesthesia (4% for induction and 2% for maintenance). The margins of the left eyelid were trimmed, and the lids were sutured closed (Ethicon 7-0 Perma-Hand Silk Sutures). Mice were monocularly deprived for 3\u20134 days prior to ocular dominance recordings.<\/p>\n In vivo extracellular recordings from primary visual cortex<\/p>\n Recordings were performed in anaesthetized mice using standard acute in vivo electrophysiological procedures65<\/a>,66<\/a>. Mice were anaesthetized with isoflurane (4% induction, 2% maintenance) and a ~1.5\u2009mm diameter craniotomy was made centred 3\u2009mm lateral to the lambda suture. Mice received 2.5\u2009mg\u2009kg\u22121 dexamethasone subcutaneously to reduce oedema. Eyes were lubricated with silicone oil (Sigma) to prevent corneal drying. To stabilize anaesthesia during recordings, mice received chlorprothixene (5\u2009mg\u2009kg\u22121, subcutaneous)65<\/a>, and isoflurane was reduced to ~1% during data acquisition. Extracellular signals were recorded using multisite silicon probes (Cambridge Neurotech, ASSY-37-H7b). Data were acquired through a PZ5 digitizer and RZ5P processor using Synapse software (Tucker-Davis Technologies). Up to three probe penetrations were performed in each animal. Recordings were always performed in the right visual cortex\u2014contralateral relative to the sutured eye. Electrodes were coated with DiD (Invitrogen) in the last penetration to enable post hoc visualization of the electrode track. Experimenters were blind to the AAV (Cre versus \u0394Cre) injected into V1.<\/p>\n Visual stimulation<\/p>\n Visual stimuli were presented on a 21\u2009\u00d7\u200912 inch LED monitor with a refresh rate of 60\u2009Hz and mean luminance of 20 lux. The monitor was centred on the binocular visual field, perpendicular to the antero-posterior axis of the mouse. Visual stimuli were generated using the Psychophysics toolbox67<\/a> in MATLAB (Mathworks). To confirm that receptive fields were in the binocular zone, contrast-modulated white noise movies were presented65<\/a>. To measure evoked responses to visual stimuli, mice were presented with full-field drifting grating moving in 8 different directions with a spatial frequency of 0.02 cycles per degree and a temporal frequency of 2\u2009cycles per second. A blank grey screen stimulus was presented to estimate spontaneous activity. Gratings were presented in pseudorandom order for 40 trials in total, alternating between the ipsilateral and contralateral eyes.<\/p>\n Analysis<\/p>\n Adequate AAV targeting of V1, as measured by viral GFP immunofluorescence, was confirmed for each animal post hoc as a criterion for inclusion in the ODP experiment. Spikes were extracted using an amplitude threshold set at 6\u00d7 s.d. above baseline noise. Recording sites (multi-units) were included for analysis if they were located within the binocular zone, defined as positions within 25\u00b0 in azimuth. Responses to drifting gratings were subtracted from responses to the blank stimulus to measure evoked firing. Responses to gratings moving in all directions were averaged to calculate the evoked firing rate of each contact site. Ocular dominance index was computed for each recording site as (Rcontra\u2009\u2013\u2009Ripsi)\/(Rcontra\u2009+\u2009Ripsi), where Rcontra and Ripsi are the evoked responses to the contralateral and ipsilateral eye, respectively. Ocular dominance index measurements are displayed per recording site (multi-unit) and averaged per animal. Statistical testing was performed in Prism 10 using the Kruskal\u2013Wallis test with Dunn\u2019s multiple comparisons correction or one-way ANOVA test with Tukey\u2019s multiple comparisons correction.<\/p>\n Bioinformatics and data analysisImaging data analysis<\/p>\n All image collection and analysis was performed by an investigator blinded to experimental conditions. For validating the astrocyte specificity of our viruses, assessing the degree of GR knockout, and quantifying the heterogeneity of GR expression in V1 astrocytes at P35 and P21 (Extended Data Fig. 8h\u2013m<\/a>), images were loaded into Fiji\/ImageJ, and the proportion of GFP+ cells containing SOX9 (SOX9+GFP+) or GR (GR+GFP+) as well as the proportion of SOX9+ cells expressing GR (GR+SOX9+) were counted manually using the multi-point tool.<\/p>\n All other image analysis was performed in Imaris v10.2 (Oxford Instruments). To quantify the cell volume and NIV of GR-con and GR-KO astrocytes (Fig. 3a,b<\/a>), the smV5 fluorescence signal was reconstructed using the surface tool. Individual astrocytes (examples shown in Extended Data Fig. 12a<\/a>) were selected for quantification: (1) if the cell boundaries in 3D space could be identified; and (2) if the cell contained a GFP-labelled nuclear membrane, indicating co-transduction with AAV2\/5-GfaABC1D-\u0394Cre-T2A-eGFP-KASH-4x6T (GR-con) or AAV2\/5-GfaABC1D-Cre-T2A-eGFP-KASH-4x6T (GR-KO). NIV per astrocyte was calculated in Imaris, as previously described36<\/a>. In brief, the astrocyte smV5 surface volume within three randomly selected 10\u2009\u00b5m \u00d7 10\u2009\u00b5m \u00d7 10\u2009\u00b5m boxes (located within the neuropil) was quantified and averaged per cell.<\/p>\n To quantify PV neuron-bound Acan, the PV fluorescence signal was reconstructed using the surface tool. The mean Acan fluorescence intensity on PV neuron surfaces per FOV was calculated in Imaris and then averaged across all FOVs per animal (Extended Data Fig. 16n,o<\/a>). To quantify PNN and PV intensity at P35 and in adults, the PV fluorescence signal was reconstructed using the surface tool, and PV neuron surface objects were manually annotated to cortical layers (L2\/3, L4, L5 and L6) based on the DAPI channel. The mean WFA or PV fluorescence intensity on PV neuron surfaces for each cortical layer was then calculated in Imaris and averaged between two coronal sections per animal (Fig. 4a\u2013d<\/a> and Extended Data Fig. 16d,h,j,l<\/a>).<\/p>\n To measure the effect of astrocyte GR knockout on SYT2+ or CB1R+ synapses on neuronal somas, the astrocyte mGFP fluorescence signal and NeuN fluorescence signal were reconstructed using the surface tool. NeuN surfaces overlapping astrocyte mGFP surfaces were then selected and duplicated into a new surface object. SYT2+ puncta (size\u2009=\u20091.3\u2009\u03bcm) and CB1R+ puncta (size\u2009=\u20091.3\u2009\u03bcm) were reconstructed using the spots tool. SYT2+ and CB1R+ spots within 0.9\u2009\u03bcm of mGFP-contacting NeuN surfaces were identified using the Matlab extension \u2018spots close to surface\u2019. The density of SYT2+ or CB1R+ spots per NeuN surface volume (\u03bcm3) was then calculated for each NeuN surface per FOV, and the median density was computed across FOVs per animal (Fig. 4e,f<\/a> and Extended Data Fig. 17a,b,e,f<\/a>).<\/p>\n To measure the effect of astrocyte GR knockout on Vglut1+HOMER1+ synapses, the astrocyte mGFP fluorescence signal was reconstructed using the surface tool, and Vglut1+ (size\u2009=\u20090.6\u2009\u03bcm) and HOMER1+ (size\u2009=\u20090.4\u2009\u03bcm) puncta were reconstructed using the spots tool. Vglut1+ and HOMER1+ spots within 0.5\u2009\u03bcm of the astrocyte mGFP surface were selected using the Matlab extension \u2018spots close to surface\u2019. Vglut1+ and HOMER1+ puncta within 0.5\u2009\u03bcm of each other were then detected using the Matlab extension \u2018co-localize spots\u2019. The density of co-localized Vglut1+HOMER1+ spots per astrocyte mGFP surface volume (\u03bcm3) was then calculated per FOV, and the median density was computed across FOVs per animal (Fig. 4g,h<\/a> and Extended Data Fig. 17c,d,g,h<\/a>).<\/p>\n SHARE-seq data processing and analysis<\/p>\n SHARE-seq data were processed using the SHARE-seq-alignmentV2 pipeline (https:\/\/github.com\/masai1116\/SHARE-seq-alignmentV2<\/a>), generating gene\u2009\u00d7\u2009nucleus matrices for the snRNA assay and fragment files for the single-nucleus ATAC assay. snRNA gene\u2009\u00d7\u2009nucleus matrices were loaded into Seurat68<\/a> (v5.3.0), and nuclei with fewer than 600 gene counts were removed. Scrublet69<\/a> was used to predict and remove doublets, and post-doublet removal, nuclei were further filtered on the basis of gene counts, retaining nuclei that have between 750 and 80,000 gene counts. Gene counts were normalized and log-transformed (NormalizeData), and the top 2,000 variable features were identified using FindVariableFeatures. Gene counts were scaled (ScaleData), regressing out the number of RNA counts, number of RNA genes, and percentage of mitochondrial gene counts per nucleus. Dimensionality reduction was performed with principal component analysis (runPCA), and a k-nearest-neighbours graph was constructed (FindNeighbors, npcs\u2009=\u200915). Nuclei were clustered to identify principal cell classes with the Louvain algorithm (FindClusters, resolution\u2009=\u20090.2) and visualized by UMAP (runUMAP, dims\u2009=\u200915). Nuclei were annotated to broad cell classes (astrocytes, endothelial cells, excitatory neurons, inhibitory neurons, microglia, oligodendrocytes, OPCs, SMC\u2013Peri, VLMC) on the basis of known marker genes. Following this initial round of clustering, excitatory neurons (Slc17a7+ clusters) and inhibitory neurons (Gad2+ clusters) were subsetted and re-clustered to identify cell subtypes, using the same procedure described above (FindClusters, resolution\u2009=\u20090.3). After clustering, any nuclei cluster lacking significant marker genes (FindMarkers) were removed, resulting in the final processed dataset (70,907 cells). Single-nucleus ATAC fragment files were loaded into Signac70<\/a> (v1.14.0), and initial 5,000\u2009bp\u2009\u00d7\u2009cell genome bin matrices were constructed using the GenomeBinMatrix function. Chromatin assay objects were then created using the CreateChromatinAssay function, and these objects were subsequently filtered to keep cell barcodes present in the snRNA assay. Peaks were called on the dataset using MACS2 (CallPeaks). Gene accessibility score (shown in Extended Data Fig. 1e,h,i<\/a>), calculated as the ATAC signal over the gene body\u2009\u00b1\u20092,000 bp, was quantified with the GeneActivity function.<\/p>\n Pseudobulk differential expression analysis between NR and DR conditions for each cell type and developmental time point was performed using edgeR glmQLFTest (Padj\u2009<\u20090.05). Differentially expressed genes were separated into cell type-shared (differential in \u22652 cell types) or cell-type-specific (differential in 1 cell type), and shared genes were then hierarchically clustered by log2(fold change) value using the ward.D2 algorithm in pheatmap for visualization in Fig. 1d<\/a>. To identify astrocyte experience-responsive genes (Fig. 2f<\/a>, Extended Data Fig. 9a\u2013c<\/a>) for comparison to GR binding sites, genes differentially expressed between NR and DR (edgeR Padj\u2009<\u20090.05) astrocytes at any time point were grouped and hierarchically clustered using the average algorithm in pheatmap with cutree\u2009=\u20094. Pseudobulk differential accessibility analysis was performed between NR and DR conditions for each cell type and developmental time point using edgeR glmQLFTest (Padj\u2009<\u20090.1), after pooling early (P7\u2013P14) and late (P21\u2013P35) time points to increase cell depth.<\/p>\n To identify transcription factors influenced by visual experience (Fig. 2a<\/a>), ATAC peaks were annotated for the presence of transcription factor binding motifs (JASPAR 2020 database) using the AddMotifs function. ChromVAR20<\/a> was then used to calculate the background-corrected, genome-wide chromatin accessibility score of all transcription factor binding motifs in individual nuclei. Experience-regulated transcription factor chromVAR scores for each cell type at P21 were then identified using MAST71<\/a> within the FindMarkers function. The top 2 unique differentially accessible transcription factor binding motifs (by Padj value) between NR and DR for each cell type are shown in a heat map in Fig. 2a<\/a>, in which each row is scaled to the maximal within-row \u2212log10(Padj) value. To predict cell identity transcription factors from SHARE-seq data (Extended Data Fig. 5a<\/a>), as done previously17<\/a>, the Pearson correlation coefficient between transcription factor chromVAR score and transcription factor expression was calculated at the cell type level for all JASPAR2020 transcription factors with matched motif and gene names. Transcription factors with a correlation coefficient >0.3 and maximal transcription factor RNA FC\u2009>1 between cell types were highlighted in red in Extended Data Fig. 5a<\/a> (top candidate transcription factors for each cell type shown in Extended Data Fig. 5b<\/a>). To identify experience-responsive, cell type-enriched transcription factor binding motifs (Extended Data Fig. 5c\u2013g<\/a>), nuclei were subsetted to late postnatal time points (P21\u2013P35), which display the majority of experience-responsive genes and chromatin regions. Experience-regulated transcription factor chromVAR scores and cell type-enriched transcription factor chromVAR scores were separately computed for each cell type using MAST71<\/a> within the FindMarkers function. To prioritize transcription factor binding motifs that are strongly influenced by experience and more accessible in specific cell types, the rank-product of these 2 statistical tests was calculated, and the top 2\u20133 transcription factors per cell type based on rank-product were selected to display in Extended Data Fig. 5c<\/a>.<\/p>\n MERFISH data processing and analysis<\/p>\n MERFISH imaging data were processed on the MERSCOPE instrument with cell segmentation using a watershed algorithm based on DAPI and PolyT staining. Individual images were then loaded into MERSCOPE Visualizer and a region of interest (ROI) around V1 was drawn to subset cells for downstream analysis. STAlign72<\/a> was used to spatially align the centroid position of individual cells across P21 and P35 tissue sections. MERFISH expression matrix and cell boundary files were then loaded and analysed in Seurat (v5.3.0). Cells with fewer than 20 gene counts were removed. Gene counts were normalized and log-transformed and then scaled (ScaleData), regressing out the number of transcript counts per cell. Dimensionality reduction and clustering were performed, as described above for SHARE-seq. Cells were annotated to broad cell classes on the basis of known marker genes. Subsequently, excitatory neurons (Slc17a7+) and inhibitory neurons (Gad2+) were subsetted and re-clustered as before to identify neuronal subtypes. Following clustering, any cell cluster lacking significant marker genes (FindMarkers) were removed, resulting in the final processed dataset (49,010 cells). Differential expression analysis between DR and NR cells at each time point was performed on pseudobulk samples and individual cells, using edgeR glmLRT test (Padj\u2009<\u20090.1) and MAST (Padj\u2009<\u20090.01), respectively.<\/p>\n CUT&RUN data processing and analysis<\/p>\n CUT&RUN sequencing data were processed with a custom bash script that involves the following steps: adapter trimming (cutadapt73<\/a> -q 30), Bowtie274<\/a> alignment to mm10 (–dovetail –very-sensitive-local –nounal –no-mixed –no-discordant –phred33), duplicate read removal (Picard MarkDuplicates) with Picard (http:\/\/broadinstitute.github.io\/picard\/<\/a>), mapping quality (samtools75<\/a> view -q 40), fragment length filtering (deepTools76<\/a> alignmentSieve –maxFragmentLength 120), and track generation (deepTools bamCoverage –binSize 1 –normalizeUsing CPM). Peaks were called on individual sample BAM files with MACS277<\/a> callpeak at a q-value threshold of 0.05 using the IgG BAM file as a local background control.<\/p>\n DiffBind78<\/a> was used to perform CUT&RUN differential peak-calling, and also to identify consensus peaks present in 2 of 3 biological replicates for GR and NFIA across cortical cell types (Fig. 2g<\/a>). edgeR79<\/a> was applied within DiffBind to identify differential GR sites, using the following parameters: 3T3 cell vehicle versus dexamethasone GR CUT&RUN (Padj\u2009<\u20090.1); astrocyte DR versus NR GR CUT&RUN (Padj\u2009<\u20090.01). Tracks of IgG, GR, or NFIA CUT&RUN signal for individual biological replicates or merged BAM file across replicates were generated using trackViewer80<\/a>. The location of GR and NFIA binding sites (Extended Data Fig. 10e,i<\/a>) relative to the NCBI RefSeq mm10 gene annotation was calculated using ChIPseeker81<\/a>, and sites were then separated into gene distal (>500\u2009bp from the transcription start site (TSS)) or promoter (\u00b1500\u2009bp around the TSS) elements for motif analysis. Motif enrichment analysis of GR and NFIA sites was performed with HOMER82<\/a>, using standard parameters. DeepTools plotHeatmap or plotProfile were used to visualize IgG, GR or NFIA CUT&RUN signal, representing CPM-normalized bigwig files for pooled samples (Fig. 2d<\/a>, Extended Data Figs. 9g<\/a> and 10d<\/a>) or individual replicates (Extended Data Figs. 7e<\/a>,\u00a08d<\/a> and 10h<\/a>). Intersections between astrocyte GR binding sites and mouse GR binding sites detected by GR chromatin immunoprecipitation with sequencing (ChIP\u2013seq) in published datasets (kidney83<\/a>, liver84<\/a>, primary brown preadipocytes85<\/a>, embryonic fibroblasts86<\/a>, mammary gland87<\/a> and primary bone marrow-derived macrophages88<\/a>) were calculated with BEDtools89<\/a> intersect, as shown in Extended Data Fig. 8f<\/a>. BETA90<\/a> (basic mode, -d 200000) was used to assess whether GR sites and GR-dependent ATAC sites are over-represented near P21 astrocyte GR-regulated genes (Padj\u2009<\u20090.05) relative to non-differential, expressed genes (Extended Data Fig. 13a,c,d<\/a>). In addition, BETA was used to calculate GR binding enrichment score for astrocyte experience-responsive genes identified in the SHARE-seq dataset (Fig. 2f<\/a> and Extended Data Fig. 9a\u2013c<\/a>), which were defined as described above. To identify whether GR binding occurs at ATAC sites that correlate with astrocyte experience-responsive genes (Extended Data Fig. 9d\u2013f<\/a>), the correlation between the accessibility of ATAC sites within 500,000\u2009bp of the TSS of experience-responsive genes and their expression was calculated with Signac LinkPeaks. Pearson correlation coefficients and P values were then segregated into 4 groups based on: (1) whether the target gene was suppressed or induced by experience; and (2) whether or not the ATAC site overlapped GR binding. Additionally, these 4 groups of ATAC sites were compared on the basis of their distance to the correlated gene TSS (Extended Data Fig. 9f<\/a>).<\/p>\n ATAC-seq data processing and analysis<\/p>\n ATAC-seq data were processed using a standard pipeline established by ENCODE (https:\/\/github.com\/ENCODE-DCC\/atac-seq-pipeline<\/a>) with default parameters. ATAC-seq differential peak-calling between GR-con and GR-KO astrocytes was performed with edgeR in DiffBind, using the following parameters: Padj\u2009<\u20090.05, abs(log2(FC)) > 0.25. DeepTools plotHeatmap was used to plot ATAC CPM for individual biological replicates at GR-regulated ATAC sites (Extended Data Fig. 13e,f<\/a>). De novo motif analysis of GR-regulated ATAC sites was performed with HOMER, using standard parameters.<\/p>\n Bulk nuclear RNA-seq data processing and analysis<\/p>\n RNA-seq reads were mapped to mm10 using STAR91<\/a> with default parameters. The number of reads mapping to gene bodies (exons and introns) was quantified with Subread featureCounts92<\/a>, using the mm10 UCSC refGene annotation. Differential gene expression between NR GR-con and NR GR-KO astrocytes at each time point (P14, P21 and P28) was performed with DESeq2 (Padj\u2009<\u20090.05, abs(log2(FC)) > 0.1) and shown in Fig. 3c<\/a>. Differential gene expression between NR GR-con and DR GR-con astrocytes at each time point (P14, P21 and P28) was performed with DESeq2 (Padj\u2009<\u20090.05) and shown in Extended Data Fig. 12d<\/a>. For comparisons of P21 astrocyte GR-regulated genes with P21 astrocyte experience-regulated genes (Extended Data Fig. 12e<\/a>), astrocyte maturation genes across other mouse brain regions (Extended Data Fig. 12h<\/a>), human astrocyte maturation genes (Fig. 3k<\/a>), or P21 astrocyte GR-regulated genes detected by snRNA-seq (Extended Data Fig. 18c<\/a>), a relaxed set of GR-regulated genes were called without a log-fold change threshold (DESeq2, Padj\u2009<\u20090.05).<\/p>\n To identify genes that change expression across V1 astrocyte development under normal conditions (Fig. 3d<\/a> and Extended Data Fig. 12g<\/a>), differential gene expression between P21 GR-con and P14 GR-con astrocytes was performed with DESeq2 (Padj\u2009<\u20090.05, abs(log2(FC)) > 0.3). Differentially expressed genes were then hierarchically clustered on the basis of mean expression across time points (P14, P21 and P28) and GR status (GR-con, GR-KO) using the ward.D2 algorithm in pheatmap and split into 4 clusters with cutree\u2009=\u20094 (shown in Fig. 3d<\/a>). The expression of genes in 2 of these 4 clusters was altered in GR-KO astrocytes and classified as GR-dependent. ClusterProfiler93<\/a> was then used to identify enrichment of GO biological process terms in GR-independent and GR-dependent astrocyte maturation genes (enrichGO, Padj\u2009<\u20090.05).<\/p>\n Mouse developmental astrocyte snRNA-seq analysis<\/p>\n Mouse brain astrocyte snRNA-seq data39<\/a> containing 68,485 astrocytes across 5 brain regions (whole cortex, motor cortex, PFC, striatum and thalamus) and 6 time points (E18.5, P4, P14, P32, P90 and 90 weeks old (not shown)) were accessed from the Broad Institute Single Cell Portal (https:\/\/singlecell.broadinstitute.org\/single_cell\/study\/SCP2719\/a-multi-region-transcriptomic-atlas-of-developmental-cell-type-diversity-in-mouse-brain<\/a>). Gene counts were normalized and log-transformed (Seurat NormalizeData) and then scaled (ScaleData). Linear dimensionality reduction was performed by principal component analysis (runPCA, npcs\u2009=\u200950), and the snRNA data were visualized with UMAP (runUMAP). The average expression level of P21 astrocyte GR-activated and GR-repressed genes identified by bulk RNA-seq (DESeq2, Padj\u2009<\u20090.05, abs(log2(FC)) > 0.1) was calculated for each cell in this snRNA-seq dataset using the Seurat AddModuleScore function. The module score expression of GR-activated and GR-repressed genes in astrocytes from different brain regions were shown by UMAP (Fig. 3e<\/a>) and line plot (Extended Data Fig. 12i,j<\/a>). To identify genes that change expression across astrocyte development de novo, pseudobulk differential expression analysis comparing the P32 and P4 time point for each brain region was performed using DESeq2, using the following parameters: Padj\u2009<\u20090.05, abs(log2(FC)) > 1. Significant genes were then intersected with P21 astrocyte GR-activated and GR-repressed genes identified by bulk RNA-seq (DESeq2 Padj\u2009<\u20090.05), and the number of overlapping genes was shown in Extended Data Fig. 12h<\/a>.<\/p>\n Human brain multiome analysis<\/p>\n Human brain single-nucleus multiome data42<\/a> containing 232,328 cells across 38 samples spanning whole cortex (first trimester only), V1, and PFC from 5 time points (first trimester, second trimester, third trimester, infancy and adolescence) were accessed from Dryad (https:\/\/doi.org\/10.5061\/dryad.2280gb612<\/a>) and loaded into Signac\/Seurat. Peaks were re-called on the entire dataset after grouping cells by time point using MACS2 (CallPeaks), and the presence of transcription factor binding motifs (JASPAR 2020 database) within these peaks was assigned (AddMotifs). ChromVAR was used to calculate the background-corrected, genome-wide accessibility of all transcription factor binding motifs within individual cells, as done previously for SHARE-seq. Differential ChromVAR transcription factor motif scores between adolescence and first trimester cells for each cell type were calculated using MAST within the FindMarkers function. For cases in which a cell type at adolescence was not present in the first trimester, the precursor cell type was used for analysis (for example, for L2\/3 IT, L4IT, L5IT and L6IT neurons, the corresponding first trimester cell types were radial glia, intermediate progenitor cells, newborn excitatory neurons, and immature IT neurons). The top 2 unique differentially accessible transcription factor motifs (by Padj value) between adolescence and first trimester for each cell type are shown in a heat map in Fig. 3g<\/a>, in which each row is scaled to the maximal within-row \u2212log10(Padj) value.<\/p>\n To identify human astrocyte maturation gene programs, pseudo-bulk differential expression analysis between adolescent and first trimester astrocytes was performed using DESeq2 with the following parameters: Padj\u2009<\u20090.01, abs(log2(FC)) > 1. Differentially expressed genes were then hierarchically clustered (shown in Fig. 3i<\/a> and Extended Data Fig. 14f<\/a>), on the basis of mean gene expression level across time points, using the ward.D2 algorithm in pheatmap and split into 4 clusters with cutree\u2009=\u20094. Human astrocyte differentially expressed genes were then overlapped with P21 mouse astrocyte GR-regulated genes (DESeq2, Padj\u2009<\u20090.05) with known human orthologues. Signac LinkPeaks was used to identify ATAC peaks between 5,000 to 300,000\u2009bp of the TSS that positively correlate (Pearson\u2019s correlation, P\u2009<\u20090.001) in accessibility with the expression level of genes in each cluster. Motif enrichment analysis was performed on correlated ATAC peaks using Signac FindMotifs, and the top 5 enriched motifs for each gene cluster are shown in Extended Data Fig. 14g<\/a>. To identify ATAC sites that change accessibility across human astrocyte development, pseudo-bulk differential accessibility analysis between adolescent and first trimester astrocytes was performed using DESeq2 with the following parameters: Padj\u2009<\u20090.05, abs(log2(FC)) > 1. Differentially accessible regions were then hierarchically clustered (shown in Fig. 3j<\/a> and Extended Data Fig. 14j<\/a>), on the basis of mean chromatin accessibility across time points, using the ward.D2 algorithm in pheatmap and split into 4 clusters with cutree\u2009=\u20094. Human astrocyte differentially accessible regions were then overlapped with P14 mouse astrocyte GR binding sites that were lifted over from mm10 to hg38 using UCSC LiftOver.<\/p>\n snRNA-seq data processing and analysis<\/p>\n snRNA-seq data were processed using the Cell Ranger Count pipeline (v8.0.1) with a custom mm10 gtf file containing the eGFP sequence for viral read mapping. Default parameters were used to align reads, count transcripts, and demultiplex nuclei. Individual Cell Ranger output files for each sample were loaded into Seurat (Read10X), and nuclei with fewer than 500 gene counts were removed. DoubletFinder94<\/a> was then used to predict and remove doublets from each sample, and nuclei with greater than 70,000 gene counts were further removed. Seurat objects for each individual sample were then merged, and downstream normalization, scaling, dimensionality reduction, and Louvain clustering steps were performed, as described above for SHARE-seq, to identify principal cell types and excitatory and inhibitory neuron subtypes. During clustering, nuclei clusters lacking significant marker genes (FindMarkers) were removed, resulting in the final processed dataset (102,330 cells). Pseudobulk differential expression analysis between NR GR-con and GR-KO samples for each cell type was performed using the edgeR glmLRT test (Padj\u2009<\u20090.05). Differentially expressed genes in excitatory neuron subtypes were separated into cell type-shared (differential in \u22652 cell types) or cell-type-specific (differential in 1 cell type), and ClusterProfiler was used to identify enrichment of GO biological process terms in the differentially expressed genes shared across excitatory neuron subtypes (enrichGO, Padj\u2009<\u20090.05) (Fig. 4k,l<\/a>). To compare the expression of astrocyte GR-dependent genes between NR and DR conditions, the scaled, mean expression of NR astrocyte GR-dependent genes (edgeR glmLRT Padj\u2009<\u20090.05) was calculated for each experimental group (NR GR-con, NR GR-KO, DR GR-con and DR GR-KO) and visualized on box plots in Fig. 4n<\/a>.<\/p>\n Reporting summary<\/p>\n