Abstract
Schwann cells play a critical role after peripheral nerve injury by clearing myelin debris, forming axon-guiding bands of Büngner, and remyelinating regenerating axons. Schwann cells undergo epigenomic remodeling to differentiate into a repair state that expresses unique genes, some of which are not expressed at other stages of Schwann cell development. We previously identified a set of enhancers that are activated in Schwann cells after nerve injury, and we determined whether these enhancers are preprogrammed into the Schwann cell epigenome as poised enhancers before injury. Poised enhancers share many attributes of active enhancers, such as open chromatin, but are marked by repressive histone H3 lysine 27 (H3K27) trimethylation rather than H3K27 acetylation. We find that most injury-induced enhancers are not marked as poised enhancers before injury indicating that injury-induced enhancers are not preprogrammed in the Schwann cell epigenome. Injury-induced enhancers are enriched with AP-1 binding motifs, and the c-JUN subunit of AP-1 had been shown to be critical to drive the transcriptional response of Schwann cells after injury. Using in vivo chromatin immunoprecipitation sequencing analysis in rat, we find that c-JUN binds to a subset of injury-induced enhancers. To test the role of specific injury-induced enhancers, we focused on c-JUN-binding enhancers upstream of the Sonic hedgehog (Shh) gene, which is only upregulated in repair Schwann cells compared with other stages of Schwann cell development. We used targeted deletions in male/female mice to show that the enhancers are required for robust induction of the Shh gene after injury.
SIGNIFICANCE STATEMENT The proregenerative actions of Schwann cells after nerve injury depends on profound reprogramming of the epigenome. The repair state is directed by injury-induced transcription factors, like JUN, which is uniquely required after nerve injury. In this study, we test whether the injury program is preprogrammed into the epigenome as poised enhancers and define which enhancers bind JUN. Finally, we test the roles of these enhancers by performing clustered regularly interspaced short palindromic repeat (CRISPR)–mediated deletion of JUN-bound injury enhancers in the Sonic hedgehog gene. Although many long-range enhancers drive expression of Sonic hedgehog at different developmental stages of specific tissues, these studies identify an entirely new set of enhancers that are required for Sonic hedgehog induction in Schwann cells after injury.
Introduction
The capacity of peripheral nerve for regeneration after injury is dependent in many respects on Schwann cells (SCs). SCs undergo major injury-induced reprogramming as they are repurposed from myelin-producing cells or nonmyelinating cells (Remak SCs) to their nerve repair objectives (Jessen and Mirsky, 2016, 2019; Arthur-Farraj and Coleman, 2021). In the repair state, SCs clear myelin debris, secrete factors to summon macrophages, promote neuronal regeneration, create axon guidance tracks (Büngner bands), and remyelinate axons (Gomez-Sanchez et al., 2017; Jessen and Mirsky, 2019). Understanding the genesis of the repair SC is important for developing approaches to facilitate regeneration of nerves, which is typically impaired in aging or in pathologic conditions such as diabetic neuropathy (Painter et al., 2014; Jessen and Mirsky, 2019; Wagstaff et al., 2021; Arthur-Farraj and Coleman, 2021).
The remarkable transition of myelinating to repair SC during nerve injury is accompanied by a unique gene expression program (Nagarajan et al., 2002; Arthur-Farraj et al., 2017; Clements et al., 2017; Toma et al., 2020; Wolbert et al., 2020). SCs in injured nerves do not merely dedifferentiate into a precursor cell, but rather transdifferentiate into a unique repair state that activates a gene program distinct from other stages of SC development (Jessen and Mirsky, 2016). One aspect of this unique state is defined by reliance on injury-specific transcription factors such as c-JUN (hereafter referred to as JUN), which is not required for Schwann cell development but is a major early response component for injury gene activation in SCs. A knock-out of Jun inhibited SC repair gene induction and caused neuronal death and reduced functional recovery after nerve injury (Arthur-Farraj et al., 2012; Fontana et al., 2012). Moreover, overexpression of JUN is sufficient to drive expression of a subset of injury genes, namely, Shh, Glia Cell Line-Derived Neurotrophic Factor (Gdnf), and Oligodendrocyte Transcription Factor 1 (Olig1; Fazal et al., 2017). JUN is therefore a key injury-induced transcription factor, although others, including STAT3 and RUNX2, may play important roles (Hung et al., 2015; Benito et al., 2017).
The novel features of the repair state extend to the use of a unique set of regulatory elements associated with the injury program. Our previous studies sought to elucidate SC reprogramming after nerve injury by assessing active enhancer regions proximal to myelin and injury genes (Hung et al., 2015). The characteristics of active enhancers include open chromatin and histone H3 lysine 27 (H3K27) acetylation (H3K27ac; Heintzman et al., 2009; Creyghton et al., 2010; Rada-Iglesias et al., 2011; Buecker and Wysocka, 2012). We used the histone mark H3K27ac to identify enhancers that become active after injury (InjuryDB) and also active enhancers in mature SCs that lose H3K27ac after injury (ShamDB). Many ShamDB enhancers are proximal to a subset of myelin genes, which decrease in expression after injury. Conversely, many InjuryDB enhancers are proximal to injury genes that are activated after nerve injury. Although a motif analysis revealed enrichment of the JUN binding motif in injury-induced enhancers (Hung et al., 2015), the scope of JUN binding in relation to injury-induced genes has not been defined.
Although H3K27ac marks actively engaged enhancers, stem cell studies identified a poised enhancer state marked by open chromatin, H3K4me1 (mono methylation of lysine 4 of histone H3), and H3K27 trimethylation (H3K27me3) instead of H3K27ac (Creyghton et al., 2010; Rada-Iglesias et al., 2011). Because H3K27me3 is a repressive mark, many of these enhancers could be activated during subsequent differentiation through active recruitment of H3K27 acetylases like CBP/p300. This raised the possibility that the injury program of Schwann cells may be preprogrammed into the epigenome through poised enhancers that could be activated after nerve injury. Our experiments test whether injury gene induction is associated with poised enhancers, and we find that pioneer transcription factor activity is likely a key mechanism underpinning gene expression changes. In addition, we find that JUN binds to many injury-induced enhancers and test the role of injury-induced enhancers in the activation of the Sonic hedgehog (Shh) gene.
Materials and Methods
Rat nerve injury surgery
All animal experiments were performed according to protocols approved by the University of Wisconsin, School of Veterinary Medicine. Two male 4-week-old Sprague Dawley rats (The Jackson Laboratory) were anesthetized under isoflurane and given an injection of 20 mg/kg of ketoprofen. Under aseptic conditions, a 5 mm incision was made through the skin and muscle layers at the proximal lateral region of the femur. The sciatic nerve was exposed and cut to replicate injury. The contralateral leg received a sham operation in which a 5 mm incision was made just through the skin and muscle layers. The skin incision was sutured with rodent surgical staples, and the rats were caged for 8 d after surgery. The nerve tissue distal to the injury or sham site was harvested for chromatin immunoprecipitation (ChIP) experiments.
ChIP
Freshly dissected rat sciatic nerve was used for ChIP sequencing (ChIP-seq) using the MNase protocol described previously (Ma et al., 2018) with 4 μg of JUN antibody (catalog #sc-1694, Santa Cruz Biotechnology; RRID:AB_631263). Two biological replicates were performed, and samples/inputs were sequenced on a Illumina HiSeq 2500 instrument at the University of Wisconsin Biotechnology Center.
Luciferase assay
Three enhancers of mouse Shh, as defined by the following mm10 coordinates: chr5:28557965–28558769; chr5:28567282–28568085; chr5:28630064–28630888, were amplified from mouse genomic DNA. Enhancer sites were cloned upstream of the pGL4 luciferase reporter containing the minimal E1B TATA promoter using Acc65I and BglII. RT4 Schwann cells were transfected with those enhancers and pRL-TK using TransIT-X2 (catalog #MIR6004, Mirus Bio) and harvested for dual-luciferase assay 48 h post-transfection (n = 3 per group). The luciferase assay was performed using the Dual-Luciferase Reporter Assay System (Promega). A human JUN expression vector under the CMV promoter in pEZ-MO2 was purchased from GeneCopoeia (catalog #EX-B0091-M02). For siRNA transfections, RT4 Schwann cells were transfected with enhancer-1 and/or 25 nmol siRNAs targeting Jun [siJun-2, DsiRNA, DesignID #rn.Ri.Jun.13.2, Integrated DNA Technologies (IDT)] and siJun-3 (DsiRNA, DesignID #rn.Ri.Jun.13.3, IDT) or a negative siRNA control (catalog #51-01-14-04, IDT).
At 48 h after transfection, total RNA was extracted and cleaned using Trizol Reagent (catalog #15596018, Invitrogen) and RNA Clean & Concentrator-5 (catalog #R1014, Zymo Research). Jun expression was analyzed by qRT-PCR with the primers forward GAGAGGAAGCGCATGAGGAAC, reverse CCTTTTCCGGCACTTGGAG.
Bioinformatics
JUN ChIP-seq reads [Gene Expression Omnibius (GEO) series GSE190858] were mapped to the reference genome rn5 using Bowtie2 (Langmead et al., 2009; Langmead and Salzberg, 2012) to produce Binary Alignment Map (BAM) files for two biological replicates, which were combined for further analysis. BAM files were filtered for mapped reads using BamTools (Quinlan and Hall, 2010; Barnett et al., 2011) and sorted into called peaks using MACS2 (Zhang et al., 2008; Feng et al., 2012; Liu, 2014). BedTools bamCoverage generated bedgraphs of ChIP-seq samples.
Overlap analyses compared peaks from MACS2 annotated to the nearest gene, intergenic region, promoter, TSS (transcription start site), TES (transcription end sites), exon, or intron via ChIPseeker (Yu et al., 2015). Pie charts showing distribution of peak sets were also generated with ChIPseeker. Heat maps were created via EAseq (Lerdrup et al., 2016). Data processing was performed in a cloud-based manner through GalaxyBiostars (Afgan et al., 2018). ChIP-seq tracks were visualized using the University of California, Santa Cruz Genome Browser (Kent et al., 2002). Previous ChIP-seq datasets for H3K27ac (ShamDB and InjuryDB; Hung et al., 2015), H3K27me3, and H3K4me3 (Ma et al., 2016, 2018) are available at GEO using the following accession numbers: GSE63103, GSE106990, and GSE84272. RNA-seq analysis of JUN overexpression data (Fazal et al., 2017) was filtered to significant genes using the Bioconductor package DESeq (Anders and Huber, 2010). Additional expression datasets (Clements et al., 2017) were used to assess Schwann cell-specific JUN target genes. Analysis of evolutionarily conserved regions in the Shh locus was performed using https://ww.dcode.org (Loots and Ovcharenko, 2007).
Generation of enhancer KO mice
The Shh 5′ enhancer deletion mouse line was generated by the Northwestern University Transgenic and Targeted Mutagenesis Laboratory using clustered regularly interspaced short palindromic repeat (CRISPR) gene editing techniques. Mice were bred and housed in a specific pathogen-free facility on a 12 h light/dark cycle and fed ad libitum in accordance with the Northwestern University Institutional Animal Care and Use Committee regulations.
gRNA identification and synthesis
The gRNA targeting the regions of interest were identified using CRISPOR online software (crispor.tefor.net; Concordet and Haeussler, 2018). These experiments used the Alt-R CRISPR-Cas9 system from IDT, according to instructions from the manufacturer. Briefly, a sequence-specific crRNA is complexed with tracrRNA (IDT, catalog #1072532) to form an individual gRNA. Each gRNA was incubated, separately, with HiFidelity Cas9 protein (catalog #1081064, IDT) to form individual ribonucleotide protein complexes (RNPs). Four CRISPR guide RNAs (gRNA) were designed to knock out the predicted Shh enhancer elements 1, 2 and 3 (see Fig. 4).
Electroporation of fertilized embryos
On day 1, RNPs containing gRNA 1a (5'-ttagtccatcacctagaaag −3') and gRNA 1b (5′- aatgcactcagataacatag-3') were introduced into C57BL/6J fertilized embryos as described, with minor modifications (Chen et al., 2016; Teixeira et al., 2018). Briefly, the RNP complexes were electroporated into fertilized embryos using the ECM 830 Square Wave Electroporation System (BTX). The final concentration of reagents was 2 μm of each gRNA and 4 μm of HiFidelity Cas9 protein. After electroporation, the cells were cultured overnight in Global media (catalog #LGGG-050, Cooper Surgical) supplemented with filtered BSA (4%) at 37°C and 5% CO2. On day 2, RNPs containing gRNA 2a (5′-taagtgtttagcctagactc-3') and gRNA 2b (5'-tctctgtgttggaccaccaa-3') were electroporated into the day1 cells using the same conditions. Electroporated cells were then transferred into pseudopregnant females.
Mouse genotyping
PCR1 amplifies the WT enhancer 2 sequence and includes primers Shh_enh2_F2 5′-CTGAAAGGGCAGCAGTTACC-3′ and Shh_enh2_R2 5′-AGTAGCTGTTCACCCCACTC-3′ (expected band size 455 bp). The second PCR confirms the large deletion on enhancers 1 and 2 and includes Shh_enh2_F2 5′-CTGAAAGGGCAGCAGTTACC-3′ and Shh_enh1_R1 5′-ACCATGGGACCTCAGAAGTG-3′ (expected band size 200 bp). The presence of this band indicated the presence of an allele with the deletion of enhancers 1 and 2. PCR3 amplifies the WT enhancer 3 sequence and includes Shh_enh3_F15' CGACCCTCAGCCAGTGAAG-3′ and Shh_enh3_R1 5′-TCTGCCAGTTCAGTCTCTCTC-3′ (expected band size 1000 bp). Finally, the deletion of enhancer 3 was confirmed with PCR4, which included primers Shh_enh3_F2 5′-TGGACAGCCCAGATAGGACT-3′and Shh_enh3_R1 5′-TCTGCCAGTTCAGTCTCTCTC-3′ (expected band size 458 bp). PCR products were run on a 2% agarose gel supplemented with GelGreen dye (Biotium, 41005-1). A subset of F0 samples was submitted to GENEWIZ for Sanger sequencing and analyzed using SnapGene software.
Mouse nerve injury surgery
Male and female 7–10-week-old triple KO homozygous mice, or C57BL/6J controls, were anesthetized with isoflurane and given an injection of BupSR (sustained release) at 0.6–1.0 mg/kg SC and carprofen (Rimadyl) at 5–10 mg/kg SC every 12 h for analgesia. Under aseptic conditions, a 5 mm incision was made through the skin and muscle layers at the proximal lateral region of the femur. The sciatic nerve was transected, and the contralateral leg received a sham operation where a 5 mm incision was made just through the skin and muscle layers. The skin incision was sutured with rodent surgical staples, and the mice were caged for 24 h after surgery. The nerve tissue distal to the injury or sham site was harvested for RT (reverse transcriptase)-qPCR analysis.
One-day injury RNA preparation, reverse transcription, and quantitative PCR
Trizol reagent was added to harvested sciatic nerve, and RNA was prepared according to the protocol of the manufacturer (catalog #15596018, Invitrogen). RNA samples were reverse transcribed on Applied Biosystems MiniAmp Thermal Cycler using Quantabio qScript cDNA Supermix. Gene expression levels were measured by quantitative PCR on Bio-Rad CFX96 Real-Time PCR Detection System using Bio-Rad iQ SYBR Green Supermix. Levels of all transcripts were normalized to Gapdh levels. Gapdh-normalized transcript levels of injury samples were normalized again to Gapdh-normalized sham transcript levels using the ΔΔCT method (Livak and Schmittgen, 2001).
Four-day injury RNA preparation, reverse transcription, and quantitative PCR
Frozen sciatic nerves were homogenized in Trizol. In-column DNase digestion and RNA purification were performed via the Zymo RNA Clean & Concentrator kit protocol (catalog #R1014, Zymo Research). RNA samples were reverse transcribed on an Applied Biosystems 2720 Thermal Cycler. Five hundred nanograms of RNA were heated to 65°C with IDT Random Hexamer Primer Cocktail (catalog #51-011826, IDT) and Promega dNTP Mix (catalog #U151B) for 5 min. The Invitrogen SuperScript II Reverse Transcriptase Kit (catalog #Y02321) and Promega Rnasin (catalog #N251B) were added to the RNA mix and incubated at 42°C for 1 h, then at 95°C for 5 min. cDNA was eluted in 100 μl of nucleotide-free water. Real-time qPCR was performed with indicated primer sets (Table 1) using the Applied Biosystems Viia7 system and the Applied Biosystems PowerSYBR Green Master Mix (catalog #4367659). All transcripts were normalized to Gapdh and wild-type 4 d injured nerves using the ΔΔCT method (Livak and Schmittgen, 2001).
Quantitative PCR primers
RNA-seq analysis
RNA-seq was performed at 1 d postinjury and sham nerve with n = 5/group (GEO series GSE209658). The quality of reads, in FASTQ format, was evaluated using the FastQC tool. Reads were trimmed to remove Illumina adapters from the 3′ ends using Cutadapt (Martin, 2011). Trimmed reads were aligned to the Mus musculus genome (mm10) using STAR (Dobin et al., 2013). Read counts for each gene were calculated using htseq-count (Anders et al., 2015) in conjunction with a gene annotation file for mm10 obtained from Ensembl (http://useast.ensembl.org/index.html).
Experimental design and statistical analysis
For RNA-seq data, normalization and differential expression were calculated using DESeq2, which uses the Wald test (Love et al., 2014). The cutoff for determining significantly differentially expressed genes was an false discovery rate adjusted p value <0.05 using the Benjamini–Hochberg method. Gene Ontology analysis was performed using Enrichr with all upregulated/downregulated genes with a nonadjusted p value of <0.05 (Xie et al., 2021) and the Molecular Signature Database hallmark gene sets (Liberzon et al., 2015). For quantitative PCR, unpaired t tests were used for comparisons between the means of two groups (WT vs KO).
Results
Most injury-induced enhancers are not poised before injury
Although we had previously identified injury-induced enhancers associated with injury genes in peripheral nerve (Hung et al., 2015), we wished to determine whether this transcriptional program was preprogrammed or marked in the Schwann cell genome before injury. Previous stem cell studies defined active and poised enhancer states that share open chromatin and histone H3K4 monomethylation, but the key distinction is that poised enhancers are marked by repressive H3K27me3 rather than H3K27 acetylation (Creyghton et al., 2010). To assess whether InjuryDB enhancers are poised before injury, we used ChIP-seq data of H3K27me3 in rat sciatic nerve (Ma et al., 2016). Rat sciatic nerve is a good model for SC enhancer characterization as neuronal nuclei are absent and ∼70–80% of peripheral nerve nuclei belong to Schwann cells (Zorick et al., 1996; Topilko et al., 1997). We screened for poised enhancers by testing whether the previously described injury-induced enhancers (Hung et al., 2015) overlap with 27,277 H3K27me3 ChIP-seq peaks (Fig. 1A). Surprisingly, only 45 peaks, or ∼1% of injury-induced enhancers are enriched for H3K27me3 before injury. Therefore, most injury-induced enhancers are not poised before injury.
Injury-induced enhancers are not poised. A, Venn diagram shows overlap between InjuryDB peaks and H3K27me3 sham peaks. Only 45 InjuryDB peaks have H3K27me3. B, Heat maps show distribution of H3K27ac Sham, H3K27ac Injury, and H3K27me3 Sham reads centered on ShamDB and InjuryDB peak locations, sorted in descending order by average read density. There is enrichment of H3K27ac at ShamDB sites in the sham condition, enrichment of H3K27ac at InjuryDB sites in the injury condition, and little enrichment of H3K27me3 at InjuryDB in the sham condition. Bottom right, Rectangular box in H3K27me3 heat map at InjuryDB enhancers denotes poised enhancers. Further analysis of poised enhancers is shown in Extended Data Figure 1-1.
Figure 1-1
Poised enhancers in injury genes. A–C, ChIP-seq data on rat peripheral nerve shows the presence of H3K27me3 on injury-induced enhancers to identify three poised enhancers in SC injury genes (highlighted in red), Sox2, Met, and Syk. Highlighted enhancers are conserved in mouse and human. Download Figure 1-1, EPS file.
In Figure 1B, the distribution of the active H3K27ac enhancer mark is shown in read density plots and heat maps in uninjured nerve (sham) and injured nerve and compared with H3K27me3 as a marker of the poised state in uninjured nerve. The plots were centered on injury-induced enhancer peaks (Hung et al., 2015) and ShamDB peaks as a control (Fig. 1B). As expected, H3K27ac is absent on injury-induced enhancers (InjuryDB) in the sham condition and dramatically increased after injury. Consistent with the peak overlap analysis, there is no preferential deposition of H3K27me3 at injury-induced enhancers in the sham condition, indicating that most injury-induced enhancers are not poised before injury.
Poised enhancers are a small subset of InjuryDB enhancers overall, and we identified only three poised enhancer peaks (of 45) that are proximal to significantly induced SC injury genes (Extended Data Fig. 1-1), SRY-Box Transcription factor 2 (Sox2), Mesenchymal Epithelial Transition Factor (c-Met), and Spleen Tyrosine Kinase (Syk). In SCs, SOX2 is normally downregulated in myelinating cells, and transgenic overexpression inhibits myelination (Le et al., 2005; Roberts et al., 2017). SOX2 is also involved in the production of fibronectin and organizing SC migration toward the distal stump of the injured nerve (Parrinello et al., 2010; Torres-Mejía et al., 2020). MET is a tyrosine kinase receptor for Hepatocyte Growth Factor in Schwann cells and promotes peripheral nerve regeneration (Ko et al., 2018). The role of SYK is unknown in SCs, but this protein is also a tyrosine kinase that assists with diverse cellular functions such as adhesion and immune signaling (Mócsai et al., 2010).
JUN binds to injury-induced enhancers
We decided to further test the hypothesis that JUN is involved in the regulation of many of our injury-induced enhancers because JUN has been shown to have a critical role in Schwann cell responses to nerve injury (Arthur-Farraj et al., 2012, 2017), and in addition, AP-1 complexes containing JUN and related family members have been shown to have the ability to open up new enhancers in other cell types (Biddie et al., 2011; Vierbuchen et al., 2017; Yukawa et al., 2020). In our previous study, ∼29% of injury-induced enhancers had a JUN/AP-1 binding motif (Hung et al., 2015). To assess whether InjuryDB enhancers are bound by JUN, a ChIP-seq analysis for JUN was performed at 8 d postinjury (dpi). This time point was chosen because some injury genes are activated over several days after injury (Ma et al., 2018), and the sustained JUN appears to be an important determinant of its activity with levels peaking at 1 week after injury (Wagstaff et al., 2021). The largest proportion of JUN ChIP-seq peaks mapped to intergenic regions (Fig. 2A). We analyzed the JUN 8 dpi peaks to determine whether JUN preferentially binds to the ShamDB or InjuryDB subset (Fig. 2B). As shown, JUN preferentially binds to InjuryDB enhancers with almost a third of those enhancers containing a JUN peak. To visualize the preferential binding of JUN on a global scale, a read density plot was generated centered on the previously defined ShamDB and InjuryDB enhancers (Fig. 2C). In line with the overlap analysis, the average read density of JUN is increased in the InjuryDB subset with minimal enrichment at ShamDB enhancers. Conversely, a similar heat map centered on the called JUN ChIP-seq peaks depicts enrichment of H3K27ac in sham and injury conditions (Fig. 2D). These plots show an increase in H3K27ac at most JUN binding sites after injury, suggesting an association between JUN binding and enhancer activation. However, there is significant binding of JUN to enhancers that were also active in the Sham condition, indicating that JUN does bind to some pre-established enhancers.
JUN preferentially binds at InjuryDB enhancers after injury. A, Pie chart of annotated JUN binding peaks identified by ChIP-seq of rat peripheral nerve at 8 d post-injury; 38.52% of JUN peaks bind at distal intergenic regions. B, Venn diagram of the overlap among JUN peaks, ShamDB, and InjuryDB. Only 80 ShamDB peaks overlap with JUN peaks, whereas InjuryDB peaks overlap with 1301 JUN peaks (p < 0.001). C, Heat maps show distribution of JUN ChIP-seq reads centered on the previously defined ShamDB and InjuryDB enhancer peaks. JUN 8 dpi peaks were assigned scores based on peak quality, log 2 normalized, and sorted by descending order. D, Heat maps show distribution of H3K27ac Sham and H3K27ac Injury ChIP-seq reads on called JUN intergenic peaks sorted by peak quality, log 2 normalized, and arranged in descending order. Further analysis of JUN binding relative to promoters of injury genes is shown in Extended Data Figures 2-1 and 2-2.
Figure 2-1
JUN binds to promoters of some injury-induced genes. A, Pie chart of JUN injury ChIP-seq peaks (8 dpi); 22.18% of JUN peaks are found at annotated promoters. B, Heat map shows distribution of H3K4me3 reads after injury centered on JUN promoter peaks. C, Venn diagram of gene promoters that have called peaks for H3K4me3 and JUN 8 dpi versus significant JUN-dependent genes from overexpression and injury-induced SC gene datasets (Clements et al., 2017; Wagstaff et al., 2021). Only ∼8% of JUN overexpression genes overlap with gene promoters containing H3K4me3 and JUN peaks. Download Figure 2-1, EPS file.
A significant proportion of JUN peaks are localized within 2 kb of transcription start sites (Extended Data Figs. 2-1, 2-2). It is important to note that H3K27ac is associated with actively engaged enhancers but tends to be constitutively associated with promoters (Heintzman et al., 2009), and very few promoters were identified as InjuryDB sites (Hung et al., 2015). Therefore, our H3K27ac analysis would likely have missed important JUN binding sites in promoter regions. As not all annotated promoters are active in Schwann cells, we verified promoter localization of JUN peaks by generating a read density heat map for H3K4me3 in peripheral nerve (Ma et al., 2016), a histone mark denoting active promoters (Extended Data Fig. 2-1B). A majority of JUN peak regions near transcription start sites show very strong H3K4me3 read densities centralized within a 2 kb window, suggesting that most of the annotated JUN promoter peaks are indeed found at active promoter regions.
One study used a transgenic overexpression approach to define genes that respond to JUN activation (Fazal et al., 2017). Additionally, RNA-seq data from sorted SCs after injury (Clements et al., 2017) show 1189 genes that are significantly induced. Using these datasets, we asked how many of our 2024 promoter-localized JUN peaks overlap with genes that are significantly upregulated in JUN overexpression mice and injury-induced SC genes (Extended Data Fig. 2-1C). Of the 65 genes that are responsive to JUN and induced significantly after injury in SCs, only 18 have active promoter JUN peaks. Additionally, only 187 SC injury-induced genes have JUN binding at their promoters. Although JUN binding sites have been identified in promoter regions (Fontana et al., 2012; Norrmén et al., 2018), most JUN-regulated genes described previously (Arthur-Farraj et al., 2012, 2017) do not have JUN binding at their promoters.
JUN activates injury-induced Sonic hedgehog enhancers
Shh is one of the most highly induced SC nerve injury genes and is a JUN-dependent gene (Arthur-Farraj et al., 2012). Shh has three distal InjuryDB enhancers that are far away from the next nearest gene, making this gene an ideal model to assess enhancer characteristics and functionality. The two proximal Shh enhancers are located at 65 and 76 kb upstream of the gene, and the farthest being ∼180 kb upstream (Fig. 3A). None of these enhancers are poised, with the first and third enhancers also containing a JUN peak. All three enhancers have conserved AP-1 binding motifs.
Injury-induced enhancers upstream of Shh are regulated by JUN. A, ChIP-seq analysis of JUN, H3K27ac, and H3K27me3 at the Shh locus (n = 2). The three InjuryDB enhancers proximal to Shh are highlighted in red. Conserved AP-1 motifs in each enhancer are shown for rat, mouse, and human. B, Luciferase reporter assay for Shh enhancers (*p < 0.05, **p < 0.01, ***p < 0.001). C, Bar graph shows Shh enhancer activation after cotransfection of a JUN expression plasmid (n = 3). Additional analysis of JUN dependence of the Shh enhancers is shown in Extended Data Figure 3-1.
Figure 3-1
Shh enhancer 1 is regulated by JUN. A, The luciferase activity of Shh enhancer 1 was assessed after cotransfecting siRNAs for Jun or a negative control siRNA into the RT4 Schwann cell line. B, Independent transfections of the siRNAs were used to evaluate their ability to downregulate endogenous Jun expression by quantitative RT-PCR. Download Figure 3-1, EPS file.
To assess whether these enhancers can be activated by JUN in vitro, we created a Luciferase reporter assay in which each Shh enhancer was fused to a minimal E1b promoter and the luciferase gene. Cotransfection of each reporter with a JUN expression plasmid was performed in the rat RT4 Schwann cell line. Shh enhancer 2 was induced by JUN expression, despite not having a called JUN peak in vivo (Fig. 3B). Although enhancer 3 was also activated by JUN, the enhancer 1 construct had high basal activity that did not respond to JUN overexpression, so we were unable to test JUN responsiveness. However, given the high basal activity of enhancer 1, we did transfection assays of the enhancer 1 construct in the presence of Jun siRNA, which downregulated the activity of this enhancer (Extended Data Fig. 3-1), consistent with the presence of JUN binding. In summary, JUN appears to regulate three Shh enhancers, consistent with the presence of conserved AP-1 motifs (Fig. 3A).
Deletion of injury-induced enhancers reduces Shh expression after nerve injury
The H3K27ac-marked Shh enhancers are distal to the Shh transcription start site, leaving open the possibility that these are enhancers for neighboring genes, or that other as yet unidentified enhancers were more important for Shh induction. We first examined the overall conservation of these enhancers and observed that at least parts of these enhancers are conserved in mammalian species, and to a lesser extent, in birds (Fig. 4A). Next, to test the function of these enhancers in vivo, we created mice harboring a deletion of all three enhancers using CRISPR technology (Fig. 4B). To avoid a massive deletion of sequences between the first and last guide RNA (∼100 kb between guide 1a, 2b), we used a sequential electroporation protocol to specifically delete the Enhancer 1/2 region (∼11 kb), followed by Enhancer 3 region (∼1.2 kb; see above, Materials and Methods). After validating the deletions and breeding to homozygosity, we then tested whether Shh was induced on nerve transection in these mice harboring these targeted enhancer deletions. In control mice, injured nerve (1 dpi) showed an ∼100-fold induction of Shh compared with uninjured nerve. Gdnf was also induced substantially in line with previous studies (Arthur-Farraj et al., 2012, 2017; Fontana et al., 2012), whereas Mbp was predictably reduced. In the triple enhancer knock-out, Shh induction was markedly reduced (Fig. 4B). Gdnf induction and Mbp reduction was similar to that in controls. Our data demonstrate that these enhancers are essential for Shh induction. As an initial test of a later time point, we analyzed the triple enhancer knock-out at 4 d post-transection. By comparing the relative levels at 4 dpi, the induction of Shh was significantly reduced, and somewhat surprisingly, JUN target gene Bdnf was significantly elevated, with similar trends shown for Gdnf, Runx2, Fgf5, and Olig1.
Deletion of Shh enhancer elements abrogates Shh induction after nerve injury. A, Alignment of the enhancer regions using the UCSC genome browser revealed partial conservation of these regions across different mammalian species. B, Diagram of Shh locus showing deletion of three upstream Shh enhancer elements, which was achieved with sequential CRISPR gene editing. gRNAs targeting enhancers 1 and 2 (triangles 1a and 1b) were electroporated into fertilized C57BL/6J mouse embryos and cultured overnight. Embryos were electroporated with gRNA 2a and 2b (triangles) the next day and transferred into pseudo pregnant females. PCR analysis confirmed the deletion of the three enhancer elements located on Chr 5. C, qRT-PCR on sciatic nerves 1 d post-transection. Dotted line indicates uninjured nerve mRNA levels. Error bars indicate SEM; ***p < 0.001, n = 5 mice each. D, qRT-PCR of Shh triple enhancer knockout compared with wild-type at 4 dpi (n = 6/group). Error bars indicate SEM; *p < 0.05, **p < 0.01.
Transcriptomic analysis of Shh-deficient nerve after injury
To test the role of the induction of Sonic hedgehog after injury, we performed RNA-seq analysis of injured sciatic nerve at 1 dpi and compared control and the triple enhancer knock-out (Fig. 5). Comparison of control and the triple enhancer knock-out at 1 dpi confirmed the virtually complete lack of induction of Shh at 1 dpi, and there were no significant changes in the induction of Jun and other AP-1 family members. However, the analysis revealed a number of significant gene expression changes as shown in the volcano plot (Fig. 5A). Genes that are downregulated with deletion of Shh enhancers included Gli1, and the Hedgehog signaling pathway (Ptch1, Cdk5r1) was a significantly enriched category among downregulated genes. The most highly enriched category was the TNF-α/NFκB pathway including Lif, Cxcl1, and several NFκB-related genes, which are normally induced at 1 d after injury. Two injury-induced Selectin genes (Sele and Selp) that are involved in immune-regulated pathways are also downregulated in the triple enhancer knock-out. Selectins and CXCL1 are involved in neutrophil regulation, and studies have shown that neutrophil infiltration precedes macrophage entry, and that neutrophils are involved in myelin debris clearance (Nadeau et al., 2011; Lindborg et al., 2017), suggesting that neutrophil activity could be affected. Interestingly, upregulated genes were enriched in cell cycle genes that are related to the G2/M checkpoint (Plk1, Aurkb, Cenpf, Mki67, Cdc25c), which is regulated by the FOXM1 transcription factor and POL and Aurora kinases. Based on these data, it is suggested that the early actions of Sonic hedgehog after injury modulate inflammatory and cell cycle pathways, although careful characterization of the knock-out will be needed to determine whether Schwann cell proliferation and macrophage/neutrophil infiltration are affected.
Transcriptomic analysis of Sonic hedgehog–deficient mice after nerve injury. A, RNA-seq was performed at 1 d postinjury in control and triple enhancer knockout (n = 5/group) sciatic nerves, and the volcano plot shows upregulated and downregulated genes relative to control. B, Gene Ontology analysis was performed separately on increased and decreased genes, and enriched categories are shown.
Discussion
During embryonic development, cell specification and differentiation are fundamental events scripted by the action of cell type-specific transcription factors that determine chromatin states and thus cell-type-specific gene expression. In contrast, how a fully differentiated cell reconfigures its chromatin landscape to execute a de novo injury-specific response is not well understood. SCs are a fitting model to study this question because of well-defined developmental and injury states. In SC, the underlying epigenomic mechanism behind induction of injury genes is not well known. The AP-1 transcription factor JUN is required for many aspects of SC responses to injury, and both loss-of-function and gain-of-function experiments have identified a series of JUN-responsive injury genes (Arthur-Farraj et al., 2012; Fazal et al., 2017). The goal of this analysis was to elucidate the connection of JUN with injury-induced enhancers, which are marked by differential H3K27ac deposition in SCs after nerve injury (Hung et al., 2015). Although previous studies have elucidated the role of several pathways regulating JUN induction (Kim et al., 2013, 2018; Norrmén et al., 2018), the status of injury-induced enhancers and the extent of JUN binding to such enhancers remains largely unexplored.
We first determined how many InjuryDB enhancers were poised. Poised enhancers can interact with their target promoters in a PRC2-dependent manner in differentiating ES cells (Cruz-Molina et al., 2017), but it is unclear whether SC injury induction works in this manner. We find that most injury-induced enhancers do not have H3K27me3, meaning that a majority of these enhancers are not poised. We did identify three injury genes that are proximal to poised enhancers, including Sox2, which is a transcription factor important for downregulation of myelin genes and SC migration during nerve injury (Le et al., 2005; Parrinello et al., 2010; Roberts et al., 2017; Torres-Mejía et al., 2020). Although reduction of repressive H3K27me3 through deletion of the EED (embryonic ectoderm development) subunit of PRC2 activated injury genes (e.g., Shh), many injury genes are associated with H3K27me3 in promoters/gene bodies rather than the injury-induced enhancers (Ma et al., 2015, 2016, 2018). Although enhancers themselves are not poised, they could be involved in reversal of polycomb repression in the Shh promoter and gene body, as distal enhancers can reverse polycomb repression at promoters (Vernimmen et al., 2011).
If injury-induced enhancers are not poised or otherwise marked before nerve injury, this raises the question of how injury-induced enhancers are formed. In other systems, pioneer transcription factors have been defined as being able to bind directly to closed chromatin regions and recruit additional transcription factors to the site (Zaret and Carroll, 2011). The JUN-containing AP-1 transcription factor has many attributes of a pioneer factor. Specifically, AP-1 facilitates glucocorticoid-receptor-(GR)-regulated transcription by binding condensed chromatin regions to allow GR binding (Biddie et al., 2011). In T-cells, AP-1 is required to open chromatin regions during T-cell activation (Yukawa et al., 2020). Interestingly, AP-1 also recruits the BAF complex, which is involved in nucleosome sliding and eviction (Vierbuchen et al., 2017). AP-1 was required to selectively open chromatin, recruit BAF, and allow for binding of cell-type-specific transcription factors in mouse embryonic fibroblast enhancers. If AP-1 is a pioneer factor that enables other transcription factors to bind these newly opened enhancers, this model would fit with early induction of JUN after injury (Shy et al., 1996; Parkinson et al., 2008) and the induction of other AP-1 subunits (Arthur-Farraj et al., 2017).
Although there is ample evidence for the importance of JUN regulation, its binding to injury-induced enhancers had not been obtained. Here, ChIP-seq analysis allows us to assess how JUN interacts with injury-regulated enhancers. In our previous characterization of injury-induced enhancers, which are marked by differential H3K27ac deposition in SCs after nerve injury, we also found enrichment of the AP-1 motif in a significant proportion of these enhancers (Hung et al., 2015). Our analysis confirms JUN binding to a similar proportion of injury-induced enhancers. Although JUN also binds to pre-existing enhancers to some extent as well as several promoters of injury-induced genes, its activity appears to be focused at enhancers. It is generally the case that transcription factors like JUN bind more avidly to non-nucleosomal DNA and therefore tend to bind to pre-existing areas of open chromatin as is found in constitutively open enhancers and promoters. In general, the most dramatic examples of gene induction are those involving genes that are silenced and/or minimally expressed before injury. The specificity of the regulation of injury genes is in part because of the activation by JUN/AP-1 of new enhancers but also because many of the injury genes are silenced or expressed at low levels before injury, which can be the result of not only inactive enhancers but also by polycomb repression of injury genes (Ma et al., 2015, 2016, 2018).
JUN is necessary for SC repair function (Arthur-Farraj et al., 2012) because it regulates an important subset of injury genes required for a proper SC repair phenotype (Fazal et al., 2017). The importance of JUN was recently reinforced by the demonstration that the impaired regeneration caused by chronic denervation or aging could be rescued through restoring the levels of JUN (Wagstaff et al., 2021). Among the JUN target genes that have been defined, one such gene is Shh, which becomes induced within 1 d after injury, and it is sustained at higher levels for >2 weeks (Hashimoto et al., 2008; Ma et al., 2018; Wagstaff et al., 2021). Shh is a unique marker of the repair state in the Schwann cell lineage as lineage tracing found no expression of Shh from neural crest to mature Schwann cells (Lin et al., 2015). The significance of Shh induction after injury has been tested using different interventions (siRNA, antibodies, cyclopamine), and effects on axon regeneration, myelin clearance, and macrophage recruitment have been shown (Pepinsky et al., 2002; Hashimoto et al., 2008; Martinez et al., 2015; Yamada et al., 2018, 2020). >After nerve injury, repair SCs undergo dynamic changes in branching and size, elongating substantially to form Büngner bands that facilitate axon regeneration (Gomez-Sanchez et al., 2017). In JUN-deficient cells, these tracks do not form appropriately; rather, irregular cellular profiles are observed (Arthur-Farraj et al., 2012). Because Shh is downstream of JUN, one possibility is that lack of SHH (Sonic hedgehog) underpins this phenotype. Consistent with this, exogenous SHH drives a change in shape and length of cultured SCs (Wagstaff et al., 2021). Finally, a positive feedback loop was identified where activation of Shh transcription by JUN leads to SHH-dependent maintenance of JUN protein levels and phosphorylation for several days after injury (Wagstaff et al., 2021).
Although further analyses are required to define the interplay between these two proteins, the cooperative response of JUN and SHH hints at how JUN target genes may autoregulate JUN at a post-transcriptional level. However, even though this effect occurs at 7 dpi, our experiments suggest that JUN target genes and some other injury responses are actually elevated with the reduction of Shh at 4 dpi. Wagstaff et al. (2021) did show a slight induction (less than twofold, not significant) of JUN at 3 dpi in the Shh knockout followed by a decrease at 7 dpi. Therefore, one possible explanation is that SHH activity initially impedes acquisition of the mature repair SC state. This could be to allow for the proper orchestration and/or timing of the multiple events occurring across various cell types in the injured nerve. Understanding the temporal regulation of JUN and AP-1 target genes could be important for nerve regeneration as JUN is critical for proper regeneration (Arthur-Farraj et al., 2012; Fazal et al., 2017; Kim et al., 2018). An alternative interpretation of our results is that Bdnf/Gdnf are increased as a compensatory mechanism for loss of SHH, although it should be noted that our results contrast with those of Hashimoto et al. (2008), who showed that SHH signaling drives Bdnf expression in injured nerve.
SHH is a potent developmental morphogen, important in many developmental contexts including limb, gut, and CNS development (Fuccillo et al., 2006; Tsukiji et al., 2014). Accordingly, there have been several efforts to identify the Shh enhancers that specify its induction in various tissues. In a particularly dramatic example, a highly conserved long-range limb-bud specific enhancer has been identified (Lettice et al., 2003), and mutation of a single transcription factor binding site in this enhancer in snakes is thought to account for their lack of limbs (Kvon et al., 2016). In the CNS, Shh expression along the rostrocaudal axis of the floor plate is driven by distinct enhancers (Anderson and Hill, 2014). Mutations in the forebrain enhancer lead to reduced binding of SIX3 and haploinsufficiency of SHH, which causes holoprosencephaly (Geng et al., 2008; Jeong et al., 2008). Developmental Shh enhancers are exquisitely modular and spread over a megabase. These long-range enhancers are highly conserved and direct the spatiotemporal regulation of Shh (Anderson and Hill, 2014; Sagai et al., 2019; Amano, 2020). However, none of these enhancers are activated in the SC injury model.
Our work has characterized a novel enhancer region for Shh, the first such enhancers that are activated only during injury, but not in developmental states, as Shh is not expressed developmentally in SCs (Arthur-Farraj et al., 2012; Lin et al., 2015). This enhancer does not overlap other known Shh enhancers, and, accordingly, mice with homozygous deletions are viable and fertile with no obvious abnormalities. However, it should be noted that SLGE (Shh lung-gut enhancer) of Shh is mapped to a position between enhancers 1 and 2 and was found to drive transgenic expression to the lung/intestine (Tsukiji et al., 2014). The SLGE has also been studied as SBE6.1, which has been shown to drive CNS expression in zebrafish, mouse, and rabbit (Benabdallah et al., 2016), and its deletion in ES cells reduces expression of Shh in differentiated neural progenitors. Although the SLGE was not marked by H3K27 acetylation in peripheral nerve, it is removed by the enhancer1/2 deletion. No developmental defects have been reported for deletion of SLGE/SBE6.1 in vivo. Interestingly, enhancer 2 was designated as SBE6.2 based on sequence conservation, but it neither drove CNS expression nor affected Shh levels when deleted in ES-cell-derived neural progenitors (Benabdallah et al., 2016). As enhancers are normally activated by more than one transcription factor, we have identified several other conserved motifs in the three enhancers, including binding sites for SOX and ETS factors. However, none of the predicted binding sites are present in all three enhancers, suggesting that they may be differentially regulated by other injury-induced pathways, and we plan to test the involvement of other such factors.
We show that Shh activation during injury is not because these enhancers are poised but more likely because of activation by pioneer transcription factors, with AP-1 as a prominent candidate. Although these enhancers are not poised, we do note that there is a basal level of H3K27ac on enhancer 2 before injury, suggesting that there is some pre-existing transcription factor binding at this enhancer, and future deletions of individual enhancers will clarify whether enhancer 2 is absolutely required. Given candidate roles of SHH in nerve injury (Pepinsky et al., 2002; Hashimoto et al., 2008; Martinez et al., 2015; Yamada et al., 2018, 2020), it is possible that the inability to properly activate these enhancers during aging or pathologic conditions would impair nerve regeneration (Wagstaff et al., 2021). Additionally, it is conceivable that nucleotide variants in these enhancers may significantly impair nerve regeneration potential.
Footnotes
This work was supported by the National Institutes of Health–National Institute of Neurological Disorders and Stroke Grant R01 NS100510 to J.S. and, in part, by a core grant (P50 HD105353) to the Waisman Center from the Eunice Kennedy Shriver National Institute of Child Health and Human Development. We thank the University of Wisconsin Biotechnology Center DNA Sequencing Facility for sequencing services and Dr. Lynn Doglio at the Northwestern Transgenic and Targeted Mutagenesis Laboratory for generation of mice.
The authors declare no competing financial interests.
- Correspondence should be addressed to John Svaren at john.svaren{at}wisc.edu
