Direct neuronal reprogramming of mouse astrocytes is associated with multiscale epigenome remodeling and requires … – Nature.com

Transgenic mice

All experimental procedures for this study were performed at the Biomedical Center, LMU Munich, in accordance with German and European Union guidelines and were approved by the government of Upper Bavaria. Primary cultures of mouse astrocyte were obtained from the cortex of R26-M2rtTA and Yy1tm2Yshi (ref. 61) mice of P56 days of age. R26-M2rtTA (no. 006965) and Yy1tm2Yshi (no. 014649) mice were obtained from The Jackson Laboratory. The mice were not selected based on their gender. The mice were fed ab libitum; housed in individually ventilated cage systems in a room with a temperature of 22C2C, 55%10% humidity and a 12-h/12-h light/dark cycle; and maintained under specific pathogen-free conditions.

Astrocytes were isolated4,32 by dissecting three postnatal mice (P56), and both the gray and white matter of the cerebral cortex were isolated, after removing the subventricular zone, striatum and hippocampus. The cortical meninges were also removed. The cortical tissue was mechanically dissociated, and the cell suspension was centrifuged at 300g, 4C, for 5min. The cell pellet was resuspended in astrocyte medium consisting of DMEM/F12 (1:1) with GlutaMAX (Thermo Fisher Scientific), 10% FBS, penicillinstreptomycin (Gibco), glucose (Gibco), 1 B27 serum-free supplement (Gibco), 10ngml1 epidermal growth factor (EGF, Gibco) and 10ngml1 basic fibroblast growth factor (bFGF, Gibco). The resulting cell suspension was plated onto a T-25 flask. The primary astrocyte culture was maintained in an incubator for 7d at 37C and 5% CO2. Thereafter, the cells were passaged using 0.05% trypsin/EDTA (Thermo Fisher Scientific) and plated onto the following poly-d-lysine (PDL) (Sigma-Aldrich) coated surfaces for the following experiments: 50,000 cells per well in a 24-well plate in 500l of media for immunocytochemistry; 200,000 cells per six-well plate for bulk-RNA-seq, bulk-ATAC-seq, 10x multiome and 10x single-cell RNA-seq experiments; and 1,000,000 cells per T-25 flask for ChIP-seq.

The plasmid FUW-TetON was modified to insert Gateway cloning sites. Mouse Ngn2, eGFP and 9S-A Ngn2 (referred to as PmutNgn2, which was a gift from A. Philpott)25 were cloned into the Gateway entry vectors (Thermo Fisher Scientific) and subsequently shuttled into the dox-inducible lentiviral expression vector FUW-TetON by employing Gateway recombination cloning technology (Thermo Fisher Scientific). The lentiviral expression vector was characterized by the presence of a tetracycline response element followed by the mammalian CMV2 promoter, which regulated the expression of the TFs and the eGFP (fluorescent reporter employed to identify transduced cells). The TF sequence was separated from the eGFP sequence by an internal ribosome entry site (IRES).

Vesicular stomatitis virus-glycoprotein (VSV-G)-pseudotyped lentiviral particles were produced by transfecting 293T cell line with the following plasmids: pCMVdR8.91 (expressing gag, pol and rev genes), pVSVG and lentiviral expression plasmid. The lentiviral particles were harvested and concentrated by ultracentrifugation at 125,000g for 2h, and the pellet containing the lentiviral particles was resuspended in 1 PBS (supplemented with 5mM MgCl2). The lentivirus was aliquoted and stored at 80C until use. The lentiviral titer was determined by a functional assay, where primary mouse astrocytes were infected with the lentivirus preparation at various dilutions, and the number of successfully infected cells was determined by immunostaining the transduced cells with an anti-GFP antibody (for TF-encoding lentiviruses) or an anti-RFP antibody (for Cre-expressing lentivirus). The viral titers used in all the experiments were in the range of 1010 to 1012 transducing units per milliliter.

After seeding the desired number of cells in PDL-coated plates, 24h later the cells were transduced with 107 to 109 transducing units per microliter of lentiviral particles. Approximately 20h after transduction, the astrocyte medium was replaced with fresh medium containing DMEM/F12 (1:1), supplemented with penicillinstreptomycin, glucose, 1 B27 and GlutaMAX (differentiation medium), and the cells were maintained in culture in a 9% CO2 incubator for a period, depending upon the experimental design. To induce the expression of the TF and fluorescent protein, dox (2gml1) was added to the differentiation medium, and the dox-containing medium was added freshly for four consecutive days.

Cells were prepared for fluorescence-activated cell sorting (FACS) by washing them once with 1 PBS followed by trypsinization (0.05% trypsin in EDTA) for 5min. The trypsinization reaction was stopped by adding astrocyte medium. The harvested cells were then washed twice with ice-cold PBS and centrifuged at 300g for 3min at 4C. The cells were resuspended in DMEM/F-12 (1:1), and a single-cell suspension was generated using a 40-m cell strainer. FACS was performed by employing a FACSAria Fusion (BD Biosciences) using a 100-m nozzle. The gating strategy was set by using forward, side scatter and untransduced astrocytes as a negative control and eGFP-expressing astrocytes as a positive control. Additionally, for Methly-HiC, astrocytes were stained for DAPI, and only cells in G0 and G1 (single DNA content) were sorted. The cells were sorted into DMEM/F-12 (1:1).

Coverslips containing astrocytes were fixed using 4% paraformaldehyde in 1 PBS for 10min at room temperature. The cells were washed twice with 1 PBS and stored for up to 3weeks at 4C before staining. The coverslips were incubated with blocking solution (3% BSA, 0.5% Triton X-100 in 1 PBS) for 30min. Thereafter, the coverslips were incubated with the primary antibody diluted (for detailed information about antibodies used, see Supplementary Table 4) using blocking solution overnight at 4C. After washing the coverslips three times with 1 PBS, they were incubated with the appropriate secondary antibody (diluted 1:500) for 1h at room temperature. The coverslips were stained with DAPI (diluted 1:1,000 in blocking solution) for 10min at room temperature. Finally, the coverslips were mounted using Aqua-Poly/Mount (Polysciences).

A Zeiss Cell Observer was employed to perform continuous live imaging of astrocyte-to-neuron conversion. The acquisition of images was performed as follows. Phase contrast images and fluorescent images (GFP) were captured every 20min and 4h, respectively, with a 10 phase contrast objective (Zeiss) and an AxioCam HRm camera. Zeiss AxioVision 4.7 software was controlled by a custom-made VBA module (TAT, Timm Schroeder, ETH Zrich)62. The movie processing and analysis was performed in ImageJ (1.53q) (National Institutes of Health).

The acquisition of microscopy images was performed using an AxioM2 epifluorescence microscope (Zeiss) or an LSM 710 laser scanning confocal microscope (Zeiss) and ZEN2 software (version 2.0.0.0, Zeiss). The quantification of iNs was performed by applying the following stringent criteria, which were previously described in Gascon et al.5. iNs had to possess a unipolar or bipolar morphology, with a process being at least three times the length of its soma. Additionally, the iNs had to be III-tubulin positive and GFAP negative. In case of the live-imaging microscopy, the time of conversion was defined as a timepoint (in hours) when a GFP+ cell acquired neuronal morphologythat is, exhibited a unipolar or bipolar morphology where the process was at least three times the length of its soma. Statistical analysis was performed in R (version 4.2.1). In Figs. 1eh,j and 7c and Extended Data Fig. 1e, statistical significance was calculated with linear regression by implementing the function lm in RStudio on log2-transformed reprogramming rate14.

The primary astrocytes, transduced with the GFP, Ngn2 or PmutNgn2 lentivirus, were obtained from the same litter of mice. In case of the primary astrocytes obtained from the Yy1tm2Yshi line for the functional studies (conditional knockouts of the candidate gene, Yy1), the wild-type, heterozygote and homozygote genotypes were obtained from same litter of mice by crossing two heterozygote mice.

No statistical methods were used to pre-determine samples sizes, but our sample sizes relied on previous experience, showing that this sample size gives sufficient statistical power5,6,17,18,19,45,63. No data were excluded from the analyses. For data in Figs. 1eh,j and 7c, the values were log transformed and, hence, assumed to be normally distributed.

All the data analysis for immunocytochemistry (Figs. 1eh and 7c) and live imaging (Fig. 1j) was blinded. The genomic experiments and associated data analysis were not blinded because they did not involve subjective measurements.

For the Methyl-HiC experiment, the cells were stained with DAPI following the intracellular staining protocol with the following modifications19. Upon fixing with 1% formaldehyde and permeabilizing the cells, they were stained with DAPI (1:1,000 dilution in wash buffer containing 1% BSA, 0.1% RNasin plus RNase inhibitor (Promega) in PBS). The cells were washed once with the wash buffer and subsequently resuspended in PBS with 1% BSA and 1% RNasin plus RNase inhibitor, filtered through a 40-m cell strainer and FACS sorted.

Approximately 30,000 events per condition were FACS sorted into DMEM/F-12 (1:1) and centrifuged at 300g for 5min at 4C. Then, the cell pellet was resuspended in TRIzol (Thermo Fisher Scientific) and further processed with an RNA Clean & Concentrator Kit (Zymo Research) to extract the RNA. The quality of the extracted RNA was determined using an Agilent RNA 6000 Pico Kit and an Agilent 2100 Bioanalyzer system. All the samples used for library preparation had an RNA integrity number (RIN) value>8.

Next, 50ng of RNA was used as the input material for library generation, and the protocol was a bulk adapted version of mcSCRB-seq64,65. cDNA was generated from the poly(A)-enriched RNA fraction using oligo-dT primers and a Maxima First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). The unincorporated primers were digested using Exonuclease I (Thermo Fisher Scientific). The resulting cDNA was pre-amplified using Terra polymerase (Takara Bio). The quality of the cDNA was determined using the Agilent 2100 Bioanalyzer system. The RNA-seq library was prepared using a NEBNext Ultra II FS DNA Library Kit for Illumina (New England Biolabs) according to the manufacturers instructions. The quality of the RNA-seq libraries was assessed using the Agilent 2100 Bioanalyzer system.

Bulk ATAC-seq libraries were generated by following the OMNI-ATAC-seq protocol66. Approximately 70,000 events were FACS sorted into tubes containing DMEM/F-12 (1:1) and centrifuged at 300g for 5min at 4C, and the cell pellet was resuspended in ATAC resuspension buffer. The cell viability and cell number were determined using a Countess automated cell counter (Thermo Fisher Scientific). Fifty thousand viable cells were used for the Tn5 transposition reaction. The transposition reaction was performed at 37C for 30min in an Eppendorf thermomixer. The transposed fragments were purified using a DNA Clean & Concentrator-5 Kit (Zymo Research). The purified transposed DNA fragments were amplified using NEBNext Ultra II Q5 Master Mix (New England Biolabs) and cleaned up using the DNA Clean & Concentrator-5 Kit. The quality of the ATAC-seq libraries was assessed using the Agilent 2100 Bioanalyzer system.

The ChIP-seq protocol was adapted from a previously described protocol67. In brief, 4 million astrocytes were fixed using 1% methanol-free formaldehyde (Thermo Fisher Scientific) at room temperature for 10min. The cross-linking reaction was terminated by the addition of 125mM glycine followed by an incubation step at room temperature for 5min. The cells were lysed by suspension in a hypotonic buffer (20mM Tris, pH 7.4; 2mM MgCl2; 5% glycerol; 0.6% NP-40) and incubation on ice for 5min with mild vortexing every 30s, which resulted in the release of the nuclei. The nuclei were resuspended in ChIP lysis buffer (20mM Tris, pH 7.4; 150mM NaCl; 1% sodium deoxycholate; 0.1% SDS; 1mM EDTA, pH 8.0) and sonicated using a Bioruptor Pico sonicator (Diagenode) with the following settings: 30s ON/OFF, 20 cycles. The sonicated chromatin was quality controlled using the Agilent 2100 Bioanalyzer system. The sonicated chromatin used for ChIP-seq ranged from 150bp to 300bp.

The chromatin was pre-cleared using Dynabeads Protein G (Thermo Fisher Scientific). After pre-clearing, 10% of the pre-cleared chromatin was set aside as the input fraction. The chromatin was incubated with 4g of mouse monoclonal anti-FLAG M2 antibody (Sigma-Aldrich) overnight at 4C on a rotating wheel (10r.p.m.). After the ChIP, Dynabeads Protein G (Thermo Fisher Scientific) was added to the ChIP sample and incubated at 4C for 3h on a rotating wheel (10r.p.m.). The ChIP sample was washed five times with LiCl was buffer (50mM Tris, pH 7.4; 1mM EDTA, pH 8; 1% NP-40; 1% sodium deoxycholate; 0.5M LiCl) followed by a single wash with TE buffer (10mM Tris, pH 8; 1mM EDTA, pH 8). All the wash steps were performed for 5min at 4C on a rotating wheel (10r.p.m.). The elution of the proteinDNA complex was performed using the elution buffer (50mM NaHCO3, 1% SDS) under the following condition: constant agitation on a thermomixer (Eppendorf) at 60g for 15min at 65C. The eluted DNA was de-crosslinked by the addition of 5M NaCl (final concentration: 210mM) and incubated overnight (not more than 15h) at 65C.

The de-crosslinked DNA was treated with RNase A (Thermo Fisher Scientific) and incubated in a thermomixer (Eppendorf) at 60g for 90min at 37C, followed by treatment with Proteinase K (Ambion) and incubated in a thermomixer (Eppendorf) at 800r.p.m. for 120min at 55C. The DNA was extracted using UltraPure Phenol:Choloroform:Isoamylalcohol (25:24:1, v/v, Thermo Fisher Scientific) following the manufacturers instructions and precipitated by ethanol precipitation (glycogen, 3M sodium acetate, pH 5.2, 100% ethanol) overnight at 20C. The DNA was resuspended in low TE buffer and quantified Qubit dsDNA HS (Thermo Fisher Scientific). One nanogram of ChIP DNA was used as starting material for library preparation with the MicroPlex Library Preparation Kit v2 (Diagenode). The quality of the ATAC-seq libraries was assessed using the Agilent 2100 Bioanalyzer system.

The Yy1, FLAG, Rad21 and H3K27Ac CUT&RUN assays were performed as previously described with specific modifications63.

In brief, 23.6105 iNs were harvested, washed twice and resuspended in wash buffer (20mM HEPES, pH 7.5; 150mM NaCl; 0.5mM spermidine; 1 Roche cOmplete). Concavalin A beads (BioMag Plus, Polysciences) were activated with bead activation buffer (20mM HEPES, pH 7.9; 10mM KCl; 1mM CaCl2; 1mM MnCl2). Cells were incubated with 10l of activated beads for 10min at room temperature. After incubation, the beads were resuspended in a cold antibody buffer (2mM EDTA in digitonin buffer) containing antibody (5g of Yy1 (D5D9Z) rabbit monoclonal antibody 46395, 2g of FLAG antibody (Sigma-Aldrich, F3165-.2MG), 5g of Rad21 (BIOZOL, GTX106012) and 1g of H3k27Ac (Abcam, 39133)), and the mixture was incubated on a nutator overnight at 4C.

On the next day, the beads were washed twice and resuspended in 0.75l of pAG-Mnase in digitonin buffer (0.1% digitonin, Thermo Fisher Scientific, in wash buffer) and incubated for 10min at room temperature on a rotator. Later, beads were washed twice with cold digitonin buffer and then resuspended in 50l of digitonin buffer containing 1l of 100mM CaCl2. The suspension was incubated for 2h at 4C on a nutator. After the incubation, 33l of STOP buffer (340mM NaCl, 20mM EDTA, 4mM EGTA, 50gml1 RNase A (Thermo Fisher Scientific), 50gml1 glycogen) was added to each reaction, and the mixture was incubated for 30min at 37C.

DNA extraction was performed using UltraPure Phenol:Chloroform:Isoamyl Alcohol (25:24:1, v/v, Thermo Fisher Scientific) and precipitated with 100% ethanol, 1l of glycogen and 1/10th volume of 3M sodium acetate for 416h at 20C. DNA was then dissolved in 10l of 1mM Tris-HCl, pH 8, and 0.1mM EDTA.

CUT&RUN libraries were prepared with an NEBNext Ultra II DNA Library Prep Kit for Illumina using 630ng of fragmented DNA. The quality of the CUT&RUN libraries was evaluated using the Agilent 2100 Bioanalyzer system.

Single-cell multiome (version 1, 10x Genomics) libraries were generated according to the manufacturers instruction manual. In case of the multiome libraries, we targeted for the recovery of 500 nuclei for the GFP, Ngn2 and PmutNgn2 conditions and 5,000 nuclei for the Astro condition.

A modified Methyl-HiC was performed19 based on previously described protocols17,18. Full details of the experimental steps can be found at https://www.protocols.io/view/methylhic-bif2kbqe/.

Pellets from frozen, fixed and FACS-sorted G0/G1 cells were thawed and then lysed on ice with 0.2% Igepal-CA630 (Sigma-Aldrich). Nuclei were subsequently permeabilized with 0.5% SDS and chromatin digested with DpnII (New England Biolabs) at 37C overnight. DpnII was heat inactivated at 62C, and then sticky ends were filled in with biotin-14-dATP (Life Technologies) before proximity ligation with T4 Ligase (New England Biolabs). Proteinase K (New England Biolabs) and NaCl was used for reverse crosslinking nuclei overnight at 68C, and DNA was afterward purified using ethanol precipitation. A Covaris S220 sonicator was next used to shear the DNA to approximately 550-bp fragments.

End repair was performed on the sonicated DNA with T4 DNA Polymerase (New England Biolabs). Approximately 0.01% of methylation controls were spiked into sample, and the reaction was bisulphite converted using an EZ DNA Methylation-Gold Kit (Zymo Research). Libraries were prepared using an Accel-NGS Methyl-Seq DNA Library kit (Swift Biosciences) according to the manufacturers instructions until the adaptor ligation step. At this point, streptavidin T1 beads (Thermo Fisher Scientific) were used for biotin pulldown of DNA, followed by stringent washes. Final libraries were amplified from the streptavidin beads using EpiMark Hot Start Taq (New England Biolabs) with Methyl-Seq indexing primers (Swift Biosciences), followed by size selection with 0.6 AMPure XP beads (Agencourt).

P19 cells were plated in 10-cm dishes. Cells were transfected using Lipofectamine 3000 with 5g of Control (Pcig2), Neurogenin2 and Neurogenin2 mutated (S-A9 TA1) DNAs and were harvested after 24h by cell scraping using cold PBS followed by centrifugation at 300g for 5min to collect the cell pellets.

The P19 cell pellets were thawed on ice and resuspended in 1 pelleted cell volume of the lysis buffer A (10mM HEPES, pH 7.9; 1.5mM MgCl2; 10mM KCl; 0.1% NP40; 1 protease inhibitor cocktail (Roche, 04 693 116 001); 50mM sodium fluoride; 0.2mM sodium orthovanadate; 0.05mM MG132 (Sigma-Aldrich, M7449); 1mM PMSF). After leaving the resuspended cells for 5min on ice, an equal volume of lysis buffer B (10mM HEPES, pH 7.9; 1.5mM MgCl2; 10mM KCl; 0.1% NP40; 1 protease inhibitor cocktail (Roche); 50mM sodium fluoride; 0.2mM sodium orthovanadate; 0.05mM MG132 (Sigma-Aldrich, M7449); 1mM PMSF) was added to leave another 5min on ice. Cells were lysed by pipetting up and down followed by passing through a 27.5-gauge needle (insulin syringe) for 1012 times on ice. This was followed by centrifugation at 15,000g for 15min, and the supernatant was collected. For in vivo samples, embryonic cortex (dorsal telencephalon) was collected at E12.5 and E14.5 to proceed with protein extraction as above.

IP was performed using 2g of anti-YY1 antibody (mouse anti-YY1; Santa Cruz Biotechnology, sc-7341) and control mouse IgG from in vivo (embryonic cortex) and in vitro (P19 cells) samples. Anti-YY1 antibody was incubated with Protein G Magnetic Dynabeads at 4C for 13h in IP 150 KCl buffer (25mM Tris, pH 7.9; 5mM MgCl2; 10% glycerol; 150mM KCl; 0.1% NP40; 0.3mM DTT; 1 protease inhibitor cocktail (Roche, 04 693 116 001), 50mM sodium fluoride; 0.2mM sodium orthovanadate; 0.05mM MG132 (Sigma-Aldrich, M7449); 1mM PMSF). Then, 0.05% NP40 was added to the protein and centrifuged at 17,530g for 15min. The supernatant was collected and added with 0.1mgml1 ethidium bromide to incubate for 30min, followed by centrifugation at 17,530g for 15min. The supernatant was then pre-cleared with Protein G Dynabeads for 1h by end-over-end rotation at 4C. After pre-clearing, protein was added to the Dynabeads, which were previously incubated with anti-YY1 antibody, followed by overnight rotation at 4C. The supernatant was removed after overnight incubation, followed by four washes using PBS with protease inhibitors (0.3mM DTT; 1 protease inhibitor cocktail (Roche, 04 693 116 001); 50mM sodium fluoride; 0.2mM sodium orthovanadate; 0.05mM MG132 (Sigma-Aldrich, M7449); 1mM PMSF). The proteins bound to the beads were eluted using 2 Laemmli buffer, by heating at 95C for 5min. Proteins were separated from beads using a magnet and proceeded to western blotting to visualize the immunoprecipitated proteins.

The immunoprecipitated proteins were run on 12% SDS-PAGE gels at 70V during stacking and 120V while resolving. The proteins were transferred to PVDF membranes (1620177, Bio-Rad) in transfer buffer (25mM Tris; 192mM glycine; 20% methanol, pH 8.3) at 40V overnight at 4C after the SDS-PAGE. Membranes were blocked in TBST (10mM Tris; 100mM NaCl, pH 7.4; 0.1% Tween 20) with 5% (w/v) skim milk for 1h at room temperature and then incubated with primary antibodies overnight at 4C. Membranes were washed 310min in TBST and then incubated for 1h at room temperature with 1/50,000 dilutions of horseradish peroxidase (HRP)-coupled secondary antibodies (anti-rabbit IgG, 7074S, Cell Signaling Technology). Membranes were washed 310min at room temperature and then processed with ECL Plus Western Blotting Reagent (29018904, GE Healthcare) before developing with X-ray film (1141J52, LabForce) and a Bio-Rad ChemiDoc MP Imaging System. The primary antibodies used were rabbit anti-YY1 (Invitrogen, MA5-32052), rabbit anti-Neurogenin2 (Invitrogen, PA5-78556) and rabbit anti-Ezh2 (Cell Signaling Technology, 5246).

Single-cell multiome reads were aligned to the Mus musculus reference genome (GRCm38, mm10), and the quantification was performed using cellranger-arc-2.0.1. Data were analyzed using Signac (version 1.7.0)68 and ArchR44. The quality control (QC) metrics are reported in Supplementary Table 1.

We eliminated low-information content cells based on the following selection criteria: cells where fewer than 1,000 genes and 1,000 unique molecular identifiers (UMIs) (from the gene expression library) and fewer than 8,000 unique fragments per cell, transcription start site (TSS) enrichment <1 and nucleosome signal <0.2 (from the ATAC library) were detected. To avoid including possible doublets in the further analysis, cells where more than 30,000 genes (from the gene expression library) and more than 125,000 unique fragments, TSS enrichment >20 and nucleosome signal >2 (from the ATAC library) were eliminated. Nucleosome signal and TSS enrichment were calculated using Signac (version 1.7.0)68 and plotted using ggplot2. Fragment lengths were calculated using ArchR44 and plotter using ggplot2. Upon filtering out the low-quality cells from all the conditions, the number of cells from the Astro condition was balanced with the other conditions.

The individual modalities (gene expression and ATAC) were normalized and processed using Signac68 and Seurat (version 4.0)33. In brief, peak calling was performed on pseudobulk aggregate per condition using MACS2. A high-quality union peak set was identified by merging the individual peaks and filtering out peaks, which overlapped with a list of blacklisted regions. The count matrix for the high-quality peak set was generated and incorporated into a Seurat object. It was subjected to TF-IDF normalization followed by SVD as described previously. For the gene expression modality, after log transformation, variance-stabilizing transformation was used to perform feature selection. Principal component analysis was performed using the first 20 dimensions. We then computed a joint neighbor graph that represents both gene expression and chromatin accessibility using FindMultiModalNeighbors. We then applied Louvain clustering to cluster cells (resolution=0.2, n.start=20, n.iter=30, algorithm=1), and the cell clusters were visualized using UMAP (min.dist=0.5, spread=1.5, n.components=2L). Cluster identity was determined based on the top 40 differentially expressed genes (MAST, minimum expression change of 0.25 and expressed by at least 25% of the cells in the cluster)69 as well as known marker genes.

Maturation pseudotime analysis was implemented on the QC-approved cells using Monocle3 (refs. 34,35,36,37). The UMAP coordinates was retained from Seurat and used to build the cds object in Monocle3. Cells in cluster iN_1 were selected as the root cells, and a trajectory graph was constructed using the following parameters: minimal_branch_len=5, maxiter=30. The change in gene expression along the constructed trajectory was calculated by fitting a generalized additive model employing cubic regression splines and REML smoothing. The resulting values were rescaled from 0 to 1.

The calculation of motif accessibility deviation scores using position weight matrices obtained from the JASPAR2000 database and Ngn2 ChIP-seq was performed as described previously19 using the ChromVar implementation in Signac68. TF footprints were calculated using ArchR44 and visualized using ggplot2.

To link putative enhancers with their target genes, we used ArchR with empirical P value estimation and k=50. We distinguish among positively correlated (r>0.35; false discovery rate (FDR)<0.1), negatively correlated (r<0.35; FDR<0.1) and non-correlated pairs (0.35

We reasoned that we can predict direct targets of a TF either by using available ChIP-seq peaks or based on the enrichment of the TF motif in the positively correlated EGPs. First, we identified all EGPs that contained the corresponding ChIP-seq peak or TF motif (either in the distal region or in the promoter region). Thereafter, we calculated the gene linkage score by adding up the r2 from each pair per gene (if the peak/motif was contained in the promoter, we used a value of r=1). To calculate enrichment, we used background ATAC peaks with similar GC content and determined significance using a hypergeometric test. A potential limitation of this method is that the significance of peak/motif enrichment for genes with very few identified pairs cannot be accurately calculated.

The experimental conditions were labeled according to the manufacturers instructions with the following CellPlex reagents from the 3 CellPlex Kit set A (10x Genomics, PN:1000261): Yy1 WT (CMO309), Yy1 KO (CMO310), Yy1 WT/Ngn2+ (CMO311) and Yy1 KO/Ngn2+ (CMO312). Approximately 25,000 events per condition were FACS sorted (Yy1 WT; untransduced, Yy1 KO; RFP+, Yy1 WT/Ngn2+; GFP+, Yy1 KO/Ngn2+; RFP+GFP+) into an Eppendorf tube. Approximately 33,000 cells were loaded onto a Chromium Next GEM ChIP G (10x Genomics, PN:2000177) to obtain a targeted cell recovery of 20,000 cells. The gene expression library (PN:3000431, single cell 3 v3) and the cell multiplexing library (PN:3000482) were prepared according to the manufacturers protocol (CG000388, Rev A). The gene expression library and the cell multiplexing library were quality controlled using the Agilent 2100 Bioanalyzer, and the libraries were sequenced according to the manufacturers specifications.

Single-cell RNA-seq reads were aligned to the Mus musculus reference genome (GRCm38, mm10), and the sample assignment and quantification were performed using cell ranger multi in cellranger-6.0.0. The QC metrics are reported in Supplementary Table 1.

We eliminated low-information content cells based on the following selection criteria: cells where fewer than 1,000 genes and 2,500 UMIs were detected. To exclude dead cells, we filtered out cells containing more than 20% mitochondrial reads. To avoid including doublets in the further analysis, cells containing more than 6,000 genes were excluded.

Seurat (version 4.0) was used to analyze the cells that passed the filtering steps. The data were normalized using SCTransform, and principal component analysis was performed using the first 25 dimensions. We applied Louvain clustering (resolution=0.6, n.start=20, n.iter=20), and the data were visualized by UMAP projection (min.dist=0.5, spread=1.5, n.components=2L). Cluster identity was determined based on the top 40 differentially expressed genes (MAST, minimum expression change of 0.25 and expressed by at least 25% of the cells in the cluster)69.

The ATAC-seq FASTQ files were demultiplexed using Je (version 1.2)70, and the demultiplexed reads were aligned to the mouse genome (GRCm38, mm10). Post-alignment read filtering, peak calling and irreproducible discovery rate (IDR)-based peak filtering were performed by implementing the ENCODE ATAC-seq pipeline. The sequencing and QC metrics are listed in the form of a supplementary data table. The bigWig coverage track was generated using deepTools (version 3.1.3)71. The plotting of the ATAC-seq signal at genomic features was performed using SeqPlots72. The QC metrics are reported in Supplementary Table 1.

The RNA-seq FASTQ files were demultiplexed using Je (version 1.2)70; demultiplexed reads were aligned to the mouse genome (GRCm 38, mm10) using STAR (version 2.7.1a)73; and read counts per gene were obtained by using the quantMode GeneCounts option. Further analysis was performed using DEseq2 (ref. 74) in RStudio. The result table for pairwise comparison between PmutNgn2 versus Ngn2 was used the input to generate the GO term enrichment bubble plot in the R package clusterProfiler75. The QC metrics are reported in Supplementary Table 1.

The ChIP-seq FASTQ files were demultiplexed using Je (version 1.2)70; demultiplexed reads were aligned to the mouse genome (GRCm38, mm10); and post-alignment read filtering, peak calling and IDR-based peak filtering were performed by implementing the ENCODE ChIP-seq pipeline. The bigWig coverage track was generated using deepTools (version 3.1.3)71. The plotting of the ChIP-seq signal at genomic features was performed using the R package SeqPlots72. The QC metrics are reported in Supplementary Table 1.

CUT&RUN data were uniformly processed using CUT&RUN tools 2.0 (ref. 76). Peaks were called using MACS2, and the bigWig coverage track was generated using deepTools (version 3.1.3)71. The QC metrics are reported in Supplementary Table 1.

FASTQ files from the Methyl-HiC were mapped to the mouse genome (GRCm38, mm10) by employing JuiceMe77. Further analysis was performed only with uniquely mapping reads (mapq score>30). After the elimination of polymerase chain reaction (PCR) duplicates, the translation of reads into a pair of fragment ends (fends) was achieved by the association of each read with its downstream fend. MethylDackel was used to assess CpG methylation, which entailed the elimination of the initial six nucleotides in the mergeContext mode. Pooling of reads from individual replicates was performed, and, for a cytosine to be considered for further analysis, it had to be in the CpG context and possess at least 10 total coverage. In case of Hi-C, exclusion of reads was based on the following criteria: mapped to the same restriction fragment and separated by less than 1kb. The QC metrics are reported in Supplementary Table 1.

Filtered fend-transformed read pairs were imported into the following genome database: mm10 after conversion into misha tracks. The Shaman package was used for read pair normalization (https://tanaylab.bitbucket.io/shaman/index.html), and the calculation of the Hi-C score was performed by employing k-nearest neighbors (kNN)22.

The calculation of contact probabilities as a function of genomic distance was previously described22. The insulation score, which is used to define insulation on the basis of observed contacts, was also previously described 19,22,78, and differential TAD boundaries were identified using insulation score19,22.

The calculation of contact matrices dominant eigenvector, which have been binned at 250kb, was previously described and performed using publicly available scripts (https://github.com/dekkerlab/cworld-dekker)79. Compartment strength was determined by plotting the log2 ratio of observed versus expected contacts (intrachromosomal separated by at least 10Mb) between AA, BB or AB domains. A ratio between the sum of observed contacts within the A and B compartments and the sum of intercompartment contacts was calculated to determine the compartment strength19.

The calculation of insulation and contact enrichment within TADs was previously described19,22.

Two complementary approaches were employed for the calculation of contact enrichment ratio at genomic feature pairs, such as Ngn2 ChIP-seq peaks or EGPs. Aggregated Hi-C maps were used to calculate the log2 ratio of observed versus expected contacts within a specified window size, which was centered on the feature of interest. The average enrichment ratio was also calculated for the following: contact strength in the center of the window versus each of the corners. Furthermore, the extraction of kNN-based Hi-C score for each pair in a 10-kb window centered around it and its representation as a scatter plot or box plots enabled the identification of pair-specific trends. Significance testing was performed by using the Wilcoxon rank-sum test.

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

Here is the original post:
Direct neuronal reprogramming of mouse astrocytes is associated with multiscale epigenome remodeling and requires ... - Nature.com