Transcriptomic analysis of loss of Gli1 in neural stem cells responding to demyelination in the mouse brain

In the adult mammalian brain, Gli1 expressing neural stem cells reside in the subventricular zone and their progeny are recruited to sites of demyelination in the white matter where they generate new oligodendrocytes, the myelin forming cells. Remarkably, genetic loss or pharmacologic inhibition of Gli1 enhances the efficacy of remyelination by these neural stem cells. To understand the molecular mechanisms involved, we performed a transcriptomic analysis of this Gli1-pool of neural stem cells. We compared murine NSCs with either intact or deficient Gli1 expression from adult mice on a control diet or on a cuprizone diet which induces widespread demyelination. These data will be a valuable resource for identifying therapeutic targets for enhancing remyelination in demyelinating diseases like multiple sclerosis.


Background & Summary
Multiple Sclerosis (MS) is the most common cause of neurological disability in young adults 1 and is characterized by inflammatory demyelination leading to axonal injury. In addition to preventing immune-mediated demyelination, a major therapeutic goal in MS is to enhance new myelin sheath formation by remyelination which prevents the loss of axons and restores function [2][3][4] . The nervous system is capable of remyelination as observed in autopsy specimens from MS patients characterized by short, thin myelin sheaths [5][6][7][8] . However, remyelination ultimately fails, particularly in MS patients with progressive disease 8,9 and the factors that limit remyelination remain poorly understood 9,10 .
There are three sources of remyelinating cells in the adult brain -neural stem cells (NSCs) [11][12][13] , oligodendrocyte progenitor cells (OPCs) [14][15][16] and mature oligodendrocytes 17,18 . NSCs reside in the subventricular zone (SVZ) while the OPCs and oligodendrocytes are present throughout the parenchyma in the adult brain and are recruited locally to regenerate myelin in demyelinated lesions 15,19 . Furthermore, NSCs can promote remyelination either by replenishing the adult OPCs recruited into lesions, or as a direct source of oligodendrocytes themselves. Consistently, prior studies have shown that NSCs in the SVZ are activated by local demyelination, migrate out and differentiate into OPCs that generate remyelinating oligodendrocytes 11,13,14,20 . Indirect marker studies of postmortem brains further suggest that NSCs proliferate during acute attacks of MS and are a source of remyelinating cells in the human brain 12,21,22 . More importantly, ablating NSCs results in axonal loss highlighting their significant functional contribution to remyelination 23 . Taken together, these studies suggest remyelination by NSCs protects axons. Thus, enhancing remyelination by NSCs is expected to provide a neuroprotective strategy in MS.
The adult SVZ is comprised of a functional, quiescent neural stem cell population, which are multipotent, divide slowly, self-renew, and give rise to proliferating, transit-amplifying cells 24 . There is considerable heterogeneity among these NSCs, indicated by their expression of specific transcription factors and generation of distinct cells during homeostasis and in response to injury 25 . One NSC subset, which is enriched in the ventral SVZ, expresses the transcription factor Gli1 and generates interneurons in the olfactory bulb and astrocytes in the healthy mouse brain 26,27 . This population of NSCs is also present in the human SVZ 27 as well as in NSCs derived from human induced pluripotent stem cells and human embryonic stem cells 28 . Remarkably, and in contrast to its fate in the healthy adult brain, this Gli1 pool of NSCs is specifically recruited to the white matter in response to demyelination and goes on to differentiate into remyelinating oligodendrocytes 20,27,29 . Inhibition of Gli1 in these cells further enhances their remyelination potential and is neuroprotective resulting in functional improvement in mouse models of MS 27,28 . Importantly, Gli1 expressing stem-like cells respond to injury in other organs like the liver, lung, bone marrow, kidney and heart with loss of Gli1 resulting in attenuated injury in these organs [30][31][32][33][34][35][36][37] . Thus, inhibition of Gli1 is potentially an important therapeutic strategy for regenerative medicine and the molecular mechanisms by which Gli1 affects stem cell repair may be shared in many different types of tissues.
We performed a transcriptomic analysis to identify Gli1-regulated genes that affect NSC recruitment and generation of remyelinating oligodendrocytes by comparing the mRNA expression between NSCs with genetic loss of Gli1 and those with intact Gli1 expression from both healthy and demyelinated adult mouse brains (Fig. 1). These data will be valuable not only for future remyelination studies but may also provide mechanistic insights into the shared principles of regeneration across organs.

Methods
Fate-mapping of neural stem cells and demyelination. All the mice had access to food and water ad libitum and were housed in a room with 12 hour dark/light cycle. The genetically modified mice belonged to C57bl/6 strain and were maintained according to protocols approved by New York University Medical Center's IACUC. The Gli1 CreERT2/+ mice (Jax # 007913) which have CreERT2 knocked into the Gli1 locus were bred with the reporter Rosa-CAG-EGFP-LoxP (RCE) (Jax# 032037) mice to generate Gli1 HET (Gli1 CreERT2/+ ;RCE) with intact Gli1 expression and Gli1 NULL (Gli1 CreERT2/CreERT2 ;RCE) mice with global knockout of Gli1. To permanently label the Gli1 pool of NSCs with GFP, we administered 5 mg tamoxifen (Sigma) in corn oil on alternate days for a total of four intraperitoneal injections to 10 week old mice 27,38 . We did not observe any labeling in the absence of tamoxifen administration. Mice used for GFP negative controls received intraperitoneal injections of corn oil without Tamoxifen. A week after the last dose of tamoxifen/corn oil, the demyelinated group was fed 0.2% cuprizone in the chow for 3 weeks while the healthy control group remained on regular chow. Cuprizone diet induces apoptosis of mature oligodendrocytes leading to maximum demyelination of the white matter corpus callosum after 5 weeks of continuous feeding 39 . Here, we used 3 weeks of cuprizone diet for our analysis since we previously found that Gli1 NULL NSCs in the SVZ have significantly higher proliferation at this timepoint compared to Gli1 HET SVZ, suggesting that maximal activation of NSCs coincides with this early stage of demyelination 27,40 . Dissection of SVZ. Brain tissues were harvested from 6 mice (3 Males and 3 Females) per genotype (Gli1 HET and Gli1 NULL ) per diet (cuprizone and control) after euthanizing them with CO 2. The brain was removed gently out of the skull and immediately transferred to a Petri dish chilled on ice and placed with the dorsal surface facing up on a chilled acrylic mouse brain matrix. The brains of mice were dissected according to previously published protocols 38 . Coronal sections of 1 mm thickness were made from the rostral end of the brain up to the dorsal hippocampus posteriorly, using a chilled sharp razor blade. The SVZ was identified as a thin layer of tissue lining the medial and lateral walls of the lateral ventricles and carefully dissected under a microscope from each coronal brain section.

Dissociation and FaCsorting of neural stem cells.
For each RNA sample, the SVZ from one female and one male mice was pooled together. Overall, we had 3 RNA samples/genotype from control diet group and 3 RNA samples/genotype from cuprizone diet group in this study. The SVZ was then dissociated into single cell suspension following the manufacturer's protocol in the papain dissociation kit (Worthington Biochemicals # LK003150) 41 . Briefly, the dissected SVZ tissue was minced using iris scissors and digested by incubating in a solution containing 20 Units/ml papain and 0.005% DNase for 30 minutes at 37 °C. Subsequently, the papain was inactivated with albumin-ovomucoid inhibitor and the digested tissue was needle triturated to obtain a cell suspension 41 . www.nature.com/scientificdata www.nature.com/scientificdata/ The cell suspension was passed through a 50 µm celltrix filter (Sysmex-Partec # 04-0042-2317) immediately before FACsorting the GFP+ NSCs using a MoFlo XDP cell sorter (Beckman Coulter) in the NYU Cytometry core facility. NSCs harvested from Gli1 NULL mice treated with corn oil served as negative controls for FACsorting. A forward scatter vs. side scatter dot plot was first used to gate the primary cell population and eliminate dead cells and debris. The gating was further refined to exclude doublets to finally select GFP+ cells (Fig. 2, Gating strategy exemplified by magenta arrows). In order to minimize any perturbation of gene expression, we performed cell sorting based on cell size and GFP expression immediately after obtaining a single cell suspension without manipulating the cells. To prevent the perturbation of gene expression, we did not add a viability dye to the cells during the FACsorting. However, we used 10 µl of the sorted GFP+ cells in phosphate buffered saline (PBS) for trypan blue cell viability assay and quantified the clear viable cells along with the blue stained dead cells using a hemocytometer 42 . All our FACsorted samples consisted of more than 95% viable cells. On average, the cell sorting yielded 17,666.7 ± 4163.3 GFP+ cells in the Gli1 HET healthy group on control diet, 7,833.3 ± 2,112.7 GFP+ cells in the Gli1 HET demyelination group on cuprizone diet, 41,333.3 ± 5033.2 GFP+ cells in the Gli1 NULL healthy group on control diet and 34,000 ± 10,816.7 GFP+ cells in the Gli1 NULL demyelination group on cuprizone diet.
Total RNa isolation. The GFP positive neural stem cells were collected in RNA extraction buffer (Arcturus PicoPure RNA isolation kit, Applied Biosystems # 12204-01) after FACsorting and total RNA was isolated following the manufacturer's instructions. The isolated RNA was eluted in 15 µl of elution buffer following on column DNase1 treatment using RNase-free DNase set (Qiagen # 79254). The total RNA was assessed for RNA integrity with a bioanalyzer (Agilent) in the NYU Genome Technology Center (Table 1). RNA samples with a RIN of more than 8 were submitted to NYU Genome Technology Center for Next Generation Sequencing (NGS).  www.nature.com/scientificdata www.nature.com/scientificdata/ Library preparation and RNa sequencing. The mRNA was purified and cDNA was synthesized with the mRNA fragments as templates using SMART-Seq V4 RNA kit (Clontech # 634889). The sequencing libraries were generated using Nextera XT DNA Library Preparation Kit (Illumina #FC-131-1024), which uses an engineered transposome to simultaneously fragment and barcode the cDNA, adding unique short adapter sequences. The barcoded cDNAs were then amplified by 6 cycles of PCR and their quality was checked for size, quantity and integrity by TapeStation system (Agilent) as well as qPCR for precise quantification. The generated libraries were then mixed for multiplexing on 3 lanes and sequenced on Illumina HiSeq 2500 platform with 150 bp paired end modules, which yielded mapped reads in the range of 16-39 million per sample.

Data transformation and downstream analysis. The data were transformed and analyzed by a bioin-
formatician in the NYU Genome Technology Center. Briefly, the raw sequencing data files were initially processed using FASTQC v.0.11.5 in order to perform quality control checks. Subsequently the Fastq files were aligned to the mm10 mouse reference genome with TopHat v2.1.1. About 93% to 96% of the RNAseq reads, including R1 reads in the 5′ to 3′ forward direction and R2 reads in the reverse direction, mapped to the mouse genome using the Ensembl gene annotation system 43 . The aligned bam files were then quantified for gene expression values utilizing HTSeq-count 0.11.1. Finally DESeq 2 v1.12 was used for analysis of the quantified count matrix 44 . Genes that had an adjusted p-value (False Discovery Rate) of less than 0.05 were considered significantly different for respective pairwise comparisons.

Data Records
The sequencing data have been deposited in NCBI's Gene Expression Omnibus and are accessible through GEO Series accession number GSE162683 45 . Table 2 lists the accession number for each sample. The data related to RNA quality, barcode lane statistics, RNA read counts, alignment statistics, DEseq 2 expression counts and the data comparison between the groups have been deposited in Figshare 46 .

Technical Validation
FaCsorting. Since Gli1 NULL mice carry two copies of the CreERT2 allele, they would be expected to have a higher probability of tamoxifen independent leaky GFP expression. Hence, we used cells isolated from the brains of Gli1 NULL mice but injected with corn oil instead of tamoxifen such that GFP is not expressed, as our negative control for FACsorting. These GFP negative cells were used to set up the gating parameters before GFP positive cells were sorted which ensured that only GFP positive neural stem cells were collected after sorting (Fig. 2). RNa quality. The quality of total RNA was assessed by RNA Integrity Number (RIN) using the Agilent Bioanalyzer. Following the standard practice in NGS, we used RNA with RIN >8 for our sequencing analysis 47 (Table 1).
Genotypic and phenotypic assessment. Before harvesting the neural stem cells, we collected the tail tissue from all the mice following euthanasia. These tissues were used for PCR analysis to reconfirm the genotypes of the mice used to isolate neural stem cells. We also obtained a higher number of GFP positive cells from Gli1 Null mice compared to Gli1 Het mice, consistent with higher proliferation rates of NSCs in the Gli1 Null mice 27 , further validating their phenotype. An analysis of differential gene expression in healthy Gli1 Het NSCs vs. healthy Gli1 Null NSCs showed a significant difference in the expression of only one gene i.e. Gli1, which not only confirmed the genotype but also confirmed the well-established lack of phenotype in the healthy Gli1 Null mice 48 (Fig. 3). In contrast, there were substantial differences (i.e. 1507 genes) in the transcriptomes of Gli1 Het vs. Gli1 Null NSCs fed with cuprizone diet for 3 weeks (Fig. 3, Table 3), consistent with their very distinct responses to demyelination. Many of these genes were downregulated in the nulls in agreement with the role of Gli1 as a transcriptional activator 49,50 . In the forebrain, Gli1 is expressed by neural stem cells in the SVZ and by mature astrocytes in the parenchyma outside the SVZ excluding the white matter tracts 26,51 . Consistently, the expression of stem cell marker genes  www.nature.com/scientificdata www.nature.com/scientificdata/ (Sox2, GFAP, Nestin) were higher than genes expressed in differentiated cells (APC, Htra1, S100β, PDGFRα, SYNGAP1) in both Gli1 HET and Gli1 NULL healthy control NSCs.
Clustering of RNaseq data. A 3D Principle Component Analysis (PCA) plot visualized by Plotly (Fig. 4) showed clustering of the samples in the control diet groups from both Gli1 Het and Gli1 Null neural stem cells; similarly the cuprizone diet RNA samples from both genotypes clustered together but away from the control diet groups. We observed an increased variance in the Gli1 Het samples of both control and cuprizone diet groups. However, the heatmaps for the differentially expressed genes (FDR < 0.05) showed clustering of Gli1 Het and Gli1 Null samples in the control groups (Fig. 5). When we compared the differentially expressed genes (FDR < 0.05) between the cuprizone groups, the expression in the Gli1 Null samples appeared similar to the expression in the control groups (Fig. 5), suggesting that demyelination results in lesser perturbation of gene expression in Gli1 Null NSCs compared to Gli1 Het NSCs.