ABSTRACT
Eisenia fetida, the common vermicomposting earthworm, shows robust regeneration of posterior segments removed by amputation. During the period of regeneration, the newly formed tissue initially contains only undifferentiated cells but subsequently differentiates into a variety of cell types including muscle, nerve and vasculature. Transcriptomics analysis, reported previously, provided a number of candidate non-coding RNAs that were induced during regeneration. We found that one such long non-coding RNA (lncRNA) is expressed in the skin, only at the base of newly formed chaetae. The spatial organization and precise arrangement of the regenerating chaetae and the cells expressing the lncRNA on the ventral side clearly support a model wherein the regenerating tissue contains a zone of growth and cell division at the tip and a zone of differentiation at the site of amputation. The temporal expression pattern of the lncRNA, named Neev, closely resembled the pattern of chitin synthase genes, implicated in chaetae formation. We found that the lncRNA has 49 sites for binding a set of four microRNAs (miRNAs) while the chitin synthase 8 mRNA has 478 sites. The over-representation of shared miRNA sites suggests that lncRNA Neev may act as a miRNA sponge to transiently de-repress chitin synthase 8 during formation of new chaetae in the regenerating segments of Eisenia fetida.
INTRODUCTION
Earthworms are a large diverse group of segmented worms that inhabit niches just under or deep within the soil. The ‘tube within a tube’ body plan of the earthworm comprises a muscular outer wall enclosing a gut within. It also has a simple vascular system to circulate blood and a nervous system comprising a nerve ganglion at the anterior end and a long ventral nerve cord running the length of the body, with ring nerves in each segment.
Earthworms vary widely in their ability to regenerate. Eisenia fetida (commonly known as red wriggler worm) regenerates nearly 2/3 of its posterior end (Xiao et al., 2011). The earthworm presents an invertebrate model of epimorphosis, a type of regeneration involving the restoration of original anatomy and polarity followed by de-differentiation, proliferation and differentiation of cells (Bely, 2014; Gazave et al., 2013; Planques et al., 2019; Xiao et al., 2011). As each segment consists of nerve, muscle, vasculature and additional specialized structures, it provides a model for studying regeneration coordinated across different tissue types. For instance, chaetae, specialized projections embedded in the skin used for gripping the soil, are found embedded in the outer body wall in close proximity to muscle and peripheral nerves in each segment (Prosser, 1934).
We have previously characterized the genome and transcriptome of the E. fetida (Bhambri et al., 2018). Injury and loss of the posterior 2/3rd of this regenerating worm was followed by apparent wound healing in 5–10 days. A stub of tissue largely consisting of a mass of undifferentiated tissue was formed by 15 days and differentiated segments were formed by 20 days post-amputation. The period between 10 and 20 days after the injury presents a time window during which cell proliferation, growth and differentiation happen simultaneously in a 4–5 mm long tissue amenable to molecular and cellular visualization. Regenerating annelids are particularly convenient for studying developmental gradients, because a single regenerating tail has many segments at varying stages of development along the anterio-posterior axis.
The transcriptome of the regenerating worm revealed signatures of rapid cell proliferation, reorganization of the extracellular matrix and differentiation of nerves. Besides these signatures, we also reported the dynamic expression of non-coding RNAs that potentially play roles in regulating the timing of expression, control and spatial organization of the transcriptome (Bhambri et al., 2018). We rationalized that novel non-coding RNAs could play an important role in restoring an undifferentiated cellular state by interfering in the function of cell type-specific microRNAs (miRNAs) that are usually associated with differentiation. In mammalian systems, several long non-coding RNAs (lncRNAs) are known to carry tandem miRNA binding sites, allowing them to ‘sponge away’ miRNAs (Paraskevopoulou and Hatzigeorgiou, 2016). Here, we report the expression pattern of selected non-coding RNAs in the regenerating earthworm. Besides validating our previous report, we focus on one lncRNA that showed a unique expression pattern at the base of the chaetae.
Chaetae are stiff chitinous structures (appendages) that originate deep within the muscular body wall, the outer ends of which are used to grip the surface and in locomotor activity (Hausen, 2005). In another annelid, Platynereis dumerilli, larval chaetogenesis starts with the appearance of chaetoblasts on the surface, which later invaginate to the base of the chaetae, forming chaetal sacs, while the surrounding cells become follicle cells (Gazave et al., 2017). We found that this lncRNA, named Neev, is expressed only in the few cells at the base of chaetae in newly regenerated segments close to the site of injury. Its expression pattern closely resembles that of chitin synthase and chitinases involved in the formation of chaetae. We therefore explored the relationship between Neev and chitin synthase expression in regenerating segments of E. fetida.
MATERIALS AND METHODS
Experimental conditions
Eisenia fetida (Savigny 1826) earthworms were originally procured from farmers engaged in vermicomposting and subsequently maintained in a plastic tray and fed with plant matter in the laboratory at around 22°C for several years. No specific permissions were required for procuring earthworms. They are not endangered species and ethical approval is not required. Medium-sized worms were collected before the experiment, rinsed in tap water to remove any soil sticking to the surface and amputated as described in our previous paper (Bhambri et al., 2018). The site of amputation was at about 2/3rd of the body length from the anterior end, thus retaining about 60 segments. After amputation, the worms were maintained in a separate container but under similar culture conditions. Regenerating tissue and about 2–3 mm of the adjacent tissue from the pre-amputated worms were collected for in situ hybridization.
Bioinformatic analysis
The sequence of Neev lncRNA was used to search for contigs in the genome assembly for E. fetida previously reported from our group (Bhambri et al., 2018). The longest contig within scaffold SACV01159372.1 (Genome ID: SACV01) containing Neev sequence was 4884 nucleotides (nt) long. Further, this contig was used to identify the similar regions from the genome of any eukaryotes using a BLASTn search optimized for short stretches of low similarity (Expect value threshold=10, Word size=11, Match score=2, Mismatch penalty=−3).
RNA isolation and qRT-PCR
The regenerated earthworm was rinsed thoroughly in running tap water followed by autoclaved Milli-Q water; regenerated and adjacent control tissue from 20–30 earthworms was collected in 1 ml Trizol kept on ice, at 15, 20 and 30 days post-amputation (dpa). A homogeneous cell suspension was made by grinding the tissue using a homogenizer; 200 μl chloroform was then added, followed by vigorous shaking for 15 s. After incubation for 10 min at room temperature for phase separation, the mix was centrifuged at 10,000 g, 15 min, 4°C. The upper aqueous layer was separated and an equal volume of isopropanol was added, incubated for 5 min, and centrifuged at 10,000 g for 10 min. The resulting pellet was washed thrice with 70% ethanol at 10,000 g for 5 min each. The air-dried pellet was dissolved in 50 μl nuclease-free water. RNA (1 μg) was used to make cDNA using a Transcriptor High Fidelity cDNA Synthesis Kit (Roche, 5081955001) and primers (FP: ATATGGTACCGTCTGCTCCCAGGGTTAG; RP: ATATGCGGCCGCCTTGTGTCGAGTGTATTCAATTGC) designed to amplify full-length transcript. Gene-specific primers listed in Table S1 were used in quantitative RT-PCR (qRT-PCR) reactions containing SYBR green master mix (Takara, RR820). The Ct values were used to calculate fold-change against spike-in control lncRNA (see Results for details) using the method described by Pfaffl (2001).
PCR cloning and probe design
PCR product from the RT reaction was cloned using TOPO™ TA Cloning™ Kit (Invitrogen, 450640) as per the manufacturer's protocol. Sanger sequencing was performed to confirm the sequence of the clone. In vitro transcription to synthesize the probe was performed using a DIG RNA (SP6/T7) Labeling Kit (Sigma-Aldrich, 11175025910) with SP6 or T7 polymerase after linearizing the plasmid using restriction enzymes. The probes were purified by using NucAway™ Spin Columns (Thermo Fisher Scientific, AM10070).
In situ hybridization
Regenerated earthworms were collected at 10, 15, 20 and 30 dpa and washed thoroughly in running tap water followed by autoclaved Milli-Q water. Earthworms were fixed overnight at 4°C in 4% (w/v) paraformaldehyde (PFA) prepared in 1× phosphate-buffered saline (PBS). After fixation, they were washed stringently in PBS with Tween 20 (PBST; 0.1% Tween 20 in PBS) with subsequent storage in 100% methanol at 4°C. Prior to hybridization, the stored earthworms were rehydrated with a methanol gradient [90%, 75%, 50%, 25% and 0% (v/v) in PBST] for 45–60 min each. Earthworms were permeabilized by 20 µg ml−1 proteinase K for 45 min at 55°C, then fixed again in 4% PFA for 20 min and blocked using hybridization buffer [50% formamide, 1.3× saline sodium citrate (SSC), 5 mmol l−1 EDTA, 5% dextran sulphate, 0.2% Tween 20, 100 µg ml−1 heparin and 50 µg ml−1 yeast t-RNA in DEPC-treated water] for 60 min at 65°C. Hybridization was performed using sense and antisense probes, prepared in hybridization buffer, overnight at 65°C in a water bath. Stringent washes were performed at 65°C with hybridization buffer thrice for 30 min each followed by washes in Tris-buffered saline with Tween (TBST; 0.1% Tween 20 in TBS) for 15 min at room temperature. After incubation at room temperature for 4 h in 1:2000 dilution of anti-digoxigenin antibody (Roche, 11376623) prepared in TBST containing 10% FBS, 15 min washes in TBST at room temperature were performed thrice and subsequently the tissue was stained using Nitro Blue Tetrazolium (NBT; working concentration 500 µg ml−1; Roche, 11383213001) and 5-bromo-4-chloro-3-indolyl phosphate (BCIP; working concentration 562.5 µg ml−1; Roche, 11383221001) in developing solution (0.1 mol l−1 NaCl, 0.1 mol l−1 Tris-HCl pH 9.5, 0.05 mol l−1 MgCl2, 0.1% Tween 20 in DEPC-treated water). Images were captured by mounting earthworms in 2.5% methylcellulose at 3.2× and 5× magnification.
RESULTS
On the basis of transcriptomics analysis in the regenerating earthworm in our previous report (Bhambri et al., 2018), we prioritized 16 potentially non-coding RNAs for further validation, as they were about 1 kb or larger and free of low complexity repeats. We designed qRT-PCR assays to detect four of the predicted lncRNA using the assembled transcript sequences (see Materials and Methods; Fig. S1). Although GAPDH is widely used as a control in gene expression studies, it was not suitable for normalization in our experiments because it is strongly upregulated during regeneration. Instead, we used a spike-in normalization method, by adding an in vitro transcribed RNA fragment to the qRT-PCR reaction. The spike-in control lncRNA was originally cloned from the zebrafish genome that has no sequence homology with the earthworm genome (Sarangdhar et al., 2017). We also verified that the primers for this fragment produce no product when provided with the earthworm cDNA as a template. In close agreement with the RNAseq data, all four lncRNAs were strongly over-expressed in the regenerating tissue (Fig. 1). As shown in the figure, the newly regenerated segments expressed the lncRNAs at 4–8 times the levels in the adjacent control segments.
Next, we cloned each lncRNA gene into the TOPO TA cloning vector and generated digoxigenin-labelled probes by incorporating digoxigenin-linked rUTP in the in vitro transcription reaction. These probes were used in in situ hybridizations with collected samples containing regenerating tissue closely juxtaposed with tissue from the worm before injury. Although there was a detectable signal in the regenerating tissue, the expression of the lncRNAs did not, in general, have a distinctive spatial pattern (Fig. S2). A notable exception was the 895 nt lncRNA, which showed a recurring pattern of expression with four spots in each segment on the ventro-lateral and ventral side in the regenerating region (Fig. 2). From the Ct values, this lncRNA was estimated to be expressed at about one-tenth the abundance of GAPDH, making it quite abundant, as lncRNAs are usually expressed at low levels. Because this abundant lncRNA was restricted to tiny spots, the expression was strong and clearly visible.
To understand the source of the signal (Fig. 3A), we made longitudinal incisions on the dorsal region and spread out the inner body wall. By gently teasing out the tissue around the signal, it was clear that the spots were at the base of newly assembled chaetae (Fig. 3B,C). Notably, no such spots were seen at the base of chaetae from non-regenerating segments (Fig. 3A). Clearly, this lncRNA was strongly but transiently induced in a very small group of cells closely associated with the chitinous setae. Because of this interesting expression pattern, we named the lncRNA Neev, which means base or foundation in Hindi.
In transcriptomics experiments, fragments of protein-coding mRNAs are sometimes erroneously annotated as non-coding transcripts (Zhao et al., 2018). Some legitimate lncRNAs may also produce functional peptides from micro-open reading frames (microORFs) (Ruiz-Orera et al., 2014). To rule out spurious annotation and detect conserved microORFs, we aligned the sequence of the Neev lncRNA to genome scaffolds assembled previously. Although there was an 82 amino acid open reading frame, it did not show any similarity to known proteins (Fig. 4). The scaffold from which the Neev gene was derived included a conserved region of 232 nt with strong similarity to many genomes including vertebrate model systems like zebrafish. This region was used to anchor genomic contigs from diverse species (Fig. 4C). However, the rest of the contig did not show any conservation, suggesting that Neev is not present in other organisms. By aligning the transcript sequence to the genomic sequence, we found that the Neev gene comprises two exons and a 191 nt long intron (Fig. 4B). Taken together, Neev codes for a lncRNA that satisfies all the current criteria, i.e. multi-exon, polyadenylated transcript longer than 200 nt that is devoid of open reading frames (ORFs) larger than 300 nt (Clamp et al., 2007; Dinger et al., 2008; Frith et al., 2006; Niazi and Valadkhan, 2012).
Next, we tried to assign a potential function to the transcript. As lncRNAs often regulate overlapping genes or genes in close proximity by RNA–DNA hybridization and recruitment of chromatin modifiers, we first checked for relevant ORFs in the 5 kb contig containing the Neev gene. Because there were no genes in this region, we speculated that the lncRNA might regulate the expression of distant genes through RNA–RNA binding. Chaetae, i.e. chitinous setae, originate in bulbous cells called chaetoblasts, which put out microvilli that are subsequently coated with a large amount of chitin, presumably produced within these cells (Schweigkofler et al., 1998). We reasoned that the chaetoblasts would also need to express chitin synthase genes transiently and in a highly regulated manner. We checked our transcriptomics data for the expression pattern of chitin synthase genes and chitinases. Amongst 29 genes with the word chitin in their name, 10 changed expression during regeneration. We retrieved the basal expression level of these genes and in agreement with our prediction, one of the chitin synthases, chitin synthase 8 (Chs8; Q4P9K9), showed a strong induction of 11- to 23-fold in the regenerating tissue, compared with the adjacent control tissue (Fig. 5).
We checked the spatial expression pattern on Chs8 by in situ hybridization in the regenerating earthworm at 10 and 15 dpa. The expression signal was strikingly similar to that of Neev, although of much lesser intensity. This was in agreement with the relative expression (as interpreted from baseMean values) in the transcriptomics data. At higher magnification, expression was also revealed at the neck of the chaetae closer to the surface, while Neev was expressed at the base (Fig. 6).
Next, we looked for potential RNA–RNA interactions that implicate the lncRNA in chitin synthase regulation. We aligned the lncRNA sequence with those of differentially expressed chitin synthases. Similarity in the anti-sense orientation would indicate that lncRNA–mRNA duplexes could potentially form, which are usually targeted for degradation. More complex regulatory mechanisms like guidance of splicing or RNA–RNA scaffold formation are also possible. In the anti-sense orientation, the mRNA of Chs8 could potentially bind to the lncRNA only at four stretches of 7 nt each.
We also looked for potential miRNA sponge-like activity in the lncRNA sequence, because sequestration of miRNAs may transiently de-repress chitin synthesis genes to facilitate regeneration of chaetae. As miRNAs tend to be highly conserved, we used the list of mouse miRNAs (Griffiths-Jones, 2004; Griffiths-Jones et al., 2006, 2008; Kozomara and Griffiths-Jones, 2011, 2014) to predict targets and verified that the earthworm genome contained sequences corresponding to the miRNAs of interest (Bhambri et al., 2018). We used the well-accepted miRNA target prediction tool miRanda (Betel et al., 2010) to identify the most frequently occurring miRNA targets in Neev (Table S2). Two miRNAs were discarded from further analysis because we could not find the corresponding region in the earthworm genome. Four miRNAs, each with more than five target sites in the Neev lncRNA (see Fig. S2; Fig. 4A) collectively had 49 sites of ΔG<−20 kcal mol−1 in the 895 nt lncRNA Neev. Next, we checked the Chs8 mRNA sequence (6202 nt) to check for binding sites against the miRNAs. The four selected miRNAs had >100 sites on average and collectively had 478 potential binding sites (ΔG<−20 kcal mol−1). To check for the rate of false positive predictions by the algorithm, we ran similar predictions for miRNA binding sites in three endo-chitinase genes, but none of them showed a comparable enrichment for binding sites of these miRNAs (Table S2). We also used 15 randomly selected 6 kb regions (matching the length of Chs8 mRNA) from the earthworm genome, to ensure that the high number of targets was not a consequence of the length of the chitin synthase mRNA. By comparing the number of sites predicted in the randomly selected controls, we conclude that binding sites for miR-667-5p and miR-7658-5p are over-represented in the Chs8 mRNA and the Neev lncRNA at a frequency significantly higher than expected by chance (P<0.001).
DISCUSSION
lncRNAs are a large group of transcripts that are more than 200 nt in length, with minimal or unclear peptide coding potential (Fang and Fullwood, 2016). lncRNAs are known to fold back into complex structures and mediate diverse functions in cells ranging from recruiting chromatin modifiers to guiding alternative splicing and sequestering miRNAs and proteins (Fernandes et al., 2019). We have previously identified several non-coding RNAs that were highly induced in the regenerating region of the earthworm from RNAseq data (Bhambri et al., 2018). The transcriptomics data were used in de novo assembly of transcript sequences, which were further classified as non-coding if they did not contain an ORF of more than 300 nt in length. In agreement with the RNAseq study, qRT-PCR validation also showed that the lncRNAs were induced in the regenerating region of the earthworm. The level of induction increased with time, till about 20 dpa, but then tapered off. This coincides with the time frame when the regenerated segments acquire almost all the features of the original segments and the transcription profile is largely restored to the normal pattern.
One of these lncRNAs, named Neev, is expressed at the base of regenerating chaetae. To the best of our knowledge, this lncRNA has no ability to encode a protein. While many dubious lncRNAs are now thought to be spurious products of abortive transcription (Consortium and The FANTOM Consortium, 2005; Ebisuya et al., 2008; Struhl, 2007), the presence of two exons in Neev suggests that its designation as a lncRNA is more reliable (Derrien et al., 2012). It is polyadenylated, and only expressed from one strand, i.e. it does not overlap with a protein-coding gene. We could not find any ORF in close proximity within the 5 kb contig containing the gene for Neev. Translating the 895 nt transcript in all potential reading frames did not reveal any peptide with even minimal homology to a known gene. Unlike the typical lncRNA, Neev is expressed at high levels, to about one-tenth of the abundant GAPDH mRNA. Comparing the expression levels from the RNAseq data, it appears that Neev is more abundant than the Chs8 gene. Taken together, the reliable expression, poor conservation and presence of multiple exons agrees with it being a functional non-coding RNA.
Non-coding RNAs are found in large numbers in every genome, frequently outnumbering their protein-coding counterparts (Derrien et al., 2012). The transient and localized expression of Neev is in agreement with the view that lncRNAs may be involved in establishing the fine spatio-temporal regulation of genes in the regenerating tissue. lncRNAs are known to regulate gene expression at various levels: by recruiting chromatin modifiers (Campos and Reinberg, 2009; Kanhere et al., 2010), directing splicing (Bernard et al., 2010; Tripathi et al., 2010), modifying stability of mRNAs by masking motifs within the mRNAs (Kung et al., 2013; Matsui et al., 2008) or sequestering miRNAs (Faghihi et al., 2010; Franco-Zorrilla et al., 2007). Some of these mechanisms inherently involve RNA–protein complexes, which cannot be predicted on the basis of RNA sequence. However, some of the mechanisms can be predicted from the sequence of mRNA and lncRNA. The direct binding of lncRNA and mRNA can be deciphered from stretches of apparent complementarity while similarity at miRNA binding sites indicates the possibility of a miRNA sponge mechanism. On the basis of the high frequency of binding sites for miR-667-5p and miR-7658-5p in the mRNA of Chs8 mRNA and Neev lncRNA, we speculate that the transient induction of the lncRNA in a few cells at the location of the chaetae temporarily releases chitin synthase mRNA from repression, and facilitates the development of chaetae in the regenerating segments (Fig. 7). Further experiments are needed to test the predicted miRNA sponge-like activity of the lncRNA in vivo.
Drawing upon other studies and our own results, we propose the following hypothetical scenario in the regenerating tissue of earthworm. Immediately following injury, a mass of undifferentiated tissue, called the blastema, caps the site of injury. Within days, the mass of cells starts differentiating to structures found within each segment even as rapidly dividing cells increase the volume of tissue. It was not clear whether the temporal period of growth and cell division is completed before differentiation ensues; our data imply that differentiation and cell proliferation happen simultaneously. The rapidly dividing cells are concentrated at the posterior tip while the differentiation zone abuts the site of injury. This is broadly in agreement with a recent report of zones within the posterior regenerating region of another annelid, Platynereis dumerilii (Gazave et al., 2013). Cells that exhaust their division potential are perhaps pushed down into the zone of differentiation where, presumably, local signals precisely position the chaetoblasts, which then adopt a transcriptional programme distinctive from that of the neighbouring cells. At the site of injury, the newly formed segment has chaetoblasts expressing the Neev lncRNA, even as the pre-existing chaetoblasts (in the adjacent segment on the other side of the site of injury) remain immune to these signals. Thus, the position of the regenerating chaetae and the pattern of Neev expression can help in inferring an invisible developmental boundary within the regenerating tissue.
Acknowledgements
The authors acknowledge Aksheev Bhambri for support with genome analysis. An anonymous reviewer is acknowledged for suggestions that improved the manuscript.
Footnotes
Author contributions
Conceptualization: S.S.P., B.P.; Methodology: S.S.P., S.Z.; Validation: S.S.P., S.Z.; Formal analysis: S.S.P.; Investigation: S.S.P.; Resources: B.P.; Data curation: S.S.P., B.P.; Writing - original draft: B.P.; Writing - review & editing: S.S.P., B.P.; Visualization: B.P.; Supervision: B.P.; Project administration: B.P.; Funding acquisition: B.P.
Funding
S.S.P. is supported through a Senior Research Fellowship from the University Grants Commission.
References
Competing interests
The authors declare no competing or financial interests.