Control of neuronal excitation–inhibition balance by BMP–SMAD1 signalling

Mice

All procedures involving animals were approved by and performed in accordance with the guidelines of the Kantonales Veterinäramt Basel-Stadt. All experiments were performed in mice on a C57Bl/6J background, except for some of the experiments performed in cultured wild-type neurons, which used RjOrl:SWISS mice (Janvier). All mice were group housed (weaning at P21–P23) under a 12-h light–dark cycle (06:00–18:00) at 21–24 °C and 50–60% humidity with food and water ad libitum. Both males and females were used at similar numbers for the experiments. Mice were randomly assigned to treatment groups. Mice that exhibited a spontaneous seizure were excluded from molecular, anatomical and slice physiology analyses.

Smad1^fl/fl mice⁵⁴, Pvalb-cre mice⁵⁵ and Ai9 mice⁵⁶ were obtained from Jackson Laboratories (Jax stock no: 008366, 017320 and 007909, respectively). Cux2-CreERT2 mice⁵⁷ were obtained from the Mutant Mouse Resource and Research Center (MMRRC). Bmp2-2xHA mice were generated using a CRISPR–Cas9 strategy⁵⁸ inserting a double HA tag at the N terminus of the mature BMP2 protein, between amino acids S292 and S293. The guide RNAs (gRNAs) used were 5′-GTCTCTTGCAGCTGGACTTG-3′ and 5′-CAAAGGGTGTCTCTTGCAGC-3′, together with a 200-bp single-stranded DNA ultramer: 5′-GACTTTTGGACATGATGGAAAAGGACATCCGCTCCACAAACGAGAAAAGCGTCAAGCCAAACACAAACAGCGGAAGCGCCTCAAGTCCGCTAGCTACCCATACGATGTTCCAGATTACGCTGGCTATCCCTATGACGTCCCGGACTATGCAGCTAGCAGCTGCAAGAGACACCCTTTGTATGTGGACTTCAGTGATGTGG-3′ (the sequence encoding the HA tags is highlighted in bold).

Surgery and drug treatments

Injections of recombinant AAVs were performed into the barrel cortex of 42–49-day-old male and female mice performed under isoflurane anaesthesia (Baxter). Mice were placed in a stereotactic frame (Kopf) and a small incision (0.5–1 cm) was made over the barrel cortex at the following coordinates targeting two sites: mediolateral (ML) ±3.0 mm and ±3.4 mm, at anteroposterior (AP) 0.6 mm and AP −1.6 mm, dorsoventral (DV) –1.5 mm from Bregma to target layers 2/3 and 4. For injections of FingR intrabodies, two injection sites restricted to layer 2/3 were used: ML ±3.0 mm and ±3.4 mm at AP –1.0 mm, DV –0.96 mm from Bregma. Recombinant AAVs (titre: 10¹²–10¹³) were injected via a glass capillary with an outer diameter of 1 mm and an inner diameter of 0.25 mm (Hilgenberg) for a total volume of 100 nl per injection site. The wound was closed with sutures (Braun, C0766194).

LMI070 (25 mg kg⁻¹, MedChemExpress, HY-19620, suspended in 20% cyclodextrin and 10% dimethyl sulfoxide (DMSO) to 5 mg ml⁻¹ concentration) was administrated by oral gavage. Clozapine N-oxide (CNO) (5 mg kg⁻¹, Sigma Aldrich, C0832) and doxycycline (50 mg kg⁻¹, Thermo Fisher Scientific, BP26531, suspended in 0.9% NaCl to 5 mg ml⁻¹ concentration) were administered by intraperitoneal injection.

Antibodies and probes

Primary antibodies were: monoclonal mouse anti-synaptotagmin-2 (Zebrafish International Resource Center, ZNP-1), rabbit anti-SMAD1 (Cell Signaling 6944, 1:100 for ChIP and 1:1,000 for western blot), H3K27ac (Abcam 4729, 1:1,000), rabbit anti-SMAD5 (Cell Signaling, 12534, 1:100 for ChIP and 1:1,000 for western blot), anti-phospho-SMAD1/5/9 (Cell Signaling 13820, 1:1,000), mouse anti-BMPR2 (BD Pharmingen, 612292, 1:1,000), rabbit anti-calnexin (StressGen, SPA-865, 1:2,000), mouse anti-MAP2 (Synaptic Systems, 188011, 1:1,000), mouse anti-CAMKII alpha (Thermo Fisher Scientific, MA1-048, 1:800), rat anti-GAPDH (Biolegend, 607902, 1:10,000), rabbit anti-NeuN (Abcam, ab177487, 1:500), mouse anti-GAD67 (Millipore MAB5406, 1:500), rabbit anti-vGLUT1 (Synaptic Systems 135303, 1:5,000), biotinylated WFA (Vector Laboratories B-1355-2, 1:500), rabbit anti-HA (Cell Signaling 3724, 1:1,000), mouse anti-GFP antibody (Santa Cruz, sc-9996, 1:1,000) and goat anti-parvalbumin antibody (Swant PVG213, 1:5,000). Secondary antibodies were: HRP goat anti-rabbit (Jackson 111-035-003, 1:10,000), HRP goat anti-rat (Jackson 112-035-143, 1:10,000), HRP goat anti-mouse (Jackson 115-035-149, 1:10,000), Alexa405 goat anti-rabbit (Thermo Fisher Scientific A-31556, 1:500), Alexa488 donkey anti-rabbit (Thermo Fisher Scientific R37118, 1:1,000), Alexa647 donkey anti-mouse (Jackson 715-605-151, 1:1,000), Alexa647 streptavidin (Thermo Fisher Scientific, S32357, 1:1,000), Cy2 Streptavidin (Jackson 016-220-084, 1:1,000), Cy3 donkey anti-mouse (Jackson 715-165-151, 1:500), Cy3 donkey anti-rabbit (Jackson 711-165-152, 1:500), Cy5 donkey anti-goat (Jackson 705-175-147, 1:500), Cy5 donkey anti-rabbit (Jackson 711-175-152, 1:500) and Cy5 donkey anti-mouse (Jackson 715-175-511, 1:500). DAPI dye was used for nuclear staining (TOCRIS Bio-Techne, 5748, 1:5,000).

Immunohistochemistry and image acquisition

Mice were deeply anaesthetized with a ketamine–xylazine mix (100 and 10 mg per kg, respectively, intraperitoneally) and were transcardially perfused with fixative (4% paraformaldehyde (PFA) in 0.1 M phosphate buffer, pH 7.4). For synapse quantifications with FingR probes the fixative also contained 15% picric acid. After perfusion, brains were post-fixed overnight in fixative at 4 °C and washed three times with 100 mM phosphate buffer.

For quantifications of parvalbumin and WFA expression and BRX reporter analyses, coronal brain slices were cut at 40 µm with a Vibratome (VT1000S, Leica). For FingR-PSD95 analysis with the Cre-dependent reporter, coronal brain slices were cut at 30 µm with a Cryostat (Microm HM560, Thermo Fisher Scientific). Brain sections were incubated for 30 min in blocking solution (0.3% Triton X-100 and 3% bovine serum albumin in phosphate-buffered saline (PBS)). Sections were incubated with primary antibodies in blocking solution overnight at 4 °C and washed three times (10 min each) with 0.05% Triton X-100 in PBS, followed by incubation for 1.5 h at room temperature with secondary antibodies in blocking solution. Sections were washed three times with PBS and DAPI dye (1.0 µg ml⁻¹) co-applied during the wash. Sections were mounted using Microscope cover glasses 24 × 60 mm (Marienfeld Superior 0101242) on Menzel-Gläser microscope slides Superfrost Plus (Thermo Fisher Scientific, J1800AMNZ) with ProLong Diamond Antifade Mountant (Invitrogen, P36970).

For S5E2 PV enhancer FingR-PSD95 quantifications, coronal brain slices were cut at 120 µm on a Vibratome (VT1000S, Leica) and cleared with CUBIC-L solution (10% w/v N-butyldiethanolamine, 10% w/v Triton X-100) for 3 h at 37 °C with gentle shaking⁵⁹. Sections were stained with goat anti-parvalbumin antibodies and mounted with Ce3D Tissue Clearing Solution (Biolegend, 427704).

For parvalbumin and WFA analysis, images were acquired on an inverted LSM700 confocal microscope (Zeiss) using 20×/0.45 and 40×/1.30 Apochromat objectives. For quantifications of the cell density of PV interneurons, tile-scan images from the barrel cortex were acquired. For synapse quantifications, images were acquired with a PlanApo 63×/1.4 oil immersion objective.

For primary neocortical neurons in culture, fixation was with 4% PFA in 1× PBS for 15 min. followed by ice-cold methanol (10 min at −20 °C). Cells were blocked (5% donkey serum, 0.3% Triton X-100 in PBS) for 1 h at room temperature and primary antibody incubation was performed overnight at 4 °C in a humidified chamber. Secondary antibody incubation was 1 h at room temperature. Imaging was performed on a widefield microscope (Deltavision, Applied Precision) with a 60× objective (NA 1.42, oil).

Image analysis

Mean intensity analyses for parvalbumin and WFA stainings were performed in ImageJ with a custom-made script in Python. In brief, H-Watershed was applied to segment PV interneurons on the basis of the tdTomato signal on the soma. To detect the WFA signal, the soma was eroded and dilated in all optical sections. After applying thresholding, parvalbumin and WFA mean intensity values were automatically calculated and displayed as arbitrary units. Integrity analysis of PNNs was done from PV interneurons with a positive WFA signal (>2,000 arbitrary units). Images were post-processed by conservative deconvolution with the Huygens Deconvolution software with the classic maximum likelihood estimation deconvolution algorithm. Quantitative analyses of the number of peaks and the distance between the peaks were performed by using plot profile function in ImageJ as described⁶⁰.

For BRX-reporter experiments, cell identity and reporter intensity were quantified with ImageJ. A region of interest was drawn around the nuclei (marked by DAPI) and the mean intensity was measured for the nuclear GFP signal and normalized to background fluorescence in the same image. Cells were identified on the basis of immunostaining for markers: mCherry (genetically restricted to PV interneurons), NeuN (marking neurons with high intensity in pyramidal cells) and GAD67 (marking all GABAergic cells).

For synapse quantification, images were post-processed by conservative deconvolution with the Huygens Deconvolution software with the classic maximum likelihood estimation deconvolution algorithm. Quantitative analysis was performed using Imaris 9.9.1 by application of spots and surface detection tool.

All data collection and image analysis were done blinded to the genotype or treatment of the mouse. Statistical analyses were done with GraphPad Prism v.9. Images were assembled using ImageJ and Adobe Illustrator software.

ChIP–seq analysis

For ChIP–seq analysis with cultured neurons, 24 × 10⁶ cells (DIV14) were cross-linked with 1% formaldehyde for 10 min at room temperature. Cross-linking was stopped by the addition of glycine solution (Cell Signaling Technology, 7005) for 5 min at room temperature. Cells were scraped, pelleted and lysed for 10 min on ice in 100 mM HEPES-NaOH pH 7.5, 280 mM NaCl, 2 mM EDTA, 2 mM EGTA, 0.5% Triton X-100, 1% NP-40 and 20% glycerol. Nuclei were pelleted by centrifugation, washed in 10 mM Tris-HCl pH 8.0, 200 mM NaCl and suspended in 10 mM Tris-HCl pH 8.0, 100 mM NaCl, 1 mM EDTA, 0.5 mM EGTA, 0.1% Na-Deoxycholate and 0.5% N-lauroylsarcosine. Chromatin was sheared using a Covaris Sonicator for 20 min in sonication buffer (SimpleChIP Plus Sonication Kit, Cell Signaling Technology, 57976) to obtain fragments in the range of 200–500 bp. After sonication, sheared chromatin was centrifuged at 16,000g for 20 min at 4 °C and dissolved in 1× ChIP buffer (Cell Signaling Technology, 57976). Input (2%) was taken and the chromatin was incubated with antibodies overnight at 4 °C. Incubation with Protein G magnetic beads, de-cross-linking and elution were performed as described in the SimpleChIP Plus Sonication Kit.

Libraries were generated using the KAPA Hyper Prep (Roche KK8504) according to the manufacturer’s instructions, and were amplified by PCR. Library quality was assessed using the High Sensitivity NGS Fragment Analysis Kit (Advanced Analytical DNF-474) on the Fragment Analyzer (Advanced Analytical). Libraries were sequenced paired-end 41 bases on NextSeq 500 (Illumina) using two NextSeq 500 High Output Kit 75-cycles (Illumina, FC-404-1005) loaded at 2.5 pM and including 1% PhiX. Primary data analysis was performed with Illumina RTA v.2.4.11 and Basecalling v.bcl2fastq-2.20.0.422. Two NextSeq runs were performed to compile enough reads (on average per sample in total: 50 million ± 2 million pass-filter reads).

ChIP–seq analysis from P35–P42 mouse cortex was performed using the SimpleChIP Enzymatic Chromatin IP Kit (Cell Signaling Technology, 9003), following the manufacturer’s instructions with slight modifications. In brief, neocortices from both hemispheres were cross-linked in 1.5% formaldehyde for 20 min at room temperature. Cross-linking was stopped by the addition of glycine solution for 5 min at room temperature. Tissue was pelleted, washed and disaggregated by using a Dounce homogenizer in 1× PBS containing protease inhibitor cocktail. Nuclei were pelleted by centrifugation and chromatin was digested by using micrococcal nuclease for 20 min at 37 °C by frequent mixing to obtain fragments in the range of 150–900 bp. Nuclei were pelleted, resuspended in 1× ChIP buffer, sonicated with Bioruptor Pico (Diagenode B01060010) to release sheared chromatin and centrifuged at 9,400g for 10 min at 4 °C. Input (2%) was taken and the chromatin was incubated with primary antibodies overnight at 4 °C. After subsequent incubation with 30 μl Protein G magnetic beads for 2 h at 4 °C, beads were washed three times with low salt, one time with high salt, one time with NP-40 buffer (8 mM Tris-HCl pH 8.0, 2 mM LiCl, 0.8 mM EDTA, 0.4% NP-40 and 0.4% sodium-deoxycholate) and one time with TE buffer (10 mM Tris-HCl, pH 8.0 and 1 mM EDTA) at 4 °C. De-cross-linking and elution were performed as described in the Enzymatic Chromatin IP Kit. Libraries were generated using the NEBNext Ultra II DNA Library Prep Kit for Illumina (New England Biolabs, E7645L) according to the manufacturer’s instructions and were amplified by PCR. Library quality was assessed using the High Sensitivity NGS Fragment Analysis Kit (Advanced Analytical, DNF-474) on the Fragment Analyzer (Advanced Analytical) and cleaned up by using 1.0× Vol SPRI beads (Beckman Coulter). Libraries were sequenced paired-end 41 bases on NextSeq 500 (Illumina) using two NextSeq 500 High Output Kit 75-cycles (Illumina, FC-404-1005). Two NextSeq runs were performed to compile enough reads (19–32 million pass-filter reads).

RNA library preparation and sequencing

Libraries of BMP2-stimulated naïve cortical cultures were prepared from 200 ng total RNA by using the TruSeq Stranded mRNA Library Kit (20020595, Illumina) and the TruSeq RNA UD Indexes (20022371, Illumina). Fifteen cycles of PCR were performed.

Quality checking was performed by using the Standard Sensitivity NGS Fragment Analysis Kit (DNF-473, Advanced Analytical) on the Fragment Analyzer (Advanced Analytical) and quantified (average concentration was 213 ± 15 nmol l⁻¹ and average library size was 357 ± 8 bp) to prepare a pool of libraries with equal molarity. The pool was quantified by fluorometry using using the QuantiFluor ONE dsDNA System (E4871, Promega) on a Quantus instrument (Promega). Libraries were sequenced single-reads 76 bases (in addition: 8 bases for index 1 and 8 bases for index 2) on NextSeq 500 (Illumina) using the NextSeq 500 High Output Kit 75-cycles (Illumina, FC-404-1005). Flow lanes were loaded at 1.4 pM of pool and including 1% PhiX. Primary data analysis was performed with Illumina RTA v.2.4.11 and Basecalling v.bcl2fastq-2.20.0.422. The NextSeq runs were performed to compile, on average per sample, 56 million ± 3 million pass-filter reads (illumina PF reads).

For the libraries from control and Smad1 mutant primary cortical cultures (four biological replicates), 100 ng total RNA was used and library preparation and quality check were performed as described above. Quantification yielded an average concentration of 213 ± 15 nmol l⁻¹ and an average library size of 357 ± 8 bp. Libraries were sequenced paired-end 51 bases (in addition: 8 bases for index 1 and 8 bases for index 2) set-up using the NovaSeq 6000 instrument (Illumina). SP Flow-Cell was loaded at a final concentration in flow lanes of 400 pM and including 1% PhiX. Primary data analysis was performed as described above and 43 million ± 5 million per sample (on average) pass-filter reads were collected on 1 SP Flow-Cell.

ChIP–seq and RNA-seq data analysis

ChIP–seq reads were aligned to the December 2011 (mm10) mouse genome assembly from UCSC⁶¹. Alignments were performed in R using the qAlign function from the QuasR package1 (v.1.14.0) with default settings⁶². This calls the Bowtie aligner with the parameters “-m 1 –best –strata”, which reports only reads that map to a unique position in the genome. The reference genome package (BSgenome.Mmusculus.UCSC.mm10) was downloaded from Bioconductor (https://www.bioconductor.org). BigWig files were created using qExportWig from the QuasR package with the bin size set to 50. Peaks were called for each ChIP replicate against a matched input using the MACS2 callpeak function with the default options. Peaks were then annotated to the closest gene and to a genomic feature (promoter, 3′-UTR, exon, intron, 5′-UTR or distal intergenic) using the ChIPseeker R package. The promoter region was defined as −3 kb to +3 kb around the annotated transcription start site. Transcripts were extracted from the TxDb.Mmusculus.UCSC.mm10.ensGene annotation R package. All analyses in R were run in RStudio v.1.1.447 running R v.3.5.1. The enrichment of BMP2-induced peaks over constitutive peaks was analysed by using default settings in the voom–limma analysis software packages⁶³. Motif enrichment analysis for BMP2-responsive peaks and constitutive peaks was performed separately by screening for the enrichment of known motifs with the default settings of HOMER⁶⁴. Output motif results with the lowest P value and highest enrichment in targets compared to the background sequences were shown for each peak set.

RNA-seq reads were aligned to mm10 using STAR and visualized in the IGV genome browser to determine strand protocol. By using QuasR’s qQCReport, read quality scores, GC content, sequence length, adapter content, library complexity and mapping rate were checked and a QC report was generated. Reads with quality scores less than 30, mapping rates lower than 65 or contaminations from noncoding RNAs were not considered for further analysis. For reads that passed QC, QuasR’s qCount function was used to count the reads that mapped to annotated exons (from Ensembl genome annotations). Each read was counted once on the basis of its start (if reads are on the plus strand) or end (if reads are on the minus strand) position. For each gene, counts were summed for all annotated exons, without double-counting exons present in multiple transcript isoforms (exon-union model). Correlations between replicates and batch structure were checked by plotting correlation heat maps, PCA plots of samples and scatter plots of normalized read counts. The EdgeR package from R was used to build a model and test for differentially expressed (DE) genes. For DE analysis, counts were normalized using the TMM method (built into edgeR). Any genes with fewer than, in total, 30 reads from all samples were dropped from further analysis. DE analyses were conducted with the voom–limma analysis software packages by using the total number of mapped reads as a scaling factor. Results were extracted from edgeR as tables and used for generating volcano or box plots in ggplot2 in RStudio.

To generate IGV genome browser tracks for ChIP–seq and RNA-seq data, all aligned bam files for each replicate of a given experiment were pooled and converted to BED format with bedtools bamtobed and filtered to be coverted into coverageBED format using bedtools. Finally, bedGraphToBigWig (UCSC-tools) was used to generate the bigWig files displayed on IGV browser tracks in the manuscript.

GO analysis was performed by using the statistical overrepresentation test and cellular component function in PANTHER (http://pantherdb.org/). All genes that were detected as expressed in RNA-seq data were used as reference. GO terms with at least ten genes and at least 1.5-fold enrichment with a false discovery rate of less than 0.05 were considered to be significantly enriched. Significant GO terms were plotted in GraphPad Prism v.9.

EEG recordings and behavioural monitoring

EEG electrodes were implanted in mice at the age of 12–16 weeks. EEG signals were recorded using two stainless steel screws inserted ipsilaterally into the skull. One was inserted 1.2 mm from the midline and 1.5 mm anterior to bregma, and the other was inserted 1.7 mm from the midline and 2.25 mm posterior from to bregma. Seven days after surgery, mice were transferred to individual behaviour cages with a 12:12 h light–dark cycle and a constant temperature of about 23 °C. Mice had access ad libitum to food and water and were allowed to recover from surgery for seven days. Analysis was performed in individual cages equipped with overhead cameras (FLIR). Mice were connected to an amplifier (A-M Systems 1600) through a commutator. EEG signals were amplified and analog filtered (Gain 500; low-pass filter, 0.3 Hz; high-pass filter, 100 Hz) and then digitized at 200 Hz using Spike2 (CED Micro1401). Spontaneous sleep–wake behaviour was monitored continuously through EEG recordings and video tracking for three weeks. Epileptic episodes were identified at first by inspecting the EEG signals, and were subsequently examined further in the simultaneous video recordings.

Statistics and reproducibility

All experiments were performed in at least three fully independent replications (on different days, with different mice or cell cultures). Details about the numbers of mice and cultures are provided in the figure legends. When single micrographs or western blots are shown, their results are representative of all independent replicates analysed. Analysis was conducted in R and with GraphPad Prism v.9. Sample sizes were chosen on the basis of previous experiments and literature surveys. No statistical methods were used to predetermine sample sizes. Exclusion criteria used throughout this manuscript were predefined. See the descriptions in the respective sections of the methods. Mice were randomly assigned to treatment groups. Appropriate statistical tests were chosen according to the sample size and the distribution of data points, and are indicated in individual experiments.

Reporting summary

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