Viral Discovery in the Invasive Australian Cane Toad (Rhinella marina) Using Metatranscriptomic and Genomic Approaches

Cane toads are poisonous amphibians that were introduced to Australia in 1935 for insect control. Since then, their population has increased dramatically, and they now threaten many native Australian species. One potential method to control the population is to release a cane toad virus with high mortality rates, yet few cane toad viruses have been characterized. This study samples cane toads from different Australian locations and uses an RNA sequencing and computational approach to find new viruses. We report novel complete picornavirus and retrovirus sequences that were genetically similar to viruses infecting frogs, reptiles, and fish. Using data generated in other studies, we show that these viral sequences are present in cane toads from distinct Australian locations. Three sequences related to circoviruses were also found in the toad genome. The identification of new viral sequences will aid future studies that investigate their prevalence and potential as agents for biocontrol.

ABSTRACT

Cane toads are a notorious invasive species, inhabiting over 1.2 million km² of Australia and threatening native biodiversity. The release of pathogenic cane toad viruses is one possible biocontrol strategy yet is currently hindered by the poorly described cane toad virome. Metatranscriptomic analysis of 16 cane toad livers revealed the presence of a novel and full-length picornavirus, Rhimavirus A (RhiV-A), a member of a reptile- and amphibian-specific cluster of the Picornaviridae basal to the Kobuvirus-like group. In the combined liver transcriptome, we also identified a complete genome sequence of a distinct epsilonretrovirus, Rhinella marina endogenous retrovirus (RMERV). The recently sequenced cane toad genome contains 8 complete RMERV proviruses as well as 21 additional truncated insertions. The oldest full-length RMERV provirus was estimated to have inserted 1.9 million years ago (MYA). To screen for these viral sequences in additional toads, we analyzed publicly available transcriptomes from six diverse Australian locations. RhiV-A transcripts were identified in toads sampled from three locations across 1,000 km of Australia, stretching to the current Western Australia (WA) invasion front, while RMERV transcripts were observed at all six sites. Finally, we scanned the cane toad genome for nonretroviral endogenous viral elements, finding three sequences related to small DNA viruses in the family Circoviridae. This shows ancestral circoviral infection with subsequent genomic integration. The identification of these current and past viral infections enriches our knowledge of the cane toad virome, an understanding of which will facilitate future work on infection and disease in this important invasive species.

IMPORTANCE Cane toads are poisonous amphibians that were introduced to Australia in 1935 for insect control. Since then, their population has increased dramatically, and they now threaten many native Australian species. One potential method to control the population is to release a cane toad virus with high mortality rates, yet few cane toad viruses have been characterized. This study samples cane toads from different Australian locations and uses an RNA sequencing and computational approach to find new viruses. We report novel complete picornavirus and retrovirus sequences that were genetically similar to viruses infecting frogs, reptiles, and fish. Using data generated in other studies, we show that these viral sequences are present in cane toads from distinct Australian locations. Three sequences related to circoviruses were also found in the toad genome. The identification of new viral sequences will aid future studies that investigate their prevalence and potential as agents for biocontrol.

INTRODUCTION

The cane toad (Rhinella marina) is a large, poisonous anuran native to Central and South America (1). It has been deliberately introduced for insect control to more than 20 countries during the last 2 centuries, including Australia, those on Caribbean and Pacific islands, the United States, Japan, Taiwan, the Philippines, and Papua New Guinea (2). Cane toad populations thrive in almost all of these regions, making them one of the most successful colonizers ever documented (3).

In 1935, 101 toads of Hawaiian origin were introduced to northeast Queensland (QLD), Australia, to control the cane beetle (2). The toads were ineffective as biocontrol agents but flourished in the tropical climate (1). As of 2007, the species is estimated to inhabit over 1.2 million km² of the Australian mainland and continues to rapidly colonize new territory (4, 5). Toads exert a negative impact on native Australian species and have caused population declines in lizards (6–9), quolls (10), snakes (11, 12), crocodiles (8, 13), and invertebrates (6, 14, 15). Accordingly, many cane toad control strategies have been trialed, yet none have managed to reduce the toad population or prevent further spread (16).

Viral biocontrol is a potential solution to the cane toad problem. Its success in mitigating the impact of invasive species was demonstrated by the use of both myxoma virus and rabbit hemorrhagic disease virus to control the European rabbit in Australia and New Zealand (17, 18). Similarly, the release of feline panleukopenia virus eliminated the feral cat population on Marion Island, South Africa (19). Key to the success of these viruses as biocontrol agents were (i) the specificity for the target species, (ii) high transmission rates, (iii) high mortality rates, and (iv) immunologically naive target populations (20). These characteristics would thus be essential for the success of a cane toad biocontrol virus. However, DNA viruses from only a single genus, Ranavirus (family Iridoviridae), have been investigated for their potential as biocontrol agents. While in vivo studies showed that these viruses were highly pathogenic and transmissible, a lack of host specificity prevented implementation (16, 20–22). Additional attempts to isolate other viral families from Australian toads have also been unsuccessful (16). The lack of a suitable biocontrol virus means that there is a compelling need to discover new cane toad-specific viruses.

Bulk RNA sequencing (RNA-Seq)—metatranscriptomics—has revolutionized the field of viral discovery (23). This is because sequencing the total RNA of a cell or sample permits the identification of known and distantly related viral transcripts without any prior knowledge of their specific sequence (23). This method has been used to identify approximately 1,500 highly divergent RNA viruses in invertebrates (24, 25) and, more recently, 218 novel vertebrate viruses (26), greatly expanding the diversity of known viral families. Metatranscriptomics is therefore appropriate for viral discovery in species whose viromes are underrepresented in sequence databases.

As a complementary approach to metatranscriptomics, bioinformatic analysis of a host genome can identify endogenous viral elements (EVEs), virus-like sequences in host genomes resulting from ancestral infections (27). Although EVEs are present in many organisms (28), they have not yet been described in cane toads. In addition to facilitating gene and evolutionary studies, EVEs identified in the cane toad genome could expand the known repertoire of viruses capable of infecting the species.

This study aimed to identify novel viral sequences among Australian cane toads, including active infections and EVEs. To this end, metatranscriptomics was first performed on total liver RNA from toads collected in QLD and Western Australia (WA), followed by screening additional transcriptome data sets from a further six locations to assess the geographical prevalence of the two novel viruses detected. Furthermore, we analyzed the recently reported cane toad genome (v2.2) (R. J. Edwards, D. Enosi Tuipulotu, T. G. Amos, D. O'Meally, M. F. Richardson, T. L. Russell, M. Vallinoto, M. Carneiro, N. Ferrand, M. R. Wilkins, F. Sequeira, L. A. Rollins, E. C. Holmes, R. Shine, and P. A. White, submitted for publication) for both retroviral and nonretroviral EVEs. Our results reveal novel cane toad viruses from the Picornaviridae, Retroviridae, and Circoviridae, expanding the virome of this important invasive species beyond the Iridoviridae.

RESULTS

Detection of a novel picornavirus (Rhimavirus A) in cane toad liver tissue.

To discover novel cane toad viruses, we performed metatranscriptomic analysis of a single library prepared from RNA of 16 cane toad livers. For this, 27,446,915 paired reads were generated and de novo assembled before screening for putative novel viral sequences with blastx. We identified an abundant transcript (7,444 nucleotides [nt]) (3.11 transcripts per million [TPM]) in our pooled transcriptome best matching Tortoise rafivirus A (ToRaV-A) (candidate genus Rafivirus, family Picornaviridae) (GenBank accession number YP_009026385), which was identified in liver, spleen, and kidney tissues of diseased tortoises (29). The assembly of raw reads onto the picornavirus-like transcript in our data set revealed complete coverage of the viral genome (mean read depth of 29.5 reads per nucleotide position). The novel putative picornavirus was denoted Rhimavirus A (Rhinella marina virus A) (RhiV-A). Based on the RhiV-A genome sequence, reverse transcriptase PCR (RT-PCR) amplification of a 302-nt region of the 5′ untranslated region (UTR) demonstrated that RhiV-A was present in 6 of 16 RNA samples (data not shown).

Classification of RhiV-A.

To classify RhiV-A, a maximum likelihood (ML) phylogenetic tree was inferred with representative viruses (n = 154) from known Picornaviridae genera, candidate genera, and newly identified picornaviruses, using a 306-amino-acid (aa) region of 3D^pol (see Fig. S1 in the supplemental material). A smaller tree was inferred using only viruses of the Kobuvirus supergroup to which RhiV-A belongs (n = 19). RhiV-A falls in a monophyletic cluster with two reptilian picornaviruses, Tortoise rafivirus A (ToRaV-A) and Hainan Gekko similignum picornavirus (HGSP), in a lineage basal to defined bird- and mammal-specific genera (Sicinivirus, Gallivirus, Kobuvirus, Sakobuvirus, and Salivirus) (Fig. 1A).

FIG 1.

Classification and genome of a novel picornavirus, RhiV-A, found in the metatranscriptome of 16 pooled cane toad livers. (A) Phylogeny of RhiV-A 3D^pol (306 aa) with sequences from the Kobuvirus supergroup used for comparison (n = 19). The tree was created using PhyML (v3.1), and bootstrap support (percent) over 1,000 replicates is shown. Defined genera (right-hand side) and the host range of the clusters (left-hand side) are shown. The bar represents amino acid substitutions per site. (B) Schematic depiction of the RhiV-A genome. The lengths of predicted proteins, cleavage sites, and conserved protein motifs are indicated (see Table 1 for further details). Residue numbering is relative to the polyprotein position (start methionine = 1).

Furthermore, we performed alignments of protein sequences from RhiV-A and related viruses. Consistent with our phylogenetic trees, RhiV-A's closest relatives were ToRaV-A and HGSP, with 50% and 47% aa identities across the full polyprotein, respectively (Table S2). Across the polyprotein, RhiV-A's next closest relatives were the newt-specific Livupivirus (31% aa identity), avian-specific Paraturdivirus (29% aa identity), and mammal-specific Kobuvirus (28% aa identity) and Salivirus (27% aa identity) (Table S2).

RhiV-A maintains classical picornavirus genomic features.

The genome of RhiV-A was annotated by comparison to related picornaviruses, and 10 nt of additional 3′ sequence was obtained with rapid amplification of cDNA ends (RACE). Its genome was 7,456 nt long with 42% GC content, exhibiting a typical picornaviral organization [5′-UTR-VP0-VP3-VP1-2A-2B-2C-3A-3B^VPg-3C^pro-3D^pol-3′-UTR-poly(A)] (Fig. 1B), with translation of the 2,241-aa polyprotein predicted to start with the initiation codon at nt positions 657 to 659 (Fig. 1B). Four of nine polyprotein cleavage sites were conserved between RhiV-A and ToRaV-A (29). The levels of homology between RhiV-A proteins and those from related genera were highest for 3D^pol (40 to 65%) and lowest for the 2A protein (9 to 22%) (see Table S2 in the supplemental material).

Following this, we looked for conserved functional motifs in the predicted protein-coding regions of the RhiV-A genome. Conserved picornavirus protein motifs were seen in the VP0, 3A, 3B^VPg, 3C, and 3C^pol protein-coding regions, suggesting conserved functionality (Table 1). RhiV-A lacked a detectable leader protease (L^pro) sequence, with the predicted polyprotein start codon positioned at the beginning of the VP0 structural coding region (Fig. 1B). This discriminates RhiV-A from other members of the Kobuvirus supergroup, which all contain an L^pro coding sequence (CDS) (30, 31). Like ToRaV-A (5.6 kDa; 53 aa), the predicted RhiV-A 2A protein was short (5.7 kDa; 54 aa) compared to other picornaviruses (∼130 aa) and lacked functional H-box and NC protease motifs, again distinguishing both viruses from the rest of the Kobuvirus supergroup (Table 1) (29, 32, 33).

TABLE 1.

Summary of predicted viral protein products encoded by the RhiV-A genome^a

Genome region	Protein product	Predicted length in RhiV-A (aa)	Predicted molecular mass in RhiV-A (kDa)	Conserved motif(s) present in RhiV-A and known function	Protein function(s) in other picornaviruses	Reference(s)
P1	VP0	458	50	Capsid myristoylation site (G2ANIT)	Capsid structural proteins	72, 73
	VP3	226	26	None known
	VP1	223	25	None known
P2	2A	54	5.7	None identified; lacks functional H-box and NC protease motifs	Cysteine protease; mediates viral polyprotein cleavage, manipulates host translational machinery, initiates viral translation	29, 32, 33, 74–76
	2B	156	17	None known	Membrane alteration, protein trafficking, calcium homeostasis	77
	2C	335	38	None known	Nucleoside triphosphatase activity, membrane alteration	78, 79
P3	3A	66	7.5	NTPase motif (G1259PPGTGKS), helicase motif (D1310DLGQ)	Alters host membrane during infection; interferes with host immune response	80
	3BVPg	33	3.7	Conserved tyrosine residue at the third aa position, where the 3BVPg–5′-UTR bond occurs	Assists viral RNA translation	81
	3Cpro	214	24	Active-site motif (G1725LCG); potential HxC catalytic triad (His40, Asp77/Asp85/Glu72, and Cys169)	Chymotrypsin-like cysteine protease; performs majority of picornavirus polyprotein cleavage	82, 83
	3Dpol	465	53	Motif A (nucleotidyl transfer and recognition) (D2012YKNYD), motif B (nucleotide selection) (G1969S), motif C (nucleotidyl transfer, active site) (Y2105GDD), conserved lysine in motif D (nucleotide incorporation) (K2128), motifs E and F (rNTP binding) (F2154LKR and K1936DELR, respectively)	RNA-dependent RNA polymerase, responsible for replication of the viral genome	84–88

^aConserved protein motifs were determined by comparison to other viruses within the Kobuvirus supergroup.

The RhiV-A 5′ UTR was 656 nt long (Fig. 1B), with no additional sequence detected using 5′ RACE. A BLASTn search revealed 80% nt identity to feline kobuvirus (GenBank accession number KF831027) over a short region (63 nt; positions 538 to 598 of the RhiV-A 5′ UTR) but did not match other viral sequences. The 5′ UTR was predicted to contain a type 1 internal ribosome entry site (IRES) using IRESPred (34). The RhiV-A 3′ UTR is 62 nt, preceding a poly(A) tail (Fig. 1B). BLASTn searches of the 3′ UTR did not show homology to other picornavirus 3′ UTRs.

RhiV-A-like transcripts are present in cane toads from diverse locations.

To assess the geographical distribution of RhiV-A, we screened for RhiV-A sequences in RNA-Seq data sets derived from cane toad spleens (n = 28) sampled from six Australian locations spanning 1,900 km of northern Australia (D. Selechnik, M. F. Richardson, R. Shine, G. P. Brown, and L. A. Rollins, submitted for publication). We detected RhiV-A-like sequences (>97.5% nt identity) in four transcriptomes (Fig. 2A) derived from three of the six locations (Fig. 2B). Transcripts represented only partial regions of the complete RhiV-A genome, ranging from 12.9 to 59.9% genomic coverage (962 to 4,466 nt) (Fig. 2A and C). In addition, the raw read coverage of these regions was very low (range, 1.5 to 25.1 reads per nt position; mean, 7.8 reads per nt position) compared to that of the original RhiV-A liver transcript (29.5 reads per nt position) (Fig. 2C).

FIG 2.

Details of RhiV-A-like sequences in four splenic transcriptomes of Australian cane toads. (A) Visual representation of the splenic RhiV-A-like sequences, aligned to the RhiV-A genome. Black bars represent coverage of the viral genome, and lines represent gaps in the alignment. (B) Three sample collection locations (red stars) of the four RhiV-A-positive toads, demonstrating the range of RhiV-A infection. Black stars represent sampling locations for the pooled liver transcriptome. Brown shading represents the current cane toad distribution, with the invasion front moving westwards through WA (71). (C) Table detailing RhiV-A sequences present in publicly available data sets. SRA accession numbers of individual transcriptomes and locations of toad collection are included. The table also includes details of the RhiV-A sequences observed in the spleen transcriptomes, including genomic coverage, regions of the virus detected, nucleotide identity, and RNA-Seq coverage of the splenic sequences to the original viral transcript identified in liver tissue.

A novel full-length epsilonretrovirus sequence was detected in cane toad liver.

There is a considerable gap in the scientific literature on amphibian retroviruses, with only two endogenous retroviruses (ERVs) identified so far in Xenopus laevis and Xenopus tropicalis (35, 36). To address this, we aimed to identify novel retroviruses in the cane toad. A retrovirus may be present as viral RNA or as a provirus inserted into the genome (37); therefore, both transcriptome and genome data sets were screened for retrovirus sequences.

We noted the presence of an abundant retrovirus-like transcript that was the length of a retroviral genome (9,859 nt; 4.53 TPM), designated Rhinella marina endogenous retrovirus (RMERV) (Fig. 3). The assembly of raw reads onto the transcript showed complete coverage (mean coverage, 21.3 reads per nt position).

FIG 3.

Properties of RMERV. (A) Unrooted RAxML phylogenetic tree of the Pol amino acid sequences (970 aa) from representative retrovirus sequences for comparison (n = 26). GenBank accession numbers are displayed. Numbers adjacent to nodes show bootstrap support (percent) from 1,000 trees. The bar represents amino acid substitutions per site. RMERV clusters monophyletically with XLERV in the genus Epsilonretrovirus. REV, reticuloendotheliosis virus; MLVMS, Moloney murine leukemia virus; FLV/GALV: feline/gibbon ape leukemia virus; PERV, porcine ERV; BFV/FFV/SFV, bovine/feline/simian foamy virus; FIV/HIV-1/SIV-2, feline/human/simian immunodeficiency virus type 1/2; HTLV-1/2/STLV-2, human/simian T-lymphotrophic virus type 1/2; ALV, avian leukosis virus; RSV, Rous sarcoma virus; MMTV, mouse mammary tumor virus; MPMV, Mason-Pfizer monkey virus; JSRV, Jaagsiekte sheep retrovirus. (B) Structure and predicted features of the RMERV genome (10,209 nt) containing three ORFs (region X/gag, pol, and env) flanked by two LTRs (471 nt each). Region X represents a potential protein-coding sequence in the same ORF as gag. gag and pol are in the same frame and separated by 57 nt. pol and env have a 13-nt overlap requiring a −1 frameshift for env translation. Conserved motifs are annotated, with amino acid numbering relative to the start residue of each encoded ORF. Italicized numbers below the genome indicate nucleotide positions. PR, protease; RT, reverse transcriptase; IN, integrase; MHR, major homology region; TM, transmembrane; PPT, polypurine tract. (C) Eight provirus sequences closely related to RMERV (>95.0% nt identity) aligned to the RMERV genome. Genomic scaffolds (s.) containing the provirus are shown above the alignment. Intact ORFs are represented by colored bars. Gray bars indicate that the ORF has lost coding capacity: a gap indicates a deletion, an arrow beneath the scaffold represents an insertion (inserted base indicated in boldface type), and a substitution is indicated with a black bar, with residues indicated above. Each provirus is flanked by two LTRs of 471 bp, except for scaffold 6576, where the 5′ LTR is absent. LTRs are annotated adjacent to the viral genome, with yellow LTRs being 100% identical and white LTRs being <100% identical. The RMERV transcript contains the complete 3′ LTR and a partial region (185 nt) of the 5′ LTR. Labeling of the nucleotide sequence is relative to the beginning of the complete 5′ LTR. Short dashed lines represent the end of a scaffold.

To classify RMERV, a phylogenetic tree was constructed using an alignment of the predicted pol amino acid sequence (1,264 aa) with sequences representing each Retroviridae genus (Fig. 3A). RMERV clusters in the subfamily Orthoretrovirinae, genus Epsilonretrovirus, in a monophyletic lineage with Xenopus laevis ERV (XLERV). This lineage was distinct from known exogenous epsilonretroviruses, including walleye dermal sarcoma virus (WDSV), walleye epidermal hyperplasia virus 1/2 (WEHV1/2), Atlantic swim bladder sarcoma virus (ASSBV), and a zebrafish ERV (ZFERV) (Fig. 3A).

Genomic features of RMERV.

RMERV exhibits a classic retroviral genome structure (long terminal repeat [LTR]-gag-pol-env-LTR) (Fig. 3B). The RMERV transcript did not contain a complete LTR, but the intact LTR sequence (471 nt) was identified in endogenous RMERV copies (Fig. 3C). After the 5′ LTR, the genome has a leader region of 736 nt, which is comparable to the size of the XLERV leader region (750 nt) (36).

A large open reading frame (ORF) of 2,622 nt succeeded the leader region; however, the predicted gag start site did not occur until 873 nt after the start of this ORF (detailed below). This left a putative protein-coding region of 873 nt (291 aa) denoted “region X” (Fig. 3B). Given that accessory proteins encoded upstream of gag exist in some epsilonretroviruses (XLERV, WEHV, WDSV, and ASSBV, ranging from 25 to 134 aa), we thought that region X may represent a retroviral accessory protein encoded in the same ORF as gag. Therefore, a BLASTp search was performed using region X as a query, generating matches to hypothetical eukaryotic proteins from X. laevis (OCT78054, OCT82261, OCT82260, OCT91396, and OCT82870) (E values of <1e⁻¹⁰; aa identity ranging from 32 to 42%). Like the equivalent region in WDSV (38), region X may also encode part of a spliced transcript, although we were unable to identify definitive splice sites.

Overall, RMERV exhibits a genomic organization more like XTERV1 than like XLERV. Like XTERV1, RMERV exhibits noncontiguous gag/pol ORFs, which are in the same frame and separated by 57 nt (70 nt in XTERV1). Additionally, both RMERV and XTERV1 exhibit overlapping pol and env ORFs (1 nt and 217 nt, respectively) (Fig. 3B), whereas these regions do not overlap in XLERV (36). This mixture of features in RMERV suggests a diversity of genome organizations within members of the genus Epsilonretrovirus, which do not strictly correlate with genetic lineage.

gag.

The predicted RMERV gag ORF begins with a Kozak sequence (translation initiation signal) at nt position 2080 of the RMERV genome (ACAUGGGU) (NetStart score, 0.606). That this codon likely represents the N terminus of RMERV Gag is further supported by a predicted myristoylation motif (G₂LAGSH) (Fig. 3B), which typically signals the start of every retroviral Gag protein (39). Translation initiation here would produce a 96-kDa protein of 873 aa, which matched closely to that of XLERV (E value = 7.49e⁻³⁷; 34% identity over 268 residues) but to no other eukaryotic or viral sequence.

Retroviral Gag is generally processed into three major products, matrix (MA), capsid (CA), and nucleocapsid (NC) (40). Due to the nonconserved nature of retroviral protease cleavage sites, the exact junctions between these products could not be determined. However, we identified two other conserved protein motifs in gag, including a major homology region (V₆₉₅QQEPGEAVEKYAARLTM; key conserved residues are in boldface) and a Cys-His box (C₈₃₈WECGSPNHLRRDC) (41, 42).

pol.

The RMERV pol CDS was 3,795 nt, encoding a predicted 140-kDa protein of 1,264 aa (Fig. 3B). It exhibited sequence homology to XLERV Pol (GenBank accession number AJ506107) (E value = 0; 50% aa identity over 212 aa) and to genomic DNA (gDNA) sequences from organisms, including the poison dart frog Dendrobates ventrimaculatus (GenBank accession number X95795) (E value = 0; 89% identity over 324 aa), X. tropicalis (GenBank accession number XM_012965051) (E value = 0; 45% identity over 784 aa), and the painted turtle Chrysemys picta bellii (GenBank accession number AC239505) (E value = 6.07e⁻¹⁴⁶; 40% identity over 792 aa).

In all retroviruses, precursor Pol is cleaved into protease (PR), reverse transcriptase (RT), and integrase (IN) (40). Based on known cleavage sites within the fish virus WDSV and an approximation of PR (∼110 aa) and RT (∼650 aa) lengths, the predicted PR/RT cleavage site is likely positioned at either G₁₁₂A or A₁₁₃E, and the predicted RT/IN cleavage site lies at either A₈₂₉K or A₈₃₄A (43). RMERV pol contained PR and RT active-site motifs (D₁₆TG and Y₃₃₂VDD₃₃₅, respectively), an IN Zn²⁺-chelating HHCC domain (H₉₃₈GPTH.CPVC₉₇₅), and a catalytic core domain (D₁₁₅₅.E₁₁₉₂) (Fig. 3B) (36, 44).

env.

The predicted RMERV env region was 1,902 nt long, encoding a 71-kDa protein of 633 aa (Fig. 3B). The putative pol and env ORFs have a 13-nt overlap, with Env translation requiring a −1 frameshift (Fig. 3B). A tBLASTn search of the env CDS revealed similarity to X. laevis syncytin-like transcripts (GenBank accession numbers XM_018224752 and XM_018257767) (E value = <3.86e⁻⁹⁶; 38% identity over 591 aa), to the XLERV env sequence (E value = 4.01e⁻⁹³; 38% identity over 591 aa), and to eukaryotic genome sequences from primates, fish, and birds (data not shown).

A critical feature of Env is the transmembrane (TM) subunit, which mediates viral entry (45). We identified a C-terminal TM region with a potential immunosuppressive motif (L₄₈₀QNRLALDMILAEKGG) (46) a and hydrophobic TM anchor domain (A₅₅₇CLVLIGLIVVGLMILCCVIPL) (Fig. 3B). Downstream (61 nt) of the env CDS, RMERV contained a polypurine tract (PPT) of 10 nt (Fig. 3B). Altogether, our findings suggest that RMERV maintains classical features of retroviral proteins while being phylogenetically distinct from known retroviruses.

RMERV-like transcripts are present in cane toads from six Australian locations.

Following the identification of the full-length RMERV transcript in our pooled liver transcriptome, we analyzed 28 spleen data sets for RMERV-like transcripts (>95.0% nt identity to the viral genome). RMERV-like sequences were present in all 28 toads, with sequence coverage ranging from 2.0 to 25.4% of the RMERV genome (see Table S3 in the supplemental material).

Full-length and truncated RMERV proviruses are present in the cane toad genome.

Part of the retroviral life cycle involves integration into the host genome, leading to retroviral sequences that accumulate during infection and can persist through generations if the germ line becomes infected (47). Therefore, we assessed the abundance of RMERV-like sequences in the cane toad genome. Eight genomic scaffolds contained a full-length proviral sequence (10,209 nt, excluding insertions or deletions) with >99.0% nt identity to the RMERV genome (Fig. 3C). In addition, 21 more genomic scaffolds contained truncated portions of the RMERV genome (>95.0% identity) (see Table S4 in the supplemental material). Fourteen of these were truncated at the 5′ end (truncations ranging from 121 to 9,262 nt), and four were truncated at the 3′ end (truncations ranging from 968 to 9,404 nt), while the remaining six scaffolds contained the RMERV LTR sequence only (Table S4). All but one endogenous copy of RMERV contained two 471-nt flanking LTRs, which were >99.6% identical to the 3′ LTR of the original transcript (Fig. 3C). LTRs of the same provirus were 100.0% identical to each other on scaffolds 3945, 6861, 9679, 15382 and >99.6% identical on scaffolds 7491, 10654, and 12245 (Fig. 3C). On scaffold 6576, only a 3′ LTR was detectable, as the scaffold ended before the 5′-LTR sequence (Fig. 3C).

Ageing of full-length RMERV insertions.

The full-length RMERV LTR divergence ranged from 0.0 to 0.4%. Using previously established methods (48), this generated an upper estimate of the most recent full-length insertion to be 1.9 million years ago (MYA) (95% confidence interval [CI], 1.3 to 2.2 MYA). The minimum LTR divergence was 0.0% and yielded a minimum integration time of 0 MYA.

Circovirus-like EVEs are present in the cane toad genome.

A BLAST search of 26,317 viral protein sequences against the cane toad genome revealed three hits to the replication initiation gene (rep) of small DNA viruses from the Circoviridae. Two hits were present on scaffold 15507: cane toad endogenous viral element 1 (CTEVE1) (length, 288 nt; scaffold nt positions 10665 to 10378) and CTEVE2 (length, 243 nt; scaffold nt positions 22512 to 22270). Scaffold 27217 contained a third EVE, CTEVE3 (length, 585 nt; scaffold nt positions 5938 to 5354). The EVEs were mapped to the Circoviridae reference viral genome, porcine circovirus 1 (PCV-1) (GenBank accession number NC_001792), aligning to the rep CDS (Fig. 4). Relative to the PCV-1 genome, CTEVE1 and CTEVE3 (CTEVE1/CTEVE3) share a 335-nt overlap, and CTEVE2/CTEVE3 share a 135-nt overlap, but the EVEs share only moderate nucleotide identity in these overlapping regions (49% and 66%, respectively) (Fig. 4). The EVEs were analyzed for the presence of stop codons to determine coding capacity: one was identified in CTEVE2 (nt positions 181 to 183), and two were identified in CTEVE3 (nt positions 121 to 124 and 268 to 270) (Fig. 4).

FIG 4.

Circovirus-like EVEs in the cane toad genome. Three circovirus-like EVEs, CTEVE1, CTEVE2 (288 nt and 243 nt, respectively, both located on scaffold 15507), and CTEVE3 (585 nt, located on scaffold 27217) were identified in the cane toad genome. An alignment of the three EVEs to the type species of the family Circoviridae, PCV-1 (GenBank accession number JX566507), is shown. Circoviruses are single-stranded DNA (ssDNA) viruses with a 2-kb circular genome encoding two proteins, Rep (replication initiation protein) and Cap (capsid protein); these are displayed above the alignment and linearized below. All EVEs align to the rep gene, labeled as a blue bar. Yellow blocks indicate circovirus-like EVEs, with numbers underneath indicating the nucleotide position relative to PCV-1. Gray shaded boxes indicate the nucleotide identity between the overlapping regions of the EVEs (264-nt position overlap between CTEVE2 and CTEVE3, sharing 49% nt identity; 122-nt position overlap between CTEVE3 and CTEVE1, sharing 66% nt identity). Black bars on the EVEs indicate a stop codon. Dashed lines show cane toad genomic sequences flanking the EVEs.

To confirm that the EVEs found were not sequencing artifacts, they were amplified using PCR from genomic DNA (gDNA) extracted from one liver used to generate the reference genome sequence. We detected three EVEs matching 100% to the sequences identified computationally (data not shown).

Phylogenetic analysis of circovirus-like EVEs.

Trees were constructed using amino acid sequences of the EVEs and 25 close relatives as determined by BLAST. For CTEVE1, the top BLAST hit was the avian beak and feather disease virus (GenBank accession number HQ641561) (65% identity over 82 aa); for CTEVE2, it was rodent circovirus (GenBank accession number KY370034) (56% identity over 82 aa); and for CTEVE3, it was finch circovirus (GenBank accession number DQ845075) (48% identity over 211 aa). In the case of CTEVE3, we noted a hit to a partial sequence from brown toad circovirus (GenBank accession number KF358282) (E value = 2.08e⁻³²), exhibiting 69% identity over 93 aa. However, the brown toad circovirus sequence was not included in the phylogenetic analysis due to its short length.

CTEVE1 clustered in an avian-like lineage (see Fig. S2A in the supplemental material). Its closest neighbor in this tree is barbel circovirus (GenBank accession number JF279961), but bootstrap support for this position was not strong (26%) (Fig. S2A). A tree of CTEVE2 indicates a potential lineage with Silurus glanis (catfish) circovirus (GenBank accession number JQ011377), rodent circovirus, fox circovirus (GenBank accession number KP260925), and canine circovirus (GenBank accession number KT734816) (Fig. S2B). There was moderate bootstrap support (48%) to suggest that CTEVE3 clusters with avian-like circoviruses, including those infecting gulls, parrots, finches, starlings, and columbids (Fig. S2C).

DISCUSSION

Cane toads pose a serious threat to ecosystems where they have been introduced, and their distribution in Australia is increasing (4). Viral biocontrol is an attractive solution but is hindered by the paucity of described cane toad viruses. This study employs a metatranscriptomic approach to identify active viral infections among invasive Australian cane toads and a genomic approach to identify past infections.

RhiV-A, a novel cane toad-infecting picornavirus.

Of the 37 defined genera in the Picornaviridae, 30 are specific to either mammals or birds, representing a historical sampling skew toward these hosts (30). Recently, 44 novel picornaviruses from fish, reptiles, and amphibians have been discovered, greatly expanding the known host diversity of the Picornaviridae (26). These novel viruses were included in our analysis to better determine the relationship of RhiV-A to these novel lower-vertebrate lineages.

RhiV-A's closest relatives are two reptilian viruses, ToRaV-A and HGSP, from tortoise and gecko, respectively. These three viruses form a cluster basal to the rest of the Kobuvirus supergroup, which infects birds and mammals (Fig. 1A; see also Fig. S1 in the supplemental material). This pattern is consistent with data from the above-mentioned study, which notes the coevolution of RNA virus and host, resulting in a phylogeny that resembles the evolution of the hosts themselves (26). Additionally, the close relationship between amphibian and reptilian viral sequences could be a result of horizontal viral transmission, facilitated by the number of physiological similarities between these groups (49). Interestingly, RhiV-A exhibits closer homology to reptilian viruses (ToRaV-A and HGSP) than to other amphibian picornaviruses (the genera Ampivirus, Livupivirus, and four viruses from newt, toad, frog, and caecilian) (Fig. 1A and Fig. S1) (26). This supports the existence of multiple and distinct amphibian-specific lineages within the Picornaviridae.

RhiV-A contained several interesting genomic features distinguishing it from the rest of the Kobuvirus supergroup. Namely, RhiV-A lacked an L^pro CDS with the predicted start codon positioned at the beginning of the VP0 structural coding region (Fig. 1B). This distinguishes it from ToRaV-A and HSGP, which have predicted L^pro sequences of 114 and 157 aa, respectively (26, 29). Despite this, the L^pro CDSs of ToRaV-A and HSGP lack the GxCG functional protease motif, suggesting a loss of L^pro function in this group of viruses (29). The RhiV-A 2A protein was short (5.7 kDa; 54 aa) compared to those of other picornaviruses (∼130 aa) and lacked functional H-box and NC protease motifs. The 2A proteins of ToRaV-A (53 aa) and HGSP (59 aa) are also notably short and lack these motifs, again suggesting a loss of 2A function and supporting the classification of these viruses into a unique cluster (Table 1 and Fig. 1B) (29, 32, 33).

Although RhiV-A has a different host range (amphibian) than other viruses in its cluster (reptilian), RhiV-A is not likely to constitute a novel genus. Distinct picornavirus genera exhibit pairwise divergence exceeding 66% within the P1 region and 64% within the 2C^hel, 3C^pro, and 3D^pol regions (30). In the P1 region, RhiV-A and ToRaV-A diverged by 51% at the amino acid level, and within the 2C^hel, 3C^pro, and 3D^pol regions, they diverged by 49%, 61%, and 45%, respectively (average, 52%) (Table S2). Despite this, the recent discovery of 44 novel picornaviruses of lower vertebrates is likely to warrant the creation of multiple novel genera and may greatly restructure existing Picornaviridae genera. The classification of RhiV-A is thus at the discretion of the ICTV.

RMERV is a novel cane toad epsilonretrovirus.

We report a full-length retrovirus sequence in our combined liver transcriptome, designated RMERV (Fig. 3). Phylogenetic analysis of the complete pol region revealed that RMERV is a member of the genus Epsilonretrovirus, a genus including exogenous oncogenic fish viruses and ERVs from fish and frogs (Fig. 3A) (50).

Our findings support that the RMERV genome is derived from a host provirus and not from an infecting retrovirus. This is because the length of the 3′ LTR of the RMERV transcript was equivalent to those of the 5′ and 3′ LTRs of the RMERV proviruses. In an exogenous retroviral transcript, it would be expected that the LTRs are shorter, as they lack the U3 and U5 regions, which are copied to complementary ends of the provirus (51). Furthermore, the genome sequence of the cane toad harbors 8 full-length (Fig. 3C) and 21 truncated genomic copies of RMERV, a number comparable with those of an X. tropicalis ERV, XTERV1 (n = 11) (35). We also identified 204 partial RMERV transcripts in publicly available splenic transcriptomes (Table S3), indicating the widespread genomic insertion and transcription of RMERV-like sequences in cane toads across Australia. All full-length and truncated RMERV copies were closely related to each other (>95.0% nt identity) and to the full-length transcript detected by RNA-Seq, indicating that they were derived from a single viral species.

Using established methods to estimate the age of retroviral insertions, we calculated the oldest of seven full-length insertions to have integrated approximately 1.9 MYA (95% CI, 1.3 to 2.2 MYA), and the four proviruses with identical LTRs presumably integrated more recently. This is relatively young for epsilonretrovirus integration; the XTERV1 integration date is approximately 14 MYA (35), and epsilon-like ERV sequences in primates are placed at 16 to 90 MYA (52).

The cane toad genome contains circovirus-like elements.

Circoviruses undergo a nuclear replication cycle with a strong dependence on host replication factors, making them predisposed to EVE formation (53, 54). We demonstrated the presence of three circovirus-like EVEs (CTEVE1 to CTEVE3) in the cane toad (Fig. 4). Analysis of the circoviral EVEs suggests that CTEVE1 and CTEVE2 share homology with circoviruses from marine organisms (barbel, catfish, and shrimp) (see Fig. S2A and S2B in the supplemental material). The shared aquatic environment of fish and amphibians could allow for horizontal viral transmission or convergent viral evolution (55). Alternatively, CTEVE3 appears to cluster with avian-like circoviruses (Fig. S2C). However, phylogenetic inferences about the EVEs should be tentative because their short lengths (243 to 585 nt) resulted in low bootstrap support within phylogenetic trees (Fig. S2).

The origins of viruses in cane toads in Australia.

To confirm the geographical distribution of RhiV-A, we identified highly similar (>97.5% nt identity) sequences (12.9 to 59.9% genome coverage) in 4 of 28 (14%) spleen transcriptome data sets. Importantly, these toads were not in contact with each other, as were the 16 sampled toads, before or after collection, proving the existence of RhiV-A infection in these locations. The three distinct locations from which these toads were derived span over 1,000 km, suggesting a widespread Australian distribution of RhiV-A stretching to the WA invasion front (Fig. 2B). Additionally, all 28 toads from the spleen data sets harbored partial RMERV transcripts, suggesting extensive integration of RMERV-like sequences in Australian toads (see Table S4 in the supplemental material).

Introduced populations often exhibit a lower diversity of pathogens than their native-range populations (56). Cane toads were derived from a small founder population (n = 101) that had undergone successive bottlenecks. Phylogenetic analysis of a common pathogen of Australian cane toads, the nematode lungworm (Rhabdias pseudosphaerocephala), has proven its South American origin (57), providing evidence that native-range infections can become widespread in Australian cane toads. Similarly, the widespread distribution of RhiV-A raises the question of whether it too arrived with founders or was transmitted horizontally from a native species (2, 58). The specific origin of RhiV-A can be further explored by screening international toad populations, and native Australian amphibians, with detailed phylogeographical analysis (57).

The aim of this project was to identify new cane toad viruses that may facilitate future studies on viral biocontrol. We have completed an important initial step by describing novel cane toad-specific viral sequences. Further studies will need to isolate these viruses and perform pathogenesis studies in vitro and in vivo to determine if they represent suitable biocontrol agents. Overall, our findings contribute toward a deeper understanding of the diversity of the significantly undersampled amphibian virome.

MATERIALS AND METHODS

Sample collection and RNA-Seq.

Adult cane toads (n = 16) were collected from QLD (S 19.2335°, E 146.7833°; n = 8) and Western Australia (WA) (S 14.83333°, E 126.6667°; n = 8). For 10 weeks, toads were held in a captive facility and subjected to the same manipulations for breeding and behavioral experiments. Following lethal injection with 150 mg/kg of body weight sodium pentobarbital, livers were excised and stored in RNAlater (Qiagen, USA) and kept cool during transport. Approximately 100 mg of tissue was power homogenized in TRIzol LS reagent (Invitrogen, Carlsbad, CA) and subjected to phenol-chloroform extraction. Total RNA was purified with the RNeasy minikit (Qiagen) and included DNase treatment with the RNase-free DNase set (Qiagen). RNA concentration and integrity were determined using UV spectrophotometry and agarose gel electrophoresis. All RNA samples were then pooled in equal mass proportions, and an RNA library was prepared after the depletion of rRNA using a RiboZero Gold kit (Epidemiology) (Illumina Inc., USA). Libraries were prepared at the Australian Genome Research Facility, Melbourne, Victoria, Australia, using a TruSeq stranded total RNA library preparation kit (Illumina Inc.). Paired-end sequencing (100-nt-long reads) was performed on an Illumina HiSeq 2500 system using v4 chemistry.

Transcriptome assembly and identification of full-length viral sequences.

Raw reads were de novo assembled into transcripts using Trinity (v2.5.1) (59). This included in silico normalization and quality control (removal of adapter sequences and trimming). Transcript quantification was performed using RSEM (v1.3.0) (60). The assembled transcriptome was annotated with Diamond (v0.9.10) (61) using the NCBI nonredundant (nr) database (October 2017) as a reference, with a maximum expectancy value (E value) cutoff of 1e⁻⁴. Annotations were filtered using the keyword “virus,” and false hits were disregarded manually by reblasting to the NCBI nt/nr database (October 2017) or inspection of sequences for premature stop codons. To visualize RNA-Seq coverage of putative viral transcripts, raw reads were mapped back to the transcript using Bowtie (v7.0.10) (62).

Analysis of additional transcriptomes for viral sequences.

Twenty-eight publicly available spleen transcriptomes of Australian cane toads from six locations were downloaded from the Sequence Read Archive (SRA) (BioProject accession number PRJNA395127), assembled, annotated, and analyzed for the presence of viral sequences as described above.

Viral sequence annotation and phylogenetic analysis.

Transcriptomic or genomic (described below) virus-like sequences were compiled, and a tBLASTn search was performed against the nr database to generate a list of close viral relatives (E value of ≤1e⁻¹⁰). Following this, the conserved polymerase region was used to approximate the position of virus-like sequences within their family of origin. If the complete polymerase sequence was unavailable (i.e., in the case of fragmented EVEs), a partial sequence was used. Representative corresponding sequences of each identified family were downloaded from the NCBI. Amino acid alignments of viral protein sequences were performed using MAFFT (v1.3.4), and trimAl (v1.4.1) (63) was used to remove ambiguously aligned regions. ML phylogenetic trees were then estimated using PhyML (v3.1) (64) or RAxML (v8.2.11) (65) with 1,000 nonparametric bootstrap replicates. Locations of viral proteins and cleavage sites within viral sequences were predicted by MAFFT alignment with related viruses. Translation initiation signals (Kozak sequences) were predicted with NetStart (v1.0) (66).

Reverse transcriptase PCR and PCR detection of virus-like sequences from cane toad tissue.

To detect picornavirus-like transcripts, total liver RNA from individuals was reverse transcribed to cDNA with the SuperScript III first-strand synthesis system (Invitrogen). cDNA generated from the sampled livers was used in a PCR targeting the 5′ UTR (Table 2). To detect circovirus-like EVEs, gDNA from the primary genome sequencing run (BioProject accession number PRJEB24695; European Nt Archive Study assembly accession number GCA_900303285) was used in three separate PCRs with primers designed to target the EVE-flanking regions. All PCRs were performed using Taq DNA polymerase (NEB, MA, USA) with 0.2 μM both forward and reverse primers (Table 2). PCR products were visualized with agarose gel electrophoresis and Sanger sequenced at the Ramaciotti Centre for Genomics at the University of New South Wales, Sydney, Australia.

TABLE 2.

Primers used in this study to amplify viral sequences from cane toad tissues

Primer pair	Forward primer sequence (5′–3′)	Reverse primer sequence (5′–3′)	Annealing temp (°C)	Target region(s)	Amplicon size (bp)
RhiV-A1/RhiV-A2	GGATCTTTCCTCTTTATGAGC	GGCATTCCTCATATTTGACTCC	45	RhiV-A 5′ UTR	302
RhiV-A3/TX30SXN (used for 3′ RACE)	TACTATCACTTCCTTGTGC	GACTAGTTCTAGATCGCGAGCGGCCGCCCa	45	RhiV-A-tagged 3′ UTR	Variable
RhiV-A4 and UAP (5′-RACE universal amplification primer)	AGTGTGTCACCCTTTAGCG	GGCCACGCGTCGACTAGTACGGGIIGGGIIGGGIIGb	50	RhiV-A cDNA, 5′ UTR tailed with poly(C) overhang	Variable
CTEVE1F/CTEVE1R	TCACTATCCCTTGTGTCC	TTTAAGATGGAGGTGATTGG	55	CTEVE1	363
CTEVE2F/CTEVE2R	ACAAGTCCCAGAGTATTTCC	CATGTCTTAAGAGTCG	45	CTEVE2	484
CTEVE3F/CTEVE3R	CTTTCACGTGATTGTTGGATG	GTCTTTCCTTTCCTCTTCC	55	CTEVE3	869

^aPrimer sequence in reference 68.

^bPrimer sequence in reference 67. I, inosine.

Rapid amplification of cDNA ends.

RACE was used to obtain complete sequences of picornavirus transcripts identified using RNA-Seq. This included 5′ RACE and 3′ RACE, which were performed as previously described (67, 68). RACE primers used are listed in Table 2.

Identification of retroviral and nonretroviral EVEs.

All EVE analyses were performed in Geneious (v10.2.3) (69). To detect proviruses (retrovirus-like EVEs), a BLASTn search (E value = ≤1e⁻¹⁰) of putative full-length retroviral genomes was performed against the cane toad genome assembly (v2.2). BLAST hits corresponding to the retroviral transcript (>95.0% identity) were compiled and inspected for the presence of ORFs, insertions, deletions, and substitutions.

Nonretroviral EVEs were identified based on previously devised methods (28). First, a list of representative viral protein sequences (DNA viruses and non-reverse-transcribing RNA viruses) was compiled (n = 24,169) based on the NCBI RefSeq database and the ICTV 2017 species list (see Table S1 in the supplemental material). This list was supplemented with protein sequences from divergent insect viruses (n = 1,445) (24, 25). Additionally, the keywords “amphibian” and “virus” were used to locate amphibian-specific viral genomes in the NCBI database (n = 353). All query sequences (n = 25,614) were downloaded and translated to protein sequences (Table S1). A tBLASTn search of all sequences was undertaken against the cane toad genome (E value = ≤1e⁻¹⁰). Duplicate BLAST hits derived from similar query proteins were discarded, with the longest hit being kept for analysis. To remove false-positive hits, a reciprocal BLASTp search was performed by querying translated hits against the nr database.

Ageing of retroviral insertions.

To estimate the integration age of the full-length RMERV proviruses, we analyzed LTRs flanking the same provirus as previously described (48). These are identical at the time of integration and are assumed to accumulate mutations at the host neutral rate of evolution (47). All proviruses containing 2 LTRs were taken, and the minimum and maximum divergences between LTRs flanking the same provirus were calculated with MAFFT alignment. This value was applied to the formula t = k/2N, where t is time (years), k is LTR divergence (nucleotide substitutions/site), and N is the neutral rate of host evolution (defined as 1.03 × 10⁻⁹ [95% CI = 0.92 × 10⁻⁹ to 1.53 × 10⁻⁹] substitutions/site/year based on estimates in frog nuclear DNA [70]).

Ethics.

Cane toads used in this study were euthanized under University of Sydney ethics permit number 2017/1151.

Accession number(s).

RNA-Seq data are deposited in the NCBI Sequence Read Archive (SRA) under BioProject accession number PRJNA419245. Viral genome sequences are deposited under GenBank accession numbers MG967619 (RhiV-A) and MG981046 (RMERV).