A Novel Taxon of Monosegmented Double-Stranded RNA Viruses Endemic to Triclad Flatworms

Planarians are freshwater flatworms, related more distantly to tapeworms and flukes, and have been developed as models to study the molecular mechanisms of stem cell biology and tissue regeneration. These worms live in aquatic environments, where they are likely to encounter a variety of viruses, bacteria, and eukaryotic organisms with pathogenic potential. How the planarian immune system has evolved to cope with these potential pathogens is not well understood, and only two types of planarian viruses have been described to date. Here, we report discovery and inaugural studies of a novel taxon of dsRNA viruses in five different planarian species. The virus in the best-characterized model species, Schmidtea mediterranea, appears to persist long term in that host while avoiding endogenous antiviral or RNAi mechanisms. The S. mediterranea virus-host system thus seems to offer opportunity for gaining new insights into host defenses and their evolution in an important lab model.

ABSTRACT

Freshwater planarians, flatworms from order Tricladida, are experimental models of stem cell biology and tissue regeneration. An aspect of their biology that remains less well studied is their relationship with viruses that may infect them. In this study, we identified a taxon of monosegmented double-stranded RNA (dsRNA) viruses in five planarian species, including the well-characterized model Schmidtea mediterranea. Sequences for the S. mediterranea virus (abbreviated SmedTV for S. mediterranea tricladivirus) were found in public transcriptome data from multiple institutions, indicating that SmedTV is prevalent in S. mediterranea lab colonies, though without causing evident disease. The presence of SmedTV in discrete cells was shown through in situ hybridization methods for detecting the viral RNA. SmedTV-staining cells were found to be concentrated in neural structures (eyes and brain) but were also scattered in other worm tissues as well. In contrast, few SmedTV-staining cells were seen in stem cell compartments (also consistent with RNA sequencing data) or early blastema tissue. RNA interference (RNAi) targeted to the SmedTV sequence led to apparent cure of infection, though effects on worm health or behavior were not observed. Efforts to transmit SmedTV horizontally through microinjection were unsuccessful. Based on these findings, we conclude that SmedTV infects S. mediterranea in a persistent manner and undergoes vertical transmission to progeny worms during serial passage in lab colonies. The utility of S. mediterranea as a regeneration model, coupled with the apparent capacity of SmedTV to evade normal host immune/RNAi defenses under standard conditions, argues that further studies are warranted to explore this newly recognized virus-host system.

IMPORTANCE Planarians are freshwater flatworms, related more distantly to tapeworms and flukes, and have been developed as models to study the molecular mechanisms of stem cell biology and tissue regeneration. These worms live in aquatic environments, where they are likely to encounter a variety of viruses, bacteria, and eukaryotic organisms with pathogenic potential. How the planarian immune system has evolved to cope with these potential pathogens is not well understood, and only two types of planarian viruses have been described to date. Here, we report discovery and inaugural studies of a novel taxon of dsRNA viruses in five different planarian species. The virus in the best-characterized model species, Schmidtea mediterranea, appears to persist long term in that host while avoiding endogenous antiviral or RNAi mechanisms. The S. mediterranea virus-host system thus seems to offer opportunity for gaining new insights into host defenses and their evolution in an important lab model.

INTRODUCTION

The phylum Platyhelminthes encompasses both parasitic and nonparasitic species of flatworms. Two taxonomic classes, Cestoda and Trematoda, include globally important parasites of humans (tapeworms and flukes), and another, Monogenea, includes parasites of fish (1, 2). Planarians, most of which are nonparasitic, fall outside these three classes, and planarian species used in lab research are specifically members of order Tricladida. Among these triclad planarians, Schmidtea mediterranea is a primary laboratory model of stem cell biology and tissue regeneration (3).

Following manual division, planarian body fragments can regenerate lost tissues to become whole animals again, thanks to a large population of pluripotent stem cells (4–6). In fact, by serially applying this method, clonal lines of S. mediterranea and other planarians can be maintained in lab colonies over extended periods, without need for sexual reproduction, and some are obligate asexuals. From a logical standpoint, such conditions would seem well suited for allowing persistent viruses to be maintained in these colonies as well, with ongoing viral transmission either vertically (directly from manually divided “mother” to regenerated progeny worms) or horizontally (through the nonsterile aqueous culture medium in which the worms are serially passaged).

To date, the only viruses reported for planarians are two unclassified small DNA viruses in Girardia tigrina (7, 8) and a recently identified large plus-strand RNA virus (order Nidovirales, family Mononiviridae) in S. mediterranea and probably Planaria torva (9). Given these limited findings, many questions remain. How many different virus species might constitute the natural planarian virome? What effects might these natural symbionts have on planarian biology? And what effects on viral biology might the planarian immune defenses play?

Recent years have seen an explosion of virus discovery, rooted in improving methods for high-throughput sequencing of biologically complex samples (10, 11). For the purposes of this study, we took account of the fact that RNA sequencing (RNA-seq) data from numerous transcriptome studies of S. mediterranea and some other planarians are available online, reflecting the utility of such studies for identifying key factors in tissue regeneration and differentiation (12–14). We therefore decided to probe these data for novel planarian viruses by bioinformatic means.

Here, we report evidence for a new taxon of monosegmented double-stranded RNA (dsRNA) viruses related to totiviruses (15) from five different species of triclad planarians, including several clonally related lab colonies of S. mediterranea. As described in detail below, we additionally used in situ hybridization, RNA interference (RNAi), and other methods to begin to address biological questions about the virus in S. mediterranea, including its distribution in different cell types and susceptibility to cure. These findings set the stage for further studies of these viruses, including their effects on different aspects of planarian biology and viral mechanisms of immune evasion.

RESULTS

Database evidence for related dsRNA viruses in five planarian species.

Progressive searches were performed on both the PlanMine database (14) and the NCBI Transcriptome Shotgun Assembly (TSA) database for Platyhelminthes (taxid 6157), starting with the RNA-dependent RNA polymerase (RdRp) sequences of several toti-like viruses (related to totiviruses, but not yet formally classified) as queries. These searches ultimately yielded the full-length coding sequences of five novel toti-like viruses from five respective species of flatworms. The flatworm species were Schmidtea mediterranea, Planaria torva, Phagocata morgani, Dendrocoelum lacteum, and Bdelloura candida. All species belonged to the order Tricladida: B. candida to suborder Maricola and the others to suborder Continenticola (suborder Cavernicola is not yet represented). Among these, B. candida, an ectosymbiont of horseshoe crab Limulus polyphemus, is the only saltwater species and the only species that is not free living (16). We propose the name tricladiviruses for these newly discovered members (Planaria torva tricladivirus [PtorTV], Phagocata morgani tricladivirus [PmorTV], Dendrocoelum lacteum tricladivirus [DlacTV], and Bdelloura candida tricladivirus [BcanTV]) of the planarian virome.

Features of the tricladivirus genomes are shown in Fig. 1. They span similar lengths (7,808 to 8,511 nucleotides [nt]) and exhibit a shared coding strategy encompassing three long open reading frames (ORFs). The longest ORF (designated ORF1) is overlapped by both of the other two: the shortest ORF (designated ORF0) overlaps the 5′ end of ORF1, and the middle-sized ORF (designated ORF2) overlaps the 3′ end of ORF1. ORF0 is found in the −1 frame relative to ORF1, and translation of these two ORFs seems likely to involve distinguishable initiation mechanisms since their putative AUG start codons are fairly closely juxtaposed in their respective frames (Fig. 1). ORF2, in contrast, is found in the +1 frame relative to ORF1 and seems likely to be translated in fusion with ORF1 via either +1 or −2 ribosomal frameshifting. The proposed +1 slippery sequence UUU_C (17) (underline indicates codon boundary in the ORF1 frame), for example, is found in the ORF1-ORF2 overlap region in four of the five viruses. Properties of the deduced viral proteins P0 and P1 as well as predicted fusion protein P1+2 are shown in Table 1 and exhibit strong similarities among the five viruses, particularly for P1 and P1+2. Pairwise sequence identity scores among the five viruses were found to be ≤16.4% for P0, 22.1 to 36.1% for P1, and 35.5– to 42.6% for the P2 region of P1+2 (see Fig. S1 in the supplemental material).

FIG 1.

Scaled diagrams of tricladivirus genomes. Overall lengths of the assembled contigs are indicated at the right (in nucleotides). The genomic RNA plus strand of each virus is shown as a thick red line. The three long ORFs in each virus (ORF0, -1, and -2) are shown as gray rectangles, labeled with the first and last nucleotide positions of each (including stop codons) in black. Evidence that ORF1 and ORF2 encode the viral CP and RdRp is described in the text. The reading frame that includes ORF1 is defined as frame 0, as labeled at left. The first in-frame AUG codon in each ORF is shown as a green line and labeled with the first nucleotide position. The proposed +1 ribosomal frameshifting site (+1fs) in the region of ORF1-ORF2 overlap in each virus is also indicated. The diagrams for the five viruses are aligned by the position of the ORF1 stop codon. The diagram for SmedTV is that for a consensus determined from the originally identified TSA and PlanMine sequences. PtorTV, Planaria torva tricladivirus; PmorTV, Phagocata morgani tricladivirus; DlacTV, Dendrocoelum lacteum tricladivirus; BcanTV, Bdelloura candida tricladivirus.

TABLE 1.

Flatworm toti-like viruses: coding features^a

Virus	ORF0 range (nt)	P0: length (aa)	π	ORF1 range (nt)	P1 length (aa)	π	ORF1+2 ranges (nt)	P1+2 length (aa)	π	Contig length (nt)
SmedTV	262–1,317	351	9.8	389–5,107	1,572	6.1	389–5,074; 5,076–7,880	2,496	8.0	7,905
PtorTV	521–1,540	339	7.1	543–5,378	1,611	5.8	543–5,345; 5,347–8,184	2,546	7.5	8,304
PmorTV	293–1,156	287	10.4	324–4,937	1,537	5.9	324–4,904; 4,906–7,740	2,471	6.8	7,808
DlacTV	678–1,544	288	10.3	700–5,436	1,578	6.2	700–5,403; 5,405–8,221	2,506	7.9	8,227
BcanTV	644–2,218	523	9.5	879–5,723	1,613	6.5	879–5,696; 5,698–8,499	2,539	8.1	8,511

^aRanges include stop codons. SmedTV, consensus sequence from several transcriptomes (see text). ORF1+2, predicted product from +1 programmed ribosomal frameshifting in the ORF1-ORF2 overlap region. aa, amino acids. π represents nucleotide diversity calculated as the average number of nucleotide differences between sequences.

We next used the five tricladivirus sequences as queries in BLASTX searches of the NCBI Reference Sequence (RefSeq) Proteins database for Viruses (taxid 10239). These searches identified two regions of significant similarities to monosegmented dsRNA viruses that have been tentatively assigned to family Totiviridae: a region in the central portion of the queries, encompassed by ORF1, with similarities to the structural (coat) protein of Atlantic salmon piscine myocarditis virus (designated PCMV1 here) or golden shiner piscine myocarditis-like virus (designated PCMV2 here) (18, 19) (top E value, 1.3e−06) and a region in the 3′ half of the queries, encompassed by ORF2, with similarities to the RdRp of Leptopilina boulardi (parasitoid wasp) toti-like virus (LbTV) (20) (top E value, 3.9e−12) (Table 2). BLASTX hits to these same two regions of the tricladivirus sequences were also found for a number of unclassified RNA viruses, mostly named as toti-like viruses in the RefSeq database, that were reported from a large transcriptome survey of invertebrates (21) (Table S1). No hits to the ORF0 region of the tricladiviruses were found.

TABLE 2.

Flatworm toti-like viruses: top hits to dsRNA viruses annotated as such in GenBank^a

Virus	P0 top hit	E value	P1 top hit	E value	P2fs top hit	E value
SmedTV	ND	>10	Piscine myocarditis-like virus	1.5e−06	Leptopilina boulardi toti-like virus	8.6e−09
PtorTV	ND	>10	Piscine myocarditis virus	1.5e−07	Leptopilina boulardi toti-like virus	4.3e−12
PmorTV	ND	>10	ND	>10	Leptopilina boulardi toti-like virus	1.0e−12
DlacTV	ND	>10	Piscine myocarditis-like virus	1.2e−04	Leptopilina boulardi toti-like virus	1.9e−06
BcanTV	ND	>10	Piscine myocarditis virus	4	Leptopilina boulardi toti-like virus	5.4e−09

^aPiscine myocarditis virus, AGA37470.1; Piscine myocarditis-like virus, YP_009229914.1; Leptopilina boulardi toti-like virus, YP_009072448.1. ND, not determined.

S. mediterranea is the best studied of the apparent hosts of the newly discovered planarian viruses, and indeed, four different DNA-based genome assemblies for S. mediterranea are currently available at NCBI (GCA_000691995.1 [asexual strain CIW4] and GCA_000181075.1, GCA_000572305.1, and GCA_002600895.1 [sexual strains S2F2 and S2]). Using the tricladivirus sequence from S. mediterranea (SmedTV) as a query for a Discontiguous MegaBLAST search of these genome assemblies failed to identify any significant similarities (E values, >10), providing evidence that SmedTV derives from an extrachromosomal source in S. mediterranea, most likely an actively replicating virus.

Sequence comparisons and phylogenetic analyses.

We used phylogenetic methods to establish the relationship between the newly found planarian viruses and a larger collection of toti-like viruses. The other viruses included in these analyses were those whose RdRps were found to have strong similarities (E values better than 1e−12) in progressive BLASTP searches of the RefSeq Proteins database, starting with the five tricladivirus RdRp sequences as queries (see Table S2 for all virus names, abbreviations, and RefSeq accession numbers). Following multiple-sequence alignments and maximum likelihood phylogenetic analyses of the RdRps of these viruses, results similar to those shown in Fig. 2A were consistently obtained, indicating that the planarian viruses constitute a discrete phylogenetic clade within this larger collection. In addition to LbTV, PCMV1, and PCMV2, which were identified in the BLASTX searches described above, the collection is notable for including Giardia lamblia virus (GLV), which is an assigned member of the genus Giardiavirus in the family Totiviridae (22, 23). Also notable are three putative arthropod viruses (Behai sea slater virus 3, Behai toti-like virus 4, and Hubei toti-like virus 16) (21) that branch most closely to the tricladivirus clade. These three viruses have genomes that encompass three long ORFs each, organized comparably to those of the planarian viruses, except that the genome length is somewhat shorter (6,618 to 6,969 nt) and that ORF1 (putative capsid protein [CP]) and ORF2 (RdRp) do not overlap (Fig. S2; identity scores for pairwise protein alignments in Fig. S1). Moreover, the same three viruses are those that were found to register top scores to the tricladiviruses in the BLASTX searches described above (see Table S1). Most of the other viruses in Fig. 2A have only two ORFs (Table S2), typical of GLV and other assigned Totivirdae family members. The genome lengths of all of the viruses in Fig. 2A range from 4,947 to 8,691 nt (Table S2), and the RdRp-encoding ORF in each virus is 3′-most in the genomic plus strand and usually overlaps the upstream ORF encoding the known or putative CP, also typical of GLV and other assigned Totivirdae family members.

FIG 2.

Unrooted radial phylograms. For each phylogram, aligned sequences were subjected to maximum likelihood phylogenetic analyses using ModelFinder, IQ-TREE, and UFBoot (57) as implemented with the “Find best and apply” option at https://www.hiv.lanl.gov/content/sequence/IQTREE/iqtree.html. Branch support values (from 1,000 bootstrap replicates) are shown as percentages; branches with support values of <70% have been collapsed to the preceding node. Scale bar indicates average number of substitutions per alignment position. (A) Deduced amino acid sequences of the translated ORF2 (RdRp) regions of each virus were aligned using MAFFT 7 (L-INS-i). The following were found to apply by ModelFinder: best-fit model according to BIC, LG+F+R5; model of rate heterogeneity, FreeRate with five categories; site proportion and rates, 0.0183 and 0.0138, 0.0662 and 0.1754, 0.2240 and 0.4485, 0.4203 and 0.9564, and 0.2712 and 1.7908, respectively. See Table S2 for summary of virus names, abbreviations, and RefSeq accession numbers. Three putative arthropod viruses branching nearest to the tricladiviruses are labeled in orange. The only formally assigned Totiviridae member with sufficient sequence similarity to appear in the tree, GLV, is labeled in cyan. Tentatively assigned totiviruses LbTV, PCMV1, and PCMV2, identified as being related to the tricladiviruses in BLASTX searches as described in the text, are labeled in green. Viruses that have been previously shown to form isometric viral capsids are marked with hexagons. (B) Nucleotide sequences of 11 newly assembled SmedTV strain sequences were aligned using MAFFT 7 (L-INS-i). The following was found to apply by ModelFinder: best-fit model according to BIC, HKY+I. The designated virus names reflect information from the respective BioProject metadata: the S. mediterranea strain name as annotated in either the metadata or an associated journal article, followed by an abbreviation for the reporting institution and the registration date(s) for the respective BioProject(s).

SmedTV in different strains of S. mediterranea and validation by amplicon sequencing.

A large number of transcriptome BioProjects with available Sequence Read Archive (SRA) data for S. mediterranea are available at the NCBI website (57 as of this writing), most of which do not have associated transcriptome assemblies deposited in the TSA database. We therefore surveyed a number of these SRA data sets in an effort to assemble SmedTV sequences from several well-annotated strains of S. mediterranea from different labs. Toward this end, we generated 11 complete coding sequences for SmedTV from selected BioProjects (registration dates 2011 to 2017) that have a sufficient number of sequence reads from SmedTV. All 11 of these assembled sequences for SmedTV could be aligned without gaps over a shared 7,858-nt region, including the expected three long ORFs described above and exhibiting >99.5% nucleotide sequence identity in pairwise comparisons. The numbers of nucleotide mismatches in the pairwise alignments ranged from 0 to 38 (Fig. S3). For example, SmedTV sequences assembled from Pearson lab BioProjects (University of Toronto), registered in 2012, 2015, and 2017 for asexual strain CIW4, showed no mismatches across this region (100% identity), suggesting durable genetic stability of the virus within a particular lab colony. By comparing all 11 SmedTV sequences, three main phylogenetic clades were identified, including two distinguishable clades of the tricladivirus from S. mediterranea annotated as CIW4 strains from different labs in the respective BioProject metadata (Fig. 2B and Fig. S3).

Among the 11 strains of S. mediterranea used for generating SmedTV sequences deposited in the NCBI database, all were annotated as asexual. By examining the SRA data sets from all BioProjects containing samples annotated as deriving from sexual strains of S. mediterranea (14 BioProjects as of this writing), on the other hand, we found none that contained large numbers of sequence reads from SmedTV: 10 of these BioProjects had no reads from SmedTV, whereas four contained only small numbers, 3 to 21. The transcriptome for sexual strain S2F2 of S. mediterranea at PlanMine (Smes) (24) was also found to register no hits to SmedTV. Based on these findings, we conclude that sexual strains of S. mediterranea in recent use by different labs are apparently not infected with SmedTV.

As a part of examining SmedTV sequences from different S. mediterranea strains, we also undertook to validate the SmedTV sequence by Sanger sequencing, starting with an RNA extract from S. mediterranea CIW4 (Pearson lab). We designed primers for reverse transcription-PCR (RT-PCR) based on the SmedTV sequences assembled from transcriptome data and then used the resulting amplicons for direct Sanger sequencing. The contig generated in this manner was 6,886 nt long and 100% identical to the transcriptome-based SmedTV-CIW4 (Pearson lab) sequences, providing validation for the transcriptome-based assemblies. However, we were unable to complete 5′ rapid amplification of cDNA ends (RACE) on the SmedTV genome to confirm/determine if the extreme terminal sequences at the 5′ end were present in the assembled transcriptomes.

Localization of SmedTV RNA within discrete cells of S. mediterranea.

The transcriptomes from which evidence for SmedTV was first obtained derive from serially passaged lab colonies of S. mediterranea. Because other organisms, such as bacteria and protists, are probably also present in such cultures (25), we tested whether SmedTV might be associated with one of them, rather than with S. mediterranea itself. We first performed PCR using SmedTV-specific primers on RNA-derived cDNA either from extensively washed worms of S. mediterranea asexual strain CIW4 or from passaged culture medium from which the worms had been excluded. An amplicon of expected size was produced only from the planarian sample, arguing against an externally contaminating source of SmedTV (Fig. S4). Notably, S. mediterranea sexual strain S2F2 also failed to yield an amplicon (Fig. S4), consistent with the paucity of sequence reads from SmedTV in the transcriptomes of sexual S. mediterranea strains described above.

Direct evidence for the presence of SmedTV in S. mediterranea CIW4 was obtained through double-fluorescence in situ hybridization (dFISH) and whole-mount in situ hybridization (WISH), using riboprobes for detecting SmedTV RNA in each case. In one set of experiments, dFISH was performed using dual riboprobes to detect both the plus and the minus strands of SmedTV, designed to nonoverlapping regions of the viral genome (Fig. 3A, top row). Within confocal sections through the worm (region of head shown in Fig. 3A), SmedTV RNA was detected in only a subset of discrete cells, and both strands consistently colocalized within these cells (101/101 SmedTV-staining cells that we counted). Interestingly, colocalization of the two strands of SmedTV RNA was most readily evident in cell nuclei, also colocalizing with the nuclear counterstain (Fig. 3A, middle row). In subsequent experiments, staining was performed to detect the SmedTV plus strand.

FIG 3.

Detection of SmedTV-staining cells in untreated worms. (A) dFISH confocal analysis of SmedTV-staining cells in the head region of worms. (Top) Separate riboprobes for detecting the SmedTV minus-strand RNA (antisense [AS]; magenta) and plus-strand RNA (sense [SE]; green) were used and showed consistent colocalization in the same cells, as expected. The area surrounded by a dotted box in each panel is enlarged in each inset to highlight this colocalization. (Middle) Nuclear staining with 4′,6-diamidino-2-phenylindole (DAPI; magenta) is included, coupled with staining for the SmedTV plus-strand RNA (green). The areas surrounded by dotted boxes 1, 2, and 3 are enlarged in the insets to highlight colocalization of the two stains. (Bottom) WISH analysis of SmedTV-staining cells in untreated worms, with staining of discrete cells but in differing numbers between worms from “low” (left) and “high” (right) cultures. Here, enrichment in eyespots is more evident in the low worm and enrichment in other neural structures (brain lobes) is more evident in the high worm (eyespots obscured). (B) dFISH confocal analysis of SmedTV-staining cells in brain lobes and eyespots. (Top [worm from low culture] and middle [worm from high culture]): SmedTV-staining cells (magenta) are visible in 20-μm Z-stacks of sectioned brain lobes (sections taken just posterior to the eyespots; brain lobes outlined with dotted lines), in some cases colocalizing in cells with the cholinergic neuron marker ChAT (green). (Bottom [worm from low culture]): SmedTV-staining cells (green) are also concentrated in eyespots, colocalizing in cells with the photoreceptor marker opsin (magenta). (C) dFISH confocal analysis of SmedTV-staining cells in stem cell-rich compartments in worm trunk regions. Staining for piwi-1 (magenta) identifies numerous stem cells in this field. Staining for SmedTV (green), in contrast, is limited to few cells, which generally do not appear to express piwi-1. The area surrounded by a dotted box is enlarged in the bottom row to show lack of colocalization between SmedTV and piwi-1, suggesting their respective presence in distinct cells. (D) Graphic illustration of previously reported single-cell RNA-seq data represented in violin plots for levels of sequence reads from SmedTV on the y axis and cell type on the x axis. SmedTV levels are higher in multiple, but not all, differentiated tissue types. Data in the lighter-shaded area are from Molinaro and Pearson (32) (head enriched), and data in the darker-shaded area are from Wurtzel et al. (31) (postpharyngeal). The generally higher levels of SmedTV seen in the Pearson lab data could be due to differences in tissue selectivity or culture conditions. (E and F) Analyses based on RNA-seq data from >50,000 single cells of S. mediterranea CIW4 (Reddien lab) in the planarian Digiworm atlas (33). An initial t-SNE expression profile plot (E) was regenerated for all cells in the atlas, and positions of 100 cells with the highest numbers of sequence reads from SmedTV are marked as red dots (top 25 cells) or cyan dots (next 75 cells). Outlines indicate previously identified clusters representing eight worm tissue classes (33): 1, neural; 2, protonephridia; 3, cathepsin⁺ cells; 4, epidermal; 5, pharynx; 6, intestine; 7, neoblast (stem cell); 8, parenchymal; and 9, muscle. A heat map (F) was regenerated for the 22 top SmedTV-expressing cells that clustered within the all-neural class of the atlas. Expression of selected marker genes (ciliated neuron [rootletin], pan-neural [pc-2], head sensory neurons [gpas], cholingerigic neurons [ChAT], photoreceptor [opsin], and stem cell [piwi-1]) are shown in rows. Cells (in columns) are ordered by SmedTV expression from highest to lowest.

The spatial distribution of SmedTV-staining cells within the whole body of S. mediterranea CIW4 was examined by WISH. We observed such cells to be scattered through the body of each worm but concentrated most clearly in the eyespots and brain lobes (Fig. 3A, bottom row). The distribution of SmedTV-staining cells was similar from worm to worm within the same culture container, but the numbers of such cells could vary between worms from different containers. In cultures with “high” numbers, there was a notable increase in staining of the head and its associated neural structures. This was quantified in worms from representative animals from both low and high cultures (Fig. S5A), revealing more than a 2-fold difference in the number of cells staining for SmedTV (P value < 0.0001) in the “high” worms (mean, 408 ± 20 cells per mm²; n = 8) versus the ‘”low” worms (mean, 180 ± 20 cells per mm²; n = 8). Moreover, this quantification was likely an underrepresentation, because such cells in the heads of “high” worms were so densely packed as to hinder accurate counting. The basis of these differing numbers of SmedTV-staining cells remains unknown but may indicate the presence of systemic host factors that regulate the levels of viral replication or expression and are subject to culture-dependent variations (26).

Colocalization of SmedTV RNA with neuronal markers in brains and eyes.

To obtain a more precise identification of the SmedTV-staining cells, we performed additional dFISH experiments. In one case, we used riboprobes for detecting the RNA of SmedTV as well as the cholinergic-neuron marker choline acetyltransferase (ChAT) (27) in tissue sections through worm heads (Fig. 3B, top two rows). ChAT expression in these images defines the brain lobe (dashed outlines) but can also be seen in the peripheral nervous system near the outer edges of the worm. Notably, cells staining for SmedTV RNA were found within both of these neuron-rich regions, in some cases colocalizing with ChAT. In addition, the numbers of SmedTV-staining cells again varied between worms from high and low cultures as defined above (Fig. S5B).

As suggested by WISH, cells staining for SmedTV RNA were also enriched in the eyespots of worms from both high and low cultures. To quantify this localization, we performed dFISH with riboprobes for detecting the RNA of either SmedTV or the photoreceptor neuron marker opsin (28) within confocal sections (Fig. 3B, bottom row). In this case, SmedTV-staining cells were clearly identified as photoreceptors via their presence in the eyespots and colocalization with opsin. Restricting quantification to worms from low cultures, we found similar levels of SmedTV-staining photoreceptors in both eyes (Fig. S5C, right, mean = 12.3 ± 1.7; left, mean = 11.6 ± 1.6; n = 7 for both eyes).

Of note, brain and eye neurons are differentiated cell types. Previously reported RNA-seq data sets from the SRA database were found in this study to contain many fewer sequence reads from SmedTV in samples from stem cell compartments. For example, one such data set (BioProject PRJNA167022) (29) included 0.7 read from SmedTV per million total reads (0.7 RPM) for the “X1” population of stem cells versus 42.7 RPM for differentiated tissues. Consistent with this finding, few SmedTV-staining cells were detected by dFISH among piwi-1-expressing stem cells (30) in confocal sections through stem cell-rich regions of worm trunks (Fig. 3C, green and magenta, respectively); moreover, the few SmedTV-staining cells that were seen in these regions appear not to express piwi-1 (Fig. 3C). Two single-cell RNA-seq data sets previously reported from different labs (BioProjects PRJNA276084 and PRJNA317292) (31, 32) supported the observation that fewer sequence reads from SmedTV were found in X1 stem cells and other neoblasts than in differentiated tissues, particularly neural cell types (Fig. 3D).

In a more recent study based on single-cell RNA-seq (BioProject PRJNA438083) (33), data were obtained from >50,000 S. mediterranea CIW4 cells to define the transcript expression profiles of different cell types. These profiles were, in turn, analyzed via t-distributed stochastic neighbor embedding (t-SNE) plots (34) to identify 44 major cell type clusters in nine tissue classes. By reexamining these data, we found that 22 of 25 cells with the highest levels of sequence reads from SmedTV (average log fold change [avg_logFC] values > 3.5) mapped to previously defined cell type clusters representing the neural class (33) (Fig. 3E). Expanding to examine the top 100 cells for SmedTV (avg_logFC values > 2.5), we found that 56 cells mapped to the neural class, with the other 44 cells distributed among seven other tissue classes: cathepsin⁺ (12 cells), protonephridia (8 cells), epidermis (6 cells), muscle (5 cells), parenchyma (5 cells), intestine (4 cells), and neoblasts (4 cells) (Fig. 3E). Among the 22 neural cells in the top 25 highest SmedTV-expressing cells, multiple neural subtypes were represented. This was indicated by differences in marker gene expression (log₂ [number of reads/transcript]) as illustrated using a heat map (Fig. 3F). Of note, half were ciliated (12/22 as marked by rootletin expression) and few were stem cells (5/22 were piwi-1 positive).

Interrogation of potential host regulators and distribution of SmedTV-staining cells during regeneration.

Given the low levels of SmedTV RNA in X1 stem cells as well as the few cells staining for SmedTV in stem cell compartments, we questioned whether SmedTV replication/expression might be repressed in these cells by the argonaute family member PIWI proteins, three of which are some of the mostly highly expressed genes in planarian stem cells (30, 35). Although no direct antiviral role for PIWI proteins has been reported specifically in planarians, they in conjunction with PIWI-interacting RNAs (piRNAs) are known to regulate gene expression, to silence transposable elements, and to suppress viral replication in many organisms (36). To address this question, we performed WISH to identify SmedTV-staining cells in GFP(RNAi) (i.e., control) worms versus worms that underwent triple knockdown of piwi-1, -2, and -3 (30, 35) by RNAi (efficiency of knockdowns described here and below confirmed by quantitative PCR [qPCR] [Fig. 4A, right]). We observed no discernible changes in the apparent numbers or distributions of SmedTV-staining cells in any of these piwi-1,2,3(RNAi) worms by WISH or when quantified by qPCR (Fig. 4A, top). Reciprocally, similar results were obtained upon ablation of stem cells by ionizing irradiation (Fig. 4A, bottom right).

FIG 4.

Potential regulators of SmedTV expression. WISH and qPCR analysis of SmedTV expression following different experimental conditions. (A) Knockdown of stem cells by combining piwi-1, -2, and 3 RNAi (top) or by lethal irradiation (bottom) had no effect on SmedTV expression. qPCR confirmation of knockdown of piwi-1, -2, and -3 and quantification of SmedTV expression are shown on the right. Normalized gene expression was calculated for worms following GFP(RNAi) or piwi-1,2,3(RNAi) knockdown at 4 days after the third feeding (3fd4) (as in panel A). Note that despite robust knockdown of piwi-1,2,3, no difference was observed in SmedTV expression levels. (B) RNAi knockdown of dsRNA helicase enzyme-encoding gene, RIG-I, does not affect expression. qPCR confirmation of knockdown of RIG-1 is shown on the right. Normalized gene expression of RIG-I and SmedTV following GFP(RNAi) or RIG-I(RNAi) knockdown was analyzed at 5fd11 (as in Fig. S7B, left and middle). Note that despite robust knockdown of RIG-1, no difference was observed in SmedTV expression levels. (C) Confocal image of combined FISH for SmedTV expression (green) and TUNEL (magenta) staining. An example of a doubly positive cell is shown on the left. An additional example illustrating that this is a rare occurrence (2/51 TUNEL-positive cells) shown on the right. Note that TUNEL-positive cells do not to double with SmedTV.

We additionally pursued RNAi knockdowns of one other candidate regulator, retinoic acid-inducible gene sRNA, RIG-I. RIG-I encodes a dsRNA helicase essential for recognition and control of many RNA viruses (37). RIG-I plays an essential role in virus-induced innate immunity of the closely related free-living planarian flatworm Dugesia japonica (38, 39). Again, however, we observed no discernible changes in the apparent numbers or distributions of SmedTV-staining cells in the RIG-I(RNAi) worms versus control(RNAi) worms examined by WISH or when quantified by qPCR (Fig. 4B).

In another effort, we tested whether SmedTV replication/expression might be affected in dying cells (e.g., derepressed). To address this possibility, we combined FISH and terminal deoxynucleotidyltransferase-mediated dUTP-biotin nick end labeling (TUNEL) staining (40) to determine whether cells undergoing apoptosis might correspond to those that stain for SmedTV (Fig. 4C). Among 51 TUNEL-positive (apoptotic) cells that we counted in confocal sections in this experiment, only 2 were also positive for SmedTV; moreover, many more cells that stained for SmedTV were seen to be TUNEL negative. We conclude that apoptosis does not routinely derepress SmedTV replication/expression and also that apoptosis is not required to reach levels of SmedTV RNA sufficient to be seen by FISH.

We next examined SmedTV-staining cells by WISH during regeneration. Heads and tails were amputated from worms, and the remaining trunk fragments were fixed at various time points (1, 3, 7, 9, and 14 days) postamputation (pa) (Fig. 5A). Interestingly, WISH demonstrated very few, if any, cells staining for SmedTV in either the head or the tail blastemas (newly regenerated tissue) until 7 dpa (Fig. 5A, second from right). By 9 dpa, strongly SmedTV-staining cells were more generally found in the regenerated head tissue, with further increases evident by 14 dpa, including in the eyespots (Fig. 5A, far right). In total, SmedTV is not strongly detected in newly regenerated tissues until regeneration is largely complete.

FIG 5.

Detection of SmedTV-staining cells in regenerating, dsRNA-fed, and homogenate-injected worms. (A) WISH analysis of SmedTV-staining cells in regenerating trunk fragments. Dotted line represents plane of amputation in each panel with tissue above (head panels) or below (tail panels) being newly regenerated. SmedTV-staining cells are largely absent from the regenerating tissues until 9 dpa, with robust expression evident by 14 dpa; arrowheads highlight weakly positive cells in the regenerating tissues at 7 dpa. (B) WISH analysis of SmedTV-staining cells in worms following a GFP(RNAi) or SmedTV(RNAi) feeding-plus-regeneration regimen (see the text). The worms shown here were analyzed at 90 days after feeding 10 (10fd90). Very few, if any, SmedTV-staining cells are visible in the SmedTV(RNAi) worm. (C) Image of an ethidium bromide (EtBr) agarose gel under UV light (colors inverted for clarity) following electrophoresis to visualize amplicons (or lack thereof) resulting from 33 cycles of PCR using SmedTV-specific (expected amplicon size of ∼1.6 kb, indicated by a solid black arrowhead) and piwi-3-specific (expected amplicon size of ∼1.8 kb, indicated by an open arrowhead) primers on cDNA generated from RNA isolated from control and SmedTV RNAi-treated worms. Lanes 1 and 15 contain a 1-kb DNA ladder. Lanes 2, 5, 8, 11, and 14 were run empty for spacing of experimental samples. Lane pairs (3/4, 6/7, 9/10, and 12/13) represent wild-type control, 7-RNAi-feed, 10-RNAi-feed, and 15-RNAi-feed SmedTV RNAi samples, respectively. The first lane of each pair contains the SmedTV reaction, and the piwi-3 positive control is in the second. Note that relative to the control (lane 3), the SmedTV band reduces until no longer detectable (lanes 6, 9, and 12), while the piwi-3 amplicon remains unchanged across all samples (lanes 4, 7, 10, and 13). (D) WISH analysis of SmedTV in S. polychroa following no or 10 injections of homogenates from S. mediterranea strain CIW4 (Pearson lab). No SmedTV-staining cells are visible in either worm.

SmedTV is susceptible to RNAi and failed to be horizontally transmitted.

Toward examining whether SmedTV may have any functional effects in S. mediterranea, we knocked down SmedTV by RNAi (41). Initial feedings with SmedTV-specific dsRNA resulted in a progressive decrease in the apparent numbers of SmedTV-staining cells detected by WISH and quantified by qPCR, though these numbers remained fairly high even after seven feedings (Fig. S6). By including regeneration steps to force tissue turnover as part of the RNAi protocol, however, numbers of SmedTV-staining cells were greatly reduced, such that after 10 feedings of SmedTV(RNAi) and two rounds of regeneration from trunk fragments (after 3 and 7 feedings), SmedTV-staining cells were no longer detected by WISH and remained undetectable until the end of the experiment at 90 days after feeding 10 (Fig. 5B). Worms treated in parallel with GFP(RNAi), in contrast, showed no discernible changes in their numbers and distributions of SmedTV-staining cells. Similarly, additional rounds of RNAi resulted in further reduction of the SmedTV transcripts relative to that of controls as evident through amplicon intensity following PCR (Fig. 5C). The SmedTV band was no longer detectable after 15 feedings of SmedTV(RNAi), while intensity of a control gene (piwi-3) amplicon remained unchanged. Despite an apparent cure of these worms of SmedTV, no morphological, behavioral, or health differences were observed.

Because the CIW4 strain of S. mediterranea is asexual and clonally maintained in culture, it is plausible that SmedTV may be transmitted only vertically from manually divided “parents” to regenerated progeny. Alternatively, SmedTV might be transmitted horizontally between worms in the same culture as an active viral infection. To begin to test these possibilities, we injected a sonicated homogenate from a high SmedTV-staining CIW4 culture into the closely related, but uninfected sexual species, Schmidtea polychroa (see Fig. S4 for PCR evidence for the absence of SmedTV in S. polychroa; in addition, two transcriptome assemblies from S. polychroa, available in the NCBI TSA database, were found to be negative for SmedTV-matching hits). Ten such injections of 10 separate worms failed to yield detectable transmission of SmedTV as determined by WISH analysis for SmedTV-staining cells in the injected worms (Fig. 5D). Despite several caveats, including a possible host barrier in S. polychroa that is not present in S. mediterranea, these results suggest to us that SmedTV may lack the capacity for horizontal transmission and that transmission may instead occur only vertically, from mother to progeny worms, and probably only by intracellular means from stem cells to their progeny cells.

DISCUSSION

Phylogeny and taxonomy.

The five newly discovered planarian viruses described here appear to represent a distinct taxon, based on their distinctive host range and evidence that they constitute a discrete phylogenetic clade. At this point in understanding these viruses, we therefore suggest that they are appropriate to classify within a new genus, for which we propose the name “Tricladivirus” to reflect the taxonomic order, Tricladida, to which their planarian hosts belong. The tricladiviruses show more distant, yet evident, phylogenetic relationships to Totiviridae family member GLV and other toti-like viruses, such as LbTV, PCMV1, and PCMV2, that have been tentatively assigned to family Totiviridae (see Fig. 2A), and we therefore find it sensible at this point to identify the tricladiviruses also as tentative members of family Totiviridae.

Proteins.

Totiviruses and most toti-like viruses have two long ORFs in their genomic plus strands (e.g., see Table S2), encoding the known or putative viral CP and the viral RdRp in their upstream and downstream ORFs. Moreover, in most of these viruses, and afforded by the fact that the 3′ end of the CP ORF overlaps the 5′ end of the RdRp ORF, the RdRp is likely translated as part of a CP/RdRp fusion protein following ribosomal frameshifting in the ORF1-ORF2 overlap region. The tricladiviruses, like some other toti-like viruses, instead have three long ORFs in their genomic plus strands, but including two ORFs that overlap and have the same coding capacities for an apparent CP (P1) and RdRp (P2) as just described (Fig. 1).

As noted above, the apparent tricladivirus CP shows significant sequence similarity to the CP of tentatively assigned Totiviridae members PCMV1 and PCMV2. Considering that PCMV1 forms nonenveloped virus particles with isometric capsids (18), we expect the tricladiviruses to do the same. Nonetheless, our efforts to visualize tricladivirus particles by transmission electron microscopy (TEM) have proven unsuccessful to date, and we acknowledge the possibility that the tricladiviruses might not form isometric capsids. Among the related viruses shown in Fig. 2A, nonenveloped virus particles with isometric capsids have also been shown for Camponotus yamaokai virus (CyV), GLV, and LbTV (20, 22, 23, 42), and none of the other viruses in Fig. 2A have been suggested to lack such capsids.

The additional ORF of the tricladiviruses, encoding smaller protein P0, is located 5′-most in the genomic plus strand. Because P0 lacks significant sequence similarities to any other proteins in the Non-Redundant protein sequence database at GenBank, the function of P0 remains unknown. As noted above, P0 is also the most variable of the three proteins among the tricladiviruses. As only CP and RdRp are essential for totivirus replication (most viruses in Fig. 2A have ORFs for only these two proteins), some auxiliary role for P0 seems possible or even likely. We speculate that it might be involved in evading one or more immune defense mechanisms of host planarians, though other possibilities remain. The additional protein of PCMV1, for example, includes a putative chemokine superfamily domain and has been speculated to play a role in modulating inflammatory responses in infected fish or in binding to chemokine receptors on the surface of fish cells during viral entry into those cells (18).

Tissue distribution and proposed transmission mechanism.

SmedTV was localized to discrete cells in the worms by numerous WISH and FISH analyses. These cells appear to be concentrated in neural structures but are scattered through other worm tissues as well (Fig. 3). In contrast, few SmedTV-staining cells were detected in stem cell populations or in early blastema tissue. The basis for this apparent tissue/cell selectivity remains unclear. Observations about the tissue distribution of SmedTV nonetheless seem to have implications for how the virus is transmitted between worms, especially given other evidence that SmedTV may not be effectively transmitted by horizontal means (i.e., no transmission by injections of tissue homogenates in this study). We think it important to recognize that the WISH and FISH assays in this study probably lack the sensitivity to detect cells in which SmedTV RNA is present but at lower levels because viral replication/expression has been somehow repressed (e.g., in stem cells). The cells in which SmedTV RNA is detected by WISH and FISH are, then, ones in which viral replication/expression has been derepressed (e.g., in photoreceptors). In this light, we propose that SmedTV is transmitted to progeny worms by most or all cells in each manually divided fragment of the mother, including stem cells and others in which SmedTV replication/expression has been repressed to lower levels (Fig. 6). As tissue regeneration proceeds in the progeny worms, SmedTV replication/expression is then derepressed along certain differentiating cell lineages, including some photoreceptors and other neurons (Fig. 6) but cells in other tissues as well (Fig. 3). Identifying host and viral factors involved in these regulatory effects is therefore a matter of great interest for future studies.

FIG 6.

Diagram of proposed SmedTV infection cycle in S. mediterranea asexual strains. The outer circle of images shows manual fission of a mother worm (left) into head (top) and tail (bottom) fragments, followed by regeneration and morphallaxis (remodeling) and then growth to yield two full-sized progeny worms (right). Circled cutaways reveal the gut branches (violet) and stem cell-rich compartments (green) in the worm at left and the optic cup (magenta) and brain lobe (cyan) in the worm at right. Across the middle of the diagram, SmedTV is represented by red hexagons and is illustrated to be present at low levels in stem cells and their immediate progeny but at higher levels in a subset of differentiated cells (shown here as higher in a photoreceptor and another neuron but not in an epithelial cell). As additionally explained in the text, we propose that SmedTV is present in most or all stem cells and their immediate progeny, but its replication/expression is repressed to lower levels until derepression occurs along certain differentiation lineages as worms undergo regeneration, morphallaxis, and growth.

Asexual versus sexual worms.

In this report, we provide evidence that asexual and sexual strains of S. mediterranea are, respectively, positive and negative for the virus. These findings simplistically suggest that SmedTV might not be able to survive meiosis, gametogenesis, fertilization, and/or embryogenesis to undergo vertical transmission during sexual reproduction. However, the origins of these respective strains should also be considered. The asexual strains now in common use by different labs are generally all thought to descend from a single cloned animal (CIW4) obtained from a fissiparous population of S. mediterranea native to Spain (43); the sexual strains, in contrast, have descended from multiple inbred lines derived from a hermaphroditic populations of S. mediterranea native to the Mediterranean from Spain to Sardinia (44, 45). These two subspecies of S. mediterranea are therefore both geographically and reproductively distinct and likely diverged long ago. The difference in their derivative strains with regard to SmedTV can thus not be attributed with confidence to their difference in reproduction alone. In addition, three of the other planarian species in which other tricladiviruses are reported here (B. candida, D. lacteum, and P. torva) are thought to reproduce at least mostly by sexual means. Considering these findings, we suspect that sexual reproduction is not a strict barrier for vertical transmission of tricladiviruses. Related viruses in Fig. 2A for which evidence of vertical transmission during sexual reproduction has been obtained are CyV and LbTV (20, 42).

Nuclear localization.

Totiviruses are generally thought to replicate in the cytoplasm (46). GLV, however, has been shown to accumulate both viral particles and viral RNA also inside the nuclei of its Giardia (protozoan) host cells (47–49). Evidence in this report for the presence of both strands of SmedTV RNA in the nuclei of planarian cells is thus consistent with these findings for GLV and also consistent with the fact that GLV and the tricladiviruses share a relatively close phylogenetic relationship (Fig. 2A). Using cNLS Mapper as implemented at http://nls-mapper.iab.keio.ac.jp/cgi-bin/NLS_Mapper_form.cgi (50), we indeed found that the apparent SmedTV CP contains a strongly predicted nuclear localization signal, RSRAKRRATSS (score, 8/10), though this sequence is not conserved in the other tricladiviruses described here.

Effects on host biology and evasion of host defenses.

Using public transcriptome data, we found evidence for SmedTV persistently infecting S. mediterranea asexual strains from a number of different labs in Canada, Germany, Spain, the United Kingdom, and the United States. Yet there is no evidence for illness in the S. mediterranea CIW4 colony used in this study (maintained at University of Toronto) and presumably not in the other colonies either. This suggests a mutualistic or commensalistic relationship between this virus and host. After apparently curing worms through serial feedings with SmedTV(RNAi), we observed no obvious effects on worm health or behavior, consistent with a commensalistic relationship; however, numerous conceivable mutualistic effects remain to be investigated.

Planarians have endogenous mechanisms to suppress exogenous elements such as viruses, including the RIG-I antiviral program (37–39) and RNAi silencing systems, which can involve piRNAs or small interfering RNAs (siRNAs) (30, 35, 36). In this study, we tested whether RIG-I or the piwi genes might be responsible for repressing SmedTV replication/expression, but we did not obtain evidence to support this hypothesis. As noted above, though, we found that SmedTV is susceptible to RNAi, which presumably acts by giving rise to SmedTV-specific siRNAs. This finding raises an interesting follow-up question, namely, how does SmedTV evade the RNAi pathway under natural circumstances, in the absence of dsRNA feeding, to maintain its persistent infection of worms? The replication strategy shared by totiviruses and many other dsRNA viruses, whereby viral dsRNA is thought to be generated only inside the viral capsid(s) and maintained there throughout the replication cycle (46), may play a role in sequestering SmedTV dsRNA from RNAi signaling factors, allowing the virus to persist. It also seems conceivable to us that the apparent nuclear localization of SmedTV might represent some sort of immune evasion strategy that this virus employs.

MATERIALS AND METHODS

Transcriptome analyses.

Queries used in the initial TBLASTN searches were the RdRp sequences of infectious myonecrosis virus (IMNV; GenBank accession no. ABN05325.1), LbTV (AGW80479.1), and PCMV1 (ADP37187.1). Using the LbTV query, eight hits with E values better than 1.3e−03 were obtained, six of which were ≥7,808 nt in length. The translated sequences of these six longest hits were then used to repeat the TBLASTN search, which identified the same six high-scoring hits ≥7,808 nt in length. These original six TSA accession numbers are GFKB01052472.1 (BcanTV), GEKK01023799.1 (PmorTV), GAKN01017247.1 (SmedTV), GBGQ01018256.1 (SmedTV), GCZZ01061652.1 (SmedTV), and GFQI01052041.1 (SmedTV). The immediately preceding search was repeated on the PlanMine database, which yielded six high-scoring hits of ≥7,887 nt in length. These original six PlanMine accession numbers are dd_Dlac_v9_194467_0_1 (DlacTV), dd_Ptor_v3_38990_1_1 (PtorTV), dd_Smed_v6_19281_0_1 (SmedTV), ka_Smed_v1_GCZZ01061652.1 (SmedTV), ox_Smed_v2_08543 (SmedTV), and to_Smed_v2_BPKG17 (SmedTV). The SmedTV accessions from the preceding searches were then compared to generate a consensus, which was used in subsequent analyses; the termini of this consensus were trimmed to sequences shared by at least four of the SmedTV accessions.

For comparing SmedTV sequences from different labs, we accessed sequence reads in the SRA database for generating our own assemblies for potentially distinguishable strains of SmedTV. BioProjects that we accessed to obtain SRA data for assembling the 11 strain sequences shown in Fig. 2B are PRJNA79031, PRJNA167022, PRJNA283132, PRJNA387022 plus PRJNA387024 plus PRJNA320389, PRJNA415947, PRJNA215411, PRJNA235907, PRJNA303146 plus PRJNA304644, PRJNA327836, PRJEB2469 plus PRJNA208441 plus PRJNA222859, and PRJNA376157. To assemble each respective sequence, we first used MegaBLAST to search the pertinent SRA data set(s) for sequence reads that approximated the SmedTV consensus sequence described in the preceding paragraph. The reads registering as hits were then assembled into a consensus contig for each strain using CAP3 (51) as implemented at https://galaxy.pasteur.fr/ or http://biosrv.cab.unina.it/webcap3/, CLC Genomics Workbench 7 (Qiagen), and/or Velvet as implemented at https://galaxy.pasteur.fr/. Both termini of the assembled contig were then extended to their final lengths by using Discontiguous MegaBLAST or BLASTN to perform progressive searches against the pertinent SRA data sets, with reads registering as hits appended progressively to the ends of the contig. Once all of the SmedTV strain sequences had been assembled, they were aligned, and terminal sequences either unique to one strain or contradictory between strains were removed, yielding the 11 final assemblies, which ranged in length from 7,888 to 7,959 nt.

Planarian culture.

Clonal populations of S. mediterranea asexual strain CIW4, S. mediterranea sexual strain S2F2, and S. polychroa were maintained under standard conditions as previously described (6). For irradiation experiments, worms were exposed to 60 Gy of gamma irradiation from a ¹³⁷Cs source, Gammacell 40 extractor irradiator (Best Theratronics).

WISH, dFISH, and FISH/TUNEL.

WISH and dFISH were performed as previously described (52–54). Riboprobes for SmedTV were prepared from two pPR-T4P plasmids into which a region of sequence from either end of the SmedTV genome had been cloned (nt positions 510 to 2094 [region 1] or nt positions 6318 to 7874 [region 2]; position numbers based on the SmedTV plus-strand consensus sequence represented in Fig. 1). These pPR-T4P plasmids were then used for PCR amplifications in which either a T7 or an SP6 promoter was introduced onto the 5′ end of the desired region and strand of SmedTV. To generate the SmedTV riboprobes, these amplicons were then lastly used for in vitro transcription with either T7 or SP6 RNA polymerase. For the colocalization analysis of SmedTV RNA in Fig. 3A, the respective riboprobes were designed to hybridize either to a 3′-proximal region of the genomic plus strand (minus-strand probe transcribed from cloned region 2) or to a 3′-proximal region of the genomic minus strand (plus-strand probe transcribed from cloned region 1). For the other experiments, only the first of these two riboprobes was used. For WISH, riboprobes were transcribed in the presence of nucleotides labeled with digoxigenin (DIG) and detected after hybridization via a chromogenic reaction (anti-DIG antibody conjugated with alkaline phosphatase, plus nitroblue tetrazolium and 5-bromo-4-chloro-3′-indolyphosphate substrates). For dFISH, respective riboprobes were transcribed in the presence of nucleotides labeled with either DIG or fluorescein isothiocyanate (FITC). In both cases, the probes were detected or their signal was amplified after hybridization via a fluorogenic reaction (anti-DIG antibody conjugated with alkaline phosphatase plus Fast Blue substrate; anti-FITC antibody conjugated with horseradish peroxidase plus tyramide substrate). Whole-mount FISH/TUNEL was performed as previously described (40).

Microscopy and image acquisition, processing, and analysis.

Colorimetric WISH stains were imaged on a Leica M165 FC fluorescent dissecting microscope with a Leica DFC7000 T digital camera. Photographs of whole animals were obtained with an Olympus SZX16 microscope equipped with a DP72 digital camera. dFISH results (whole animal and sections) were photographed with a Quorum Spinning Disk Confocal 2 (Olympus IX81 microscope and Hamamatsu C9100-13 electron multiplying charge-coupled-device [EM-CCD] camera). Raw images were captured using Perkin Elmer Volocity (confocal) software and stitched together for whole-animal images. Images were postprocessed in Adobe Photoshop, and figures were assembled in Adobe Illustrator. Linear adjustments (brightness and contrast) were made for images of animals labeled by WISH, dFISH, or FISH/TUNEL in order to best represent actual results. These adjustments were identical within a given experiment in which comparisons were drawn between conditions. Cell counts (cells positive for SmedTV plus-strand RNA, alone or with SmedTV minus-strand RNA, ChAT, opsin, piwi-1, or TUNEL) were quantified using ImageJ software (http://rsb.info.nih.gov/ij/). Graphs and statistics were generated using GraphPad Prism software. Significance was determined by a 2-tailed Student t test with equal or unequal variance as specified. To eliminate any bias due to difference in detection threshold with different development techniques, only cells that were completely encompassed within Z-projections were counted for the SmedTV plus- and minus-strand RNA colocalization analysis.

t-SNE plot and heat map.

Plots were produced using transcriptome data from >50,000 single S. mediterranea strain CIW4 cells from the planarian Digiworm atlas (33) (Gene Expression Omnibus accession number GSE111764). To find single cells with the highest numbers of reads from SmedTV (PlanMine accession number dd_Smed_v4_19281_0_1), a combination of functions (FetchData and WhichCells) from R package Seurat version 3.1 (34) were used to extract the barcodes of 100 cells with avg_logFC values for SmedTV of >2.5, including 25 cells with values of >3.5. The positions of these cells were identified in a t-SNE plot for all cells in the original data set, using the same principal components and cluster resolutions as in the original report (33). For the top 22 SmedTV cells mapping to the neural cluster, the number of reads per transcript for selected marker genes was also used to generate a heat map using the pheatmap R package. The log₂ expression of the selected genes (in rows) is shown in Fig. 3 within ordered SmedTV-expressing cells (columns).

RNAi.

dsRNA-expressing Escherichia coli HT115(DE3) cultures were prepared by transformation with respective pPR-T4P plasmids, mixed with homogenized calf liver, and fed to animals as previously described (6). The pPR-T4P plasmid containing the region 2 insert was used for the SmedTV(RNAi) experiments. Unless otherwise stated, animals were fed every 3 days and rinsed each day between feedings. The numbers of feeds (f) and days (d) after which the phenotypes were analyzed varied for each treatment and are listed in the figures with an annotation of XfdY, where X is the number of feeds and Y is the numbers of days after the last feed. For the piwi(RNAi) experiments, equal amounts of dsRNA-expressing bacterial cultures for piwi-1, -2, and -3 were combined for feedings (55). For the SmedTV(RNAi) experiments, worms were subjected to two rounds of head and tail amputation to force tissue turnover after three and seven feeds. Trunk fragments were given 1 week to regenerate before recommencement of RNAi feedings. In all experiments, an RNAi vector with green fluorescent protein (GFP)-coding sequences was used as a negative control. Effective knockdowns were confirmed by quantitative real-time PCR as follows. Reverse transcription reactions were conducted on total RNA extracted from approximately 10 worms using an Invitrogen SuperScript III reverse transcriptase kit. Quantitative real-time PCR was then performed in biological triplicate on a Bio-Rad CFX96 Touch real-time PCR detection system with Roche SYBR green PCR master mix per the manufacturer’s instructions. Expression was normalized to GFP(RNAi) worms and the threshold cycle (2^−ΔΔCT) method was used for relative quantification. Glyceraldehyde-3-phosphate dehydrogenase (GAPDH) expression was used as a reference as previously described (56).

Injections of worm homogenates.

Homogenate was generated by sonicating pooled S. mediterranea CIW4 worms in physiological salt (150 mM NaCl, 10 mM MgCl₂, 10 mM Tris [pH 7.5]) with four pulses at power level 3 on a Fisher Scientific model 100 sonic dismembrator. The homogenate was cooled on ice for 10 s between pulses and gently centrifuged to pellet any large debris. The resulting supernatant was injected into the mesenchyme of S. polychroa using a Drummond Scientific Nanoject mounted on a micromanipulator. Worms were immobilized using a cold plate and injected with 2 to 6 pulses of homogenate (32 nl each) until gut branches were filled. Worms were injected 10 times total, with 2 to 3 days between injections.