Research Article

Acoel genome reveals the regulatory landscape of whole-body regeneration

See allHide authors and affiliations

Science  15 Mar 2019:
Vol. 363, Issue 6432, eaau6173
DOI: 10.1126/science.aau6173

Acoel-regeneration regulatory landscapes

Some animals, including some types of worms, can undergo whole-body regeneration and replace virtually any missing cell type. Gehrke et al. sequenced and assembled the genome of Hofstenia miamia, a regenerative acoel worm species (see the Perspective by Alonge and Schatz). They identified a variable motif corresponding to regulation of the early growth response (egr) gene that was involved in regeneration. RNA interference experiments and validation in a second species showed that the protein Egr is a pioneer factor that stimulates regeneration.

Science, this issue p. eaau6173; see also p. 1152

Structured Abstract

INTRODUCTION

Although all animals can heal wounds, some are capable of reconstructing their entire bodies from small fragments of the original organism. Whole-body regeneration requires the interplay of wound signaling, stem cell dynamics, and positional identity, all of which have been investigated at the protein-coding level of the genome. Little is known about how the noncoding portion of the genome responds to wounding to control gene expression and to launch the process of whole-body regeneration. Understanding how these control points (regulatory regions) are activated and then operate during regeneration would uncover how genes connect into networks, ultimately restructuring entire body axes. Networks of transcriptional regulatory genes can reveal important mechanisms for how animals can grow new skin, muscles, or even entire brains.

RATIONALE

To identify regulatory regions involved in whole-body regeneration, we sequenced the genome of the highly regenerative acoel Hofstenia miamia, commonly known as the three-banded panther worm. Equipped with this genome, we reasoned that applying the assay for transposase-accessible chromatin using sequencing (ATAC-seq) would identify regulatory regions that change in response to amputation and during whole-body regeneration. Further, by analyzing the sequence motifs contained within these regulatory regions, we sought to predict which transcription factors (TFs) control regeneration gene networks.

RESULTS

The Hofstenia genome assembly totals 950 megabases of sequence, with sufficient contiguity for functional genomics. ATAC-seq data revealed thousands of chromatin regions that respond dynamically during regeneration. A genome-wide scan for TF binding motifs in these regions identified the EGR (early growth response) motif as the most dynamic. By combining RNA interference (RNAi) and RNA-seq, we predicted a set of Egr target genes in Hofstenia. We found that most of these target genes contained EGR binding motifs in neighboring regions of regeneration-responsive chromatin, which failed to respond under egr-RNAi. This functional validation allowed us to build a gene regulatory network (GRN) with Egr as a direct master regulator of downstream regeneration genes. Lastly, by quantifying the binding probabilities of TFs at individual motifs, we identified targets of TFs further downstream of Egr, extending the regeneration GRN.

CONCLUSION

Using our regulatory data, we inferred a GRN for launching whole-body regeneration in the acoel H. miamia, where the master regulator Egr acts as a putative pioneer factor to directly activate wound-induced genes. This network includes homologs of genes that are involved in regeneration in other species, suggesting that it can serve as a template for direct comparisons of regeneration pathways across distantly related animals. Our approach of combining genome-wide assays for chromatin accessibility with functional studies can be applied to extend the network further in time in Hofstenia regeneration and to construct GRNs for regeneration in other systems.

The regulatory landscape of whole-body regeneration.

Hofstenia represents the sister-lineage of other bilaterians and regenerates extensively. We sequenced the genome of H. miamia and used ATAC-seq to identify thousands of regeneration-responsive regions of chromatin. Combining motif analysis, ATAC-seq, and RNAi, we identified Egr as a master regulator of regeneration in Hofstenia and inferred an Egr-controlled GRN for regeneration.

Abstract

Whole-body regeneration is accompanied by complex transcriptomic changes, yet the chromatin regulatory landscapes that mediate this dynamic response remain unexplored. To decipher the regulatory logic that orchestrates regeneration, we sequenced the genome of the acoel worm Hofstenia miamia, a highly regenerative member of the sister lineage of other bilaterians. Epigenomic profiling revealed thousands of regeneration-responsive chromatin regions and identified dynamically bound transcription factor motifs, with the early growth response (EGR) binding site as the most variably accessible during Hofstenia regeneration. Combining egr inhibition with chromatin profiling suggests that Egr functions as a pioneer factor to directly regulate early wound-induced genes. The genetic connections inferred by this approach allowed the construction of a gene regulatory network for whole-body regeneration, enabling genomics-based comparisons of regeneration across species.

The capacity to replace all missing tissues, i.e., whole-body regeneration, is present in nearly all animal phyla (1). Whereas candidate gene studies and high-throughput transcriptomics have yielded insight into genes involved in several aspects of regeneration, including wound response (2, 3), stem cell dynamics (4), and repatterning of tissue identities (57), how these genes are connected is largely unknown. A detailed understanding of the DNA binding site logic of regulatory transcription factors is needed to elucidate connections between genes, in particular to distinguish between direct (i.e., control via binding to cis-regulatory regions of a target gene) and indirect (i.e., control of a downstream gene via an intermediate gene) regulatory relationships. Active “tissue regeneration enhancer elements” have been documented in zebrafish heart and fin regeneration (8, 9), as well as in Drosophila imaginal discs (10), but we lack understanding of how the epigenome responds during whole-body regeneration.

We focused on the model system Hofstenia miamia, an acoel worm that possesses the ability to regenerate its entire body (11) (Fig. 1A). Besides its regenerative capacity, we chose Hofstenia because of its plentiful and accessible embryos (Fig. 1A) and its phylogenetic position as the likely sister group (Xenacoelomorpha) to all other bilaterians (Nephrozoa) (1113) (Fig. 1B).

Fig. 1 The genome of the acoel Hofstenia miamia and global chromatin dynamics of whole-body regeneration.

(A) Hofstenia miamia adult, zygote, and eight-cell embryo. Scale bars represent 500 μm for adult, 100 μm for embryos. (B) Schematic phylogenetic tree showing the placement of acoels (Phylum: Xenacoelomorpha) as the likely sister group to all other bilaterians, according to (12). (C) Principal components analysis of metazoan genomes showing that Hofstenia contains a standard complement of animal genes. Each dot represents a species with a sequenced genome and is colored green (protostome), blue (deuterostome), or brown (nonbilaterian species). For a full list of species in this graph, see fig. S2A and table S1, C and D. (D) Schematic of chromatin profiling workflow for heatmap in (E). Gray box indicates a region of chromatin (“peak”) that is regeneration-responsive (14). (E) Heatmap of all chromatin peaks that show significant changes during tail fragment (“new head”) regeneration. Groups of chromatin regions that show similar temporal patterns of opening or closing are highlighted by black boxes. See fig. S2, B to F, for head fragment (“new tail”) data. (F) Bar plots showing the proportions of total genome sequence, of static peaks, and of peaks dynamic during regeneration contained within promoters, exons, introns, and intergenic regions. Most regeneration-responsive peaks lie in introns and intergenic space. (G) Number of “new” (i.e., not present in any previous time point) chromatin changes at each time point. Animal silhouettes were obtained from PhyloPic (http://phylopic.org/).

Assembly and annotation of the Hofstenia miamia genome

We sequenced the genome of Hofstenia at an average coverage of 89.6× via Illumina sequencing of paired-end [300- and 500–base pair (bp) inserts] and mate pair (3- and 7-kb inserts) libraries derived from a single individual and a pool of five animals, respectively (table S1A) (14). The sequenced animals were descendants of a wild population of individuals, and we estimated the polymorphism level as 0.426% (or 1 single nucleotide polymorphism every 235 bases) (fig. S1A). The draft assembly, achieved by the sequential application of SOAPdenovo (15), PRICE (16), SSPACE (17), and the Chicago method (18), totals 950 Mb of sequence (11.8% gaps) (14). Half of the sequence (N50) is contained in 294 scaffolds longer than 1 Mb in this assembly (fig. S1A), and this contiguity was validated by mapping long-read Oxford Nanopore sequencing and by flagging potential misjoins using REAPR (14). To predict protein-coding genes, we generated a new transcriptome assembly to train Augustus (19) and recovered 22,632 gene models, 97% of which were supported by transcriptome data (14). The Hofstenia genome is 53% repetitive, of which the majority are long terminal repeats derived from retrotransposons (fig. S1B and table S1B). The Hofstenia genome annotation contained 90% of the expected metazoan BUSCO (20) genes (84% complete BUSCOs), yielding a representative protein-coding gene set (fig. S1A) (14). A principal components analysis on gene content in 36 metazoan genomes nested the Hofstenia genome among nonbilaterian and protostome genomes, indicating that it contains a standard complement of animal genes (Fig. 1C, fig. S2A, and table S1, C and D).

Chromatin is highly dynamic during regeneration

To understand chromatin dynamics during the process of whole-body regeneration, we applied the assay for transposase-accessible chromatin followed by sequencing (ATAC-seq) (21) to identify changes in the epigenome following injury. To capture how Hofstenia chromatin responds during regeneration, animals were amputated transversely, and wound sites from both resulting fragments (anterior-facing wound sites from tail fragments and posterior-facing wound sites from head fragments) were harvested for ATAC-seq at “zero” (control), 3, 6, 12, 24, and 48 hours post-amputation (hpa) (Fig. 1D, fig. S2, B to E, and table S2). This time course encompasses many aspects of regeneration, including the wound response (i.e., transcriptional changes initiated immediately upon amputation) at early time points and the formation and patterning of new tissue at later time points (11, 22).

Regeneration induced dynamic patterns of change in chromatin over time, including regions of chromatin (or “peaks”) that “opened” or “closed.” We identified nearly 18,000 changes to chromatin regions (adjusted p < 0.05, Wald test) (14) during initiation of head regeneration in tail fragments (see fig. S2, E and F, for head fragment data analysis) (Fig. 1E). This regeneration-responsive chromatin was most frequently located in introns and intergenic space (Fig. 1F and table S3). Notably, the majority of peaks appeared within the first 6 hpa (Fig. 1G and fig. S2F). As regeneration progressed, the number of newly emerging dynamic peaks decreased, until 24 to 48 hpa, when chromatin regions tended to close (Fig. 1G, fig. S2F, and table S3).

We sought to identify if there were specific transcription factor (TF) binding motifs associated with regions of change during the first 48 hours of regeneration. We used chromVAR (23) to identify matches to known TF binding sites in the Hofstenia genome and to calculate a genome-wide “chromatin variability” score, a measure of how dynamic chromatin regions containing a particular motif were during regeneration. We reasoned that TFs with binding sites that become more or less accessible during regeneration, i.e., whose binding sites lie in variable chromatin, are candidate regulators of regeneration. Of the 386 motifs that we examined with chromVAR, some were more variably accessible in head fragments (e.g., TCF) or in tail fragments (e.g., NFY), likely corresponding to TF families that play important roles in tail and head formation, respectively (table S4, A and B). Many motifs showed highly variable accessibility in both head and tail fragment data (fig. S3A and table S4C). Notably, the binding site for an early growth response (Egr) protein emerged as the most significantly variable (adjusted p < 10−10, Brown-Forsythe test), i.e., associated with the most dynamic chromatin, in our combined analyses of head and tail regeneration in Hofstenia (Fig. 2A, fig. S3A, and table S4C). Regions containing EGR binding motifs were closed at 0 hpa, then open at 3 hpa, with these sites remaining accessible until 48 hpa, when the chromatin regions encompassing these motifs closed (Fig. 2B and fig. S3B).

Fig. 2

Chromatin dynamics, expression, and functional assays implicate egr as a master regulator of regeneration in Hofstenia. (A) Plot of chromVAR chromatin variability scores for 386 transcription factor binding motifs across all time points of regeneration assayed in both head and tail fragments, indicating EGR as the most variable motif. (B) t-Distributed stochastic neighbor embedding (t-SNE) plot showing groupings of samples, including biological replicates, based on chromatin accessibility of the consensus peak set. Samples are colored on the basis of their time point. Right panel shows the accessibility of EGR binding motifs overlaid on the same t-SNE plot, indicating opening of chromatin (red color) at sites by 3 hpa, followed by closing (blue color) at 48 hpa (see fig. S3B for combined tail and head fragment analysis). (C) Left: egr was expressed at the wound sites of head and tail fragments (red arrows) at 1 hpa relative to 0 hpa (black arrows) (26/28). Scale bars, 500 μm. Right: Compared to control RNAi animals that regenerated normally (white arrows), 100% (n = 150) egr-RNAi head and tail fragments failed to regenerate (yellow arrows). Phenotypes shown here were observed 7 days post-amputation (dpa). Dashed line in schematic shows plane of amputation. Scale bars, 200 μm. (D) Heatmap of RNA-seq data showing the expression of 13 validated wound-induced genes in tail fragment samples during regeneration, including 6 hpa egr-RNAi animals. Twelve out of 13 wound-induced genes failed to be up-regulated (p < 0.05, likelihood ratio test) at 6 hpa in the egr-RNAi condition. Additionally, egr, which was targeted via RNAi, was also down-regulated relative to controls. (E) In situ hybridization validation of egr-mediated control of a subset of genes shown in (D). egr-RNAi animals failed to show wound-induced expression of runt (20 out of 22), follistatin (19 out of 19), nlk (22 out of 23), and mtss-1 (19 out of 21) relative to controls. Scale bars, 200 μm.

Egr is a master regulator of regeneration

As the EGR motif emerged as the top candidate from the chromVAR analysis and Egr homologs have known roles upon wounding in other animals (2430), we sought to understand the expression dynamics and function of this family of TFs during Hofstenia regeneration. The Hofstenia genome and transcriptome only contains a single gene encoding an Egr ortholog, egr (fig. S3C). In situ hybridization showed that this gene was up-regulated at both anterior- and posterior-facing wound sites of amputated animals as early as 1 hpa (Fig. 2C). Notably, wound-induced egr expression occurred in multiple cell types, including epidermis, muscle, neurons, and neoblasts, the stem cell population that underlies the regenerative capacity of Hofstenia (fig. S3D). We inhibited the expression of egr using RNA interference (RNAi) (egr-RNAi) and found that egr-RNAi animals did not regenerate blastemas, resulting in a failure to make new heads or new tails (100%, n = 150) (Fig. 2C and fig. S3E).

These data suggest that egr might lie at the top of a gene regulatory network (GRN) associated with the initiation of regeneration. To test this hypothesis, we performed RNA sequencing (RNA-seq) at 0, 1, 3, 6, and 12 hpa using the same amputation strategy as for the ATAC-seq experiment. We identified 61 genes that were significantly up-regulated (adjusted p < 0.05, likelihood ratio test) (14) in both head and tail fragments in at least one of the four time points. egr was the most significantly wound-induced gene at 1 hpa in both head and tail fragments (Fig. 2D and table S5). Twenty genes from this list yielded RNA probes with detectable expression, enabling us to validate by in situ hybridization 13 genes, in addition to egr, as wound-induced genes (up-regulated at wound sites by 6hpa) (fig. S3F and table S5).

To determine whether the wound-induced activity of these 13 genes is under the control of egr, we performed RNA-seq after inhibition of egr via RNAi. Twelve out of 13 of these genes showed significantly lower expression at 6 hpa in egr-RNAi animals compared to controls (Fig. 2D and table S5). To validate these results, we performed in situ hybridization for four potential target genes after egr-RNAi and amputation and found a decrease in expression when egr was inhibited (Fig. 2E and fig. S3G). Given that EGR binding sites show high chromatin variability, egr is wound-induced within 1 hpa in multiple cell types, is necessary for regeneration, and lies upstream of the majority of other validated early (1 to 12 hpa) wound-induced genes, we infer that egr is likely a master regulator of early regeneration.

Egr directly regulates wound-induced genes

We next sought to understand the mechanism(s) by which Egr protein regulates other wound-induced genes. We filtered our list of genome-wide regeneration-responsive chromatin peaks for those that contained a predicted EGR binding site and assigned each peak to its closest gene (table S7, A to C) (14). To distinguish between spurious sites and true binding events, we performed ATAC-seq on egr-RNAi animals at 6 hpa. By focusing on peaks that experienced a significant decrease (p < 0.05, Wald test) in accessibility in egr-RNAi relative to control RNAi (which represent likely bona fide Egr targets), we could identify the Egr regeneration cistrome, i.e., the genome-wide cis-acting targets of Egr protein during regeneration (table S7, C and D) (14).

Among the list of putative egr targets, we found that 10 out of the 12 genes affected by egr-RNAi, and egr itself, had a predicted EGR binding site present in a neighboring dynamic peak (Fig. 3A and table S7). Analysis of ATAC-seq data in egr-RNAi at 6 hpa revealed that these dynamic peaks depend upon the action of Egr, as the regeneration-responsive chromatin regions that contain EGR-binding sites failed to open at 6 hpa in egr-RNAi animals and instead resembled the chromatin profile of the control (time 0) fragments (Figs. 3A and fig. S4). These experiments provide functional validation of predicted binding sites in the loci of wound-induced genes and suggest that Egr directly regulates their transcription.

Fig. 3 Egr is a direct transcriptional regulator of wound-induced genes and likely has pioneer factor activity.

(A) ATAC-seq tail fragment profiles at 0 hpa (blue), 6 hpa (red), and egr-RNAi 6 hpa (green) at the loci of validated wound-induced genes. Predicted EGR sites are denoted as black tick marks below. A 10-kb window containing the egr locus is shown, with magnified views (~800 bps) of regeneration-responsive peaks in the loci of egr and other wound-induced genes shown below. See fig. S4 for images of all peaks in the context of larger surrounding genomic regions. (B) Heatmaps of RNA-seq (transcript reads) and ATAC-seq (genomic DNA reads) profiles of validated wound-induced genes, showing correspondence of statistically significant changes of both gene expression profile and corresponding chromatin accessibility (tables S3E, S5A, and S7) (14). (C) Frequency distribution of PIQ-derived chromatin opening index scores for 439 JASPAR motifs at the 6 hpa time point of tail fragment regeneration. Green arrows indicate factors among the top 11 TFs that have previously reported pioneer factor activity, with the score for the EGR motif highlighted in red.

We propose that the induction of egr expression upon amputation launches a cascade of transcriptional events, with the 10 wound-induced genes identified here as a subset of direct Egr targets. The Egr targets encode diverse molecular functions, including homologs of two transcription factors (runt, deaf1), components of several signaling pathways (follistatin, nrg-1, nrg-2, nlk), and cytosolic proteins with varied functions (p-protein, ankrd, mtss-1), as well as a protein with unclear homology (wound induced expression-1; wie-1). These targets are likely up-regulated in different cell types to mediate varied downstream effects: for example, follistatin was specifically expressed and up-regulated in longitudinal muscle after injury in Hofstenia (figs. S3H and S6B), much like its ortholog in planarians (31).

This preliminary regeneration GRN appears to have an autoregulatory loop, as the egr locus itself has an EGR binding site contained within a regeneration-responsive peak that was lost upon egr-RNAi. Two wound-induced genes, the homologs of a growth factor receptor (ntrk3) and a tumor necrosis factor (TNF) receptor-associated factor (traf), showed significantly diminished expression in egr-RNAi, but the lack of EGR binding sites in their loci suggests either the presence of cryptic EGR binding sites or that they are indirectly regulated by Egr. The one validated wound-induced gene, a homolog of an intraflagellar transport protein (ift80), that did not lose its wound-induced expression activity in the egr-RNAi condition also did not have a regeneration-responsive chromatin region containing an EGR site, suggesting that alternative, egr-independent pathways also operate during early stages of Hofstenia regeneration (Fig. 3B). We noted a correspondence of chromatin profiles at EGR binding sites with RNA expression levels of the associated genes in our analysis of validated wound-induced genes (Fig. 3B), suggesting a causal relationship between specific chromatin loci and their target genes.

Egr shows pioneer factor-like activity

Peaks containing EGR sites in the loci of wound-induced genes appeared to have a nearly binary response, where chromatin at most EGR sites seemed to switch from “closed” to “open” (e.g., in the loci of egr, nlk, nrg-1) (Fig. 3A, fig. S4, and table S7). These chromatin differences were not observed in the egr-RNAi condition, where chromatin failed to open at these loci at 6 hpa relative to control RNAi (Fig. 3, A and B). The pronounced opening of chromatin at EGR sites during regeneration suggests that Egr could be functioning as a pioneer factor.

Pioneer factors are TFs that bind to closed chromatin, resulting in an opening of the region to provide access to other TFs (32). To assess pioneer-like activity of TFs in Hofstenia in a systematic and unbiased manner, we used the protein interaction quantitation (PIQ) algorithm (33), which calculates a “chromatin opening index” that has been used to both validate and predict new pioneer TFs in human data, on the basis of the frequency of transposase cutting sites surrounding TF footprints (14, 33). Pioneer factor footprints show increased levels of local transposase integration relative to those of other TFs. When applied to 6 hpa ATAC-seq data in Hofstenia, we found that the EGR motif exhibited a high chromatin opening index, ranking among the top five of the 439 binding sites analyzed. Further, EGR was nested among other TFs that have known pioneer activity (Fig. 3C and table S8). Thus, we suggest that egr is a master control gene that regulates other wound-induced genes by acting as a pioneer factor in Hofstenia, representing a potentially previously unknown role for this protein family. We note that Hofstenia Egr can also likely act as a more traditional TF regulator by binding to open promoter regions of certain targets (e.g., follistatin).

Extending the Egr-controlled regeneration GRN

Connecting Egr and its direct targets allowed us to assemble the first tier of a GRN for early stages of regeneration. Our data indicate that Egr directly regulates at least two other genes encoding TFs, runt and deaf1. We next sought to predict and validate targets of these two Egr-controlled TFs. deaf1-RNAi animals regenerated normally, but runt-RNAi animals failed to regenerate tails (28 out of 33) and heads (19 out of 29). The phenotype of runt-RNAi was binary: either complete or no regeneration (Fig. 4A). Furthermore, runt homologs are known to be wound-induced and are required for regeneration in planarians (2).

Fig. 4

runt is a downstream target of Egr, is wound-induced and necessary for regeneration, and directly controls nrg-1. (A) Left: In situ hybridization showing that runt is wound-induced at 3 hpa (red arrows) relative to the 0 hpa control (black arrows). Scale bars, 300 μm. Right: In contrast to normal regeneration in controls (white arrows), runt-RNAi resulted in lack of tail (28 out of 33) and head regeneration (19 out of 29) (yellow arrows). Phenotypes observed 7 days post-amputation (dpa) are shown. Dashed line in schematic shows plane of amputation. Scale bars, 200 μm. (B) Violin plot comparing the distribution of fold change of all regeneration-responsive peaks at 6 hpa (relative to 0 hpa) at EGR and RUNT sites. Regeneration-responsive peaks that contain RUNT sites show significantly lower magnitude of change compared to those that contain EGR sites. Asterisk indicates statistical significance in a two-sample t test. (C) Distribution (viewed as density plots) of RUNT binding scores derived from PIQ in the 0 hpa, 6 hpa, and 6 hpa egr-RNAi data samples. RUNT sites that become highly “bound” at 6 hpa, and are dependent on egr, are indicated by a black arrow. (D) Schematic of the nrg-1 promoter with ATAC-seq peaks from 0 hpa, 6 hpa, and 6 hpa egr-RNAi data samples shown below. The location of a predicted RUNT binding site in this locus is shown as a black line, with the corresponding PIQ binding score for each condition shown next to the site. PIQ scores above 0.7 indicate a “bound” state for the TF at the given site. (E) In situ hybridization of the nrg-1 gene in head and tail fragments at 6 hpa in control and runt-RNAi conditions. Normal wound-induced expression in controls is highlighted by black arrows, loss of expression in runt-RNAi is highlighted by red arrows (18 out of 38). Animals were amputated as per the schematic in (A). Scale bars, 200 μm.

runt was the only detected Runt family homolog in Hofstenia (fig. S6A). The predicted RUNT binding motif exhibited a relatively low chromatin opening index score (table S8), suggesting that changes in ATAC-seq peaks associated with this motif during regeneration might be less substantial than for those associated with Egr. We found that regeneration-responsive chromatin peaks were significantly lower in peaks with RUNT sites relative to those for EGR sites (Fig. 4B) (p < 10−10, two-sample t test). These results suggest that Runt is not acting as a pioneer factor and thus would not be expected to elicit large changes in chromatin accessibility.

We sought to assess changes in the binding of Runt protein to its motif genome-wide without relying on large changes in chromatin accessibility. We used PIQ to assess, for every RUNT site in open chromatin in the Hofstenia genome, whether the Runt protein was predicted to be “bound” (binding score ≥ 0.7) or “unbound” (binding score < 0.7) (33) on the basis of transposase accessibility at binding sites (footprinting) (table S9, A and B). By plotting the distribution of binding indices for RUNT sites in 0 hpa, 6 hpa, and 6 hpa egr-RNAi data, we recovered a distinct population of TF binding sites with high binding indices at 6 hpa but not at 0 hpa, which remain “unbound” in egr-RNAi animals (Fig. 4C and table S9A). Because egr is required for activation of runt, this population of sites represents RUNT sites that become “newly” bound during regeneration, likely upon Runt up-regulation by Egr.

Finally, to predict potential targets of bound RUNT sites during regeneration we sorted RUNT sites in descending order of the magnitude of change between the 6- and 0-hpa binding indices and examined the top 10% as a conservative estimate of “newly bound” sites. To identify the RUNT sites most dependent on the Egr-initiated wound response, we ranked these sites by the magnitude of change in binding indices between 6- and 6-hpa egr-RNAi data (table S9). Notably, the second-highest ranked RUNT site (“newly” bound and downstream of egr) in this list was closest to the gene neuregulin-1 (nrg-1), which was identified as a wound-induced gene in Hofstenia in our RNA-seq analysis. nrg-1 is an epidermal growth factor (EGF) family ligand whose homologs are known to operate in regeneration in planarians (34) and vertebrates (35, 36) (Fig. 4D, fig. S6C, and table S9). Our ATAC-seq analysis had identified nrg-1 as being under direct control by Egr in Hofstenia (Fig. 3A and fig. S5A). Owing to the substantial change in RUNT binding score at this locus, we hypothesized that nrg-1 is also under direct control of Runt in Hofstenia (Fig. 4D). To test this, we inhibited runt via RNAi, amputated animals, and assessed the expression of nrg-1 at 6 hpa using in situ hybridization. We found that nrg-1 expression was lost or decreased in the absence of runt activity (18 out of 38), validating that egr–runt–nrg-1 functions as a regulatory cascade in Hofstenia (Fig. 4E and fig. S5B).

Regulatory networks and the evolution of whole-body regeneration

Egr family members are well known for their roles as immediate early genes (IEGs) in many injury and regeneration contexts (24, 25), including in neuronal (26), cardiac (27), liver (24), and immune cells (28) of vertebrates, and have been reported as significantly induced during injury and regeneration in cnidarians (29), planarians (2, 3), and sea stars (30). Here, we identified whole-body regeneration regulatory elements in the acoel Hofstenia to construct a regeneration GRN launched upon wounding and controlled by the master regulator Egr (Fig. 5A).

Fig. 5 Model and evolution of a gene regulatory network for early regeneration.

(A) Model of a gene regulatory network. egr (purple color at wound sites) is wound-induced early, forms an autoregulatory loop, and acts as a likely pioneer factor to directly regulate a suite of downstream genes. This includes the TF runt (green color at wound sites), which binds to the promoter of nrg-1 to up-regulate its expression. Purple shapes and lines indicate Egr protein and its activity, respectively; green shape and line indicate Runt protein and its activity. Blue disks indicate nucleosomes. (B) chromVAR chromatin variability scores for 6 hpa ATAC-seq data from the planarian S. mediterranea and the acoel H. miamia. Select motifs and their associated TF families that are highly variable in both taxa are shown in red. (C) Schematic comparing a portion of the early regeneration network described in this work to what is known in other species capable of whole-body regeneration. Blue color indicates a role in regulating genome-wide chromatin accessibility as determined in this study, purple indicates that the gene is known to be wound-induced. Black arrows indicate direct regulation determined by this study, and gray circles indicate unknown or unstudied expression or function. Sea star and cnidarian silhouettes were obtained from PhyloPic (http://phylopic.org/).

To assess the relevance of our findings to other species, we generated ATAC-seq data for the regenerating planarian (Schmidtea mediterranea) and found that, as observed in Hofstenia, the EGR motif is among the most variably accessible during regeneration (Fig. 5B and table S10). This finding indicates that TFs of the Egr family might also globally regulate chromatin and act as master regulators of regeneration in planarians, which is consistent with the early wound-induced expression of several Egr homologs, one of which is necessary for neural regeneration, in this system (2, 37) (Fig. 5C). Furthermore, homologs of the Egr-controlled GRN are either required for regeneration or are up-regulated during regeneration in distantly related species, including vertebrates (35, 36), planarians (2, 34, 38, 39), and sea stars (30) (Fig. 5C and fig. S6). Constructing regeneration GRNs in additional animals will be crucial to understanding if these networks are conserved or evolved convergently.

Understanding the gene networks that underlie complex processes requires a knowledge of regulatory interactions. In this study, we used the acoel genome to map functional binding sites and to construct a GRN initiating whole-body regeneration. By viewing whole-body regeneration through an epigenomic lens, we uncovered a regulatory mechanism underlying dynamic gene transcription during regeneration. This approach can be applied to established and emerging model systems to gain a deeper mechanistic understanding of the transcriptional cascades that underlie the phenomenon of regeneration.

Material and methods summary

Short-read sequence data were assembled to generate a draft genome for Hofstenia miamia, which was validated using long-read sequencing. ATAC-seq data were obtained from wound sites at “zero” (control), 3, 6, 12, 24, and 48 hpa. These data were used to define regions of chromatin that changed significantly as regeneration progressed. Motif analysis in these regions was used to rank TF binding sites on the basis of dynamic overlying chromatin. The requirement of the top-ranked TF (Egr) for regeneration in Hofstenia was studied via RNAi. RNA-seq data were obtained from wild-type and control and egr-RNAi regenerating worms to identify wound-induced genes and the subset of those under Egr control. The results of these analyses were validated by RNAi and in situ hybridization. To assess direct transcriptional control, regeneration-responsive chromatin regions were assigned to their nearest gene, and ATAC-seq data from control and egr-RNAi worms were analyzed to test for significant differences. To extend the GRN, we used binding site footprinting in our ATAC-seq data to identify dynamically bound TFs, which were linked to their nearest gene as putatives targets and validated using RNAi.

Supplementary Materials

www.sciencemag.org/content/363/6432/eaau6173/suppl/DC1

Materials and Methods

Figs. S1 to S6

Tables S1 to S10

References (4164)

References and Notes

  1. Material and methods are available as supplementary materials.
Acknowledgments: We thank The Bauer Core Facility at Harvard University for technical support with sequencing, T. Sackton (Harvard University) for bioinformatics/statistical assistance, J. Gaspar (Harvard University) for NGmerge and removeChrom scripts, A. Freedman (Harvard University) for assistance with RNA-seq analysis, N. Jeffery and T. Ryan Gregory (University of Guelph) for genome size estimates, and A. Hochwagen (New York University) for help with preparing high-quality genomic DNA. We thank T. Capellini (Harvard University) and all members of the Srivastava Lab for critical reading of the manuscript. Funding: M.S. is supported by the Milton Fund of Harvard University, Searle Scholars Program, Smith Family Foundation, and the National Science Foundation (award no. 1652104). A.R.G. is supported by the Helen Hay Whitney Foundation. Y.-J.L. is supported by the Human Frontier Science Program. D.S.R. and A.B. were supported by NIH grant 1R01HD080708-01, and A.B. was supported by the Biomedical Big Training Program at UC Berkeley and NIH grant 5T32LM012417-03. D.S.R acknowledges support from the Marthella Foskett Brown Chair in Biological Sciences. P.W.R. is an investigator of the Howard Hughes Medical Institute. Author contributions: A.R.G. and M.S. conceived the project and designed experiments. M.S. and P.W.R. initiated the genome and regeneration transcriptome sequencing. A.R.G performed ATAC-seq and RNA-seq library construction and data analysis, chromVAR and PIQ bioinformatics, and Oxford Nanopore Technology sequencing analysis. A.R.G., E.N., L.R., and A.G. performed in situ hybridization and RNAi experiments. Y.-J.L., A.B. J.G.R., D.S.R. and A.R.G. performed genome assembly and statistics. Y.-J.L. constructed an improved transcriptome. R.E.H. performed phylogenetic analysis of genes. A.R.G. and M.S. wrote the paper with input from all of the authors. Competing interests: D.S.R is a member of the scientific advisory board of Dovetail Genomics LLC and a minor shareholder. Data and materials availability: The Hofstenia genome assembly has been deposited at GenBank under the accession SCFE00000000. Other genomic and transcriptomic data for this project can be found under NCBI BioProject PRJNA512373. Data for Schmidtea ATAC experiments can be found under BioProject PRJNA515075. Additionally, Hofstenia data are available for download and genome browsing via http://srivastavalab.rc.fas.harvard.edu. Source code for ATAC-seq, RNA-seq, chromVAR, PIQ, Oxford Nanopore, and REAPR pipelines can be found at zenodo (40). Hofstenia are distributed under a nonrestrictive materials transfer agreement allowing work for academic and nonprofit purposes.
View Abstract

Stay Connected to Science

Navigate This Article