SUMMARY
Parvoviruses (family Parvoviridae) are small, non-enveloped DNA viruses that infect a broad range of animal species. Comparative studies, supported by experimental evidence, show that many vertebrate species contain sequences derived from ancient parvoviruses embedded in their genomes. These ‘endogenous parvoviral elements’ (EPVs), which arose via recombination-based mechanisms in infected germline cells of ancestral organisms, constitute a form of ‘molecular fossil record’ that can be used to investigate the origin and evolution of the parvovirus family. Here, we use comparative approaches to investigate 198 EPV loci, represented by 470 EPV sequences identified in a comprehensive in silico screen of 752 published vertebrate genomes. We investigated EPV loci by constructing an open resource that contains all of the data items required for comparative sequence analysis of parvoviruses and uses a relational database to represent the complex semantic relationships between them. We used this standardised framework to implement reproducible comparative phylogenetic analysis of combined EPV and virus data. Our analysis reveals that viruses closely related to contemporary parvoviruses have circulated among vertebrates since the Late Cretaceous epoch (100-66 million years ago). We present evidence that the subfamily Parvovirinae, which includes ten vertebrate-specific genera, has evolved in broad congruence with the emergence and diversification of major vertebrate groups. Furthermore, we infer defining aspects of evolution within individual parvovirus genera - mammalian vicariance for protoparvoviruses (genus Protoparvovirus), and inter-class transmission for dependoparvoviruses (genus Dependoparvovirus) - thereby establishing an ecological and evolutionary perspective through which to approach analysis of these virus groups. We also identify evidence of EPV expression at RNA level and show that EPV coding sequences have frequently been maintained during evolution, adding to a growing body of evidence that EPV loci have been co-opted or exapted by vertebrate species, and especially by mammals. Our findings offer fundamental insights into parvovirus evolution. In addition, we establish novel genomic resources that can advance the development of parvovirus-related research - including both therapeutics and disease prevention efforts - by enabling more efficient dissemination and utilisation of relevant, evolution-related domain knowledge.
INTRODUCTION
Parvoviruses (family Parvoviridae) are a diverse group of small, non-enveloped DNA viruses that infect a broad and phylogenetically diverse range of animal species [1, 2]. The family includes numerous important pathogens of humans and domesticated species, including erythroparvovirus B19 (fifth disease), carnivore protoparvovirus 1 (canine parvovirus) and carnivore amdoparvovirus 1 (Aleutian mink disease). Parvoviruses are also being developed as next-generation therapeutic tools - rodent protoparvoviruses (RoPVs) are promising anticancer agents that show natural oncotropism and oncolytic properties [3, 4], while adeno-associated virus (AAV), a non-autonomously replicating dependoparvovirus, has been successfully adapted as a gene therapy vector, and parvoviruses are leading candidates for the further development of human gene therapy [5, 6].
Parvoviruses have highly robust, icosahedral capsids (T=1) that contain a linear, single-stranded DNA genome typically ~5 kilobases (kb) in length. Parvovirus genomes are typically very compact and generally exhibit the same basic genetic organization comprising two major gene cassettes, one (Rep/NS) that encodes the non-structural proteins, and another (Cap/VP) that encodes the structural coat proteins of the virion [2]. However, some genera contain additional open reading frames (ORFs) adjacent to these genes or overlapping them in alternative reading frames. The genome is flanked at the 3’ and 5’ ends by palindromic inverted terminal repeat (ITR) sequences that are the only cis elements required for replication.
Recent years have seen many important advances in understanding of parvovirus evolution and diversity, driven primarily by dramatic increases in the availability of DNA sequence data and investments in deriving novel adeno-associated virus capsid for gene therapy applications. Metagenomic sequencing has enabled the discovery of numerous novel parvovirus species, which in turn has led to the taxonomic re-organization of the Parvoviridae to include additional subfamilies and genera [1]. In addition, whole genome DNA sequencing has revealed that DNA sequences derived from parvoviruses are widespread in animal genomes [7–13]. These endogenous parvoviral elements (EPVs) are thought to have arisen when parvovirus infection of germline cells (i.e., gametes, gamete producing cells, or early-stage embryos) led to integration of parvovirus-derived DNA into chromosomal DNA so that it was subsequently inherited as a newly acquired allele. Integration of parvovirus DNA can occur via cell-mediated, non-homologous recombination but may also be mediated by the activities of virus-encoded proteins [14, 15]. Comparative genomic studies have shown that EPV sequences often occur as orthologous loci in multiple related host species, demonstrating that they were incorporated into the germline of a common ancestor. Thus, species divergence times – which are in part based on evidence from the fossil record - provide a robust method of deriving minimum age estimates for EVE insertions. Many EVEs represent virus lineages that have not been described previously and may be extinct [7, 16]. However, others clearly represent members of contemporary virus groups, and age estimates obtained for these EVEs provide insights into their long-term evolutionary history [7, 17–20].
The extent to which EPVs have reached fixation through positive selection - as opposed to incidental factors such as founder effects, population bottlenecks, and genetic hitchhiking - remains unclear. Potentially, EPV genetic information might sometimes be co-opted or “exapted” as has been reported for EVEs derived from other virus groups, including retroviruses (family Retroviridae) [21, 22] and polydnaviruses (family Polydnaviridae) [23]. Recent studies have revealed that two distinct, fixed EPVs in the germline of (i) the degu (Octodon degus) and (ii) family Elephantidae (elephants) – both of which encode an intact Rep protein ORF - exhibit similar patterns of tissue-specific expression in the liver [24, 25]. These observations suggest that expression of Rep protein or mRNA might – in some way - be physiologically relevant. More broadly, incorporation of parvovirus-derived DNA into animal germlines may provide a novel DNA substrate for the evolution of new genes – for example, guinea pigs (Cavia porcellus) encode a predicted polypeptide gene product comprising a partial myosin9-like (M9l) gene fused to a 3’ truncated, EPV-encoded replicase [26].
Comparative genomic analysis can reveal key insights into the biology and evolution of viral species. EVE data are critically important components of these studies as they provide calibrations in geologic time. Unfortunately, making effective use of these data is challenging for a variety of reasons. This reflects a general lack of reproducibility and reusability in computational genomics [27], particularly where rapidly evolving and highly divergent sequences are involved [28, 29]. To address these issues we previously developed GLUE (Genes Linked by Underlying Evolution), a sequence data-centric bioinformatics environment computational genomics, with a focus on variation, evolution, and sequence interpretation [30]. Here, we used GLUE to create ‘Parvovirus-GLUE’, an extensible, open resource for comparative genomic analysis of parvovirus and EPV sequence data. We catalogue hundreds of EPV sequences in published whole genome sequence (WGS) data using a standardized nomenclature system to systematize the parvovirus fossil record. We capture these data in Parvovirus-GLUE and use reproducible approaches to examine their genomic and phylogenetic characteristics, revealing new insights into parvovirus ecology and evolution.
RESULTS
Creation of open resources for reproducible genomic analysis of parvoviruses
Comparative genomic analyses generally entail the construction of complex data sets comprising molecular sequence data linked to other kinds of information. Usually these include genome feature annotations and multiple sequence alignments (MSAs) as well as other diverse kinds of data. We used GLUE to create Parvovirus-GLUE [31], an open accessible online resource for comparative genomic analysis of parvoviruses and EPVs that preserves the ‘state’ of our data so that our analyses can be precisely and widely replicated (Fig. S1a-b). Furthermore, hosting of the Parvovirus-GLUE project in an openly accessible online version control system (GitHub) provides a platform for ongoing development of this resource by multiple collaborators, following practices established in the software industry (Fig. S1c).
The Parvovirus-GLUE project incorporates all of the data items required for broad comparative sequence analysis of parvoviruses, including: (i) a set of reference sequences representing all known parvovirus species (Supplementary); (ii) sequence and isolate-specific information (e.g. host species, vector species) in tabular form; (iii) a standardized set of parvovirus genome features and their coordinates within selected reference genome sequences; (iv) a set of MSAs incorporating all sequences in the project. Loading the project into the GLUE ‘engine’ generates a relational database that not only contains the data items associated with our analysis, but also represents the complex semantic links between them (Fig. S1a). Reproducible comparative genomic analysis of parvoviruses can be implemented by using GLUE’s command layer to coordinate interactions between the Parvovirus-GLUE database and bioinformatics software tools. The resource can be installed on all commonly-used computing platforms, and is also fully containerised via Docker [32]. It can be used as a local, stand-alone tool or as a robust foundation for the development of genome analysis-based reporting tools for potential use in human and animal health (Fig. S1b) - e.g. see HCV-GLUE [30], RABV-GLUE [33].
Systematic recovery of the parvovirus ‘fossil record’
We screened in silico WGS data of 738 vertebrate species and recovered a total of 595 EPV sequences. EPV sequences ranged from near full-length virus genomes to fragments ~150-300 nucleotides (nt) in length (Fig. 1). We classified EPVs into taxonomic groups based on their phylogenetic relationships to contemporary viruses (Fig. 2). Among EPVs that were >200nt in length, the majority (192/198=0.97%) were either unambiguous members of contemporary genera or members of groups that emerge as sister clades to these genera. We used comparative approaches to resolve these sequences into sets of orthologs, revealing that they represent at least 198 distinct germline incorporation events (Table 1, Tables S1-S6, Supplementary). As part of these efforts, we investigated the genes flanking each putatively novel locus. We identified common sets of genes flanking most EPV loci (Fig. 2). We applied unique identifiers to all EPVs identified in our study using a systematic nomenclature system that was originally developed for endogenous retroviruses (ERVs) but has more recently been applied to EVEs [19, 20]. This nomenclature captures information about orthology, enforcing a higher level of order on our data set by unambiguously associating EPV sequences with genomic loci. The semantic links between these data items are recorded in the Parvovirus-GLUE database, so that they are available for retrieval and manipulation in computational analyses.
(a) Protoparvovirus-derived EPV loci shown relative to the canine parvovirus (CPV) genome; (b) Dependoparvovirus-derived EPVs loci shown relative to the adeno-associated virus 2 (AAV-2) genome; (c) EPV loci derived from Amdoparvovirus-like viruses shown relative to the Aleutian mink disease (AMDV) genome; (d) Erythroparvovirus-derived loci shown relative to the parvovirus B19 genome; (e) EPVs derived from unclassified parvoviruses shown relative to a generic parvovirus genome. (f) Icthamaparvovirus-derived loci shown relative to Syngnathus scovelli parvovirus (SscPV); EPV locus identifiers are shown on the left. Solid bars to the right of each EPV set show taxonomic subgroupings below genus level. Where numbers are shown to the immediate right, the sequence shown is a consensus and numbers indicate how many individual orthologs sequences were used to create the consensus. Boxes bounding EPV elements indicate either (i) the presence of an identified gene (see Tables S1-S6), (ii) an uncharacterised genomic flanking region, or (iii) a truncated contig sequence (see key). EPV locus identifiers use six letter abbreviations to indicate host species (Table S8). Abbreviations: NS = non-structural protein; VP = capsid protein; ORF = open reading frame. ITR=Inverted terminal repeat; PLA2 = phospholipase A2 motif.
A maximum likelihood phylogeny showing the reconstructed evolutionary relationships between contemporary parvoviruses and the ancient parvovirus species represented by endogenous parvoviral elements (EPVs). Panels (A) and (B) show a more detailed view of subclades (labelled I and II) within the phylogeny shown in panel (C). The complete phylogeny, which is midpoint rooted for display purposes, was reconstructed using a multiple sequence alignment spanning 270 amino acid residues positions of the Rep protein and the LG likelihood substitution model. Coloured brackets indicate the established parvovirus genera recognised by the International Committee for the Taxonomy of Viruses. Bootstrap support values (1000 replicates) are shown for deeper internal nodes only. Scale bars show evolutionary distance in substitutions per site. Taxa labels are coloured based on taxonomic grouping as indicated by brackets, unclassified taxa are shown in black. Viral taxa are shown in bold, while EPV taxa are show in regular text. Abbreviations: PV=Parvovirus; HHV=Human herpesvirus; AAV=Adeno-associated virus; AMDV=Aleutian mink disease; CPV=canine parvovirus; BPV=bovine parvovirus; BrdPV=Bearded dragon parvovirus; MdPV=Muscovy duck parvovovirus; SlPV=slow loris parvovirus. TS=transcription strategy; MTSP=Multiple transcriptional start positions;STSP+=single transcription start position plus additional strategies; HOMO=homoteleomeric; HETERO=heteroteleomeric.
EPVs were identified in the genomes of all major groups of terrestrial vertebrates (Table 1) except agnathans, crocodiles, or amphibians. However, they occur much more frequently in mammals than in other groups. The majority of the EPVs identified in vertebrate WGS data derived from genera within the subfamily Parvovirinae, but we also identified rare examples of EPVs derived from subfamily Hamaparvovirinae (genus Icthamaparvovirus) (Table 2, Fig. S5). The identification of icthamaparvovirus-derived EPVs in snakes provides the first evidence that the host range of this viral genus extends to reptiles. Furthermore, orthologous copies of this EPV were identified in multiple snake species, providing a minimum age of 62 My for the Ichthamaparvovirus genus, and by extension the Hamaparvovirinae subfamily. Among Parvovirinae-derived EPVs, those derived from proto- and dependoparvoviruses predominate. Other genera (Erythyroparvovirus, Amdoparvovirus [34]) are also represented in the mammalian germline, but are relatively rare (<1% of species examined).
Previous studies have shown that some EPV loci express RNA with the potential to encode polypeptide gene products, either as unspliced viral RNA [12, 24, 25], or as fusion genes comprising RNA sequences derived from both host and viral sources [26]. We examined coding potential in EPVs and identified numerous sequences capable of encoding uninterrupted polypeptide sequences of 300 amino acids (aa) or more, with some ranging up to 722aa (Table S7). Furthermore, screening of RNA databases revealed evidence for expression of EPV RNA in several previously unreported EPVs) (Supp. Doc 1).
EPVs reveal the deep evolutionary origins of the subfamily Parvovirinae
We performed a comprehensive phylogenetic analysis of the Parvoviridae, using all available information, including EPVs. Phylogenies revealed three robustly supported sub-lineages within subfamily Parvovirinae, each encompassing multiple genera, as follows: (i) “Amdo-Proto”: Amdo- and Protoparvovirus; (ii) “Ave-Boca”: Ave- and Bocaparvovirus; (iii) “ETDC”: Erythro-, Tetra-, Dependo- and Copiparvovirus (Fig. 2).
Next, we compiled estimates of EPV integration dates (Table S1-S6) to provide an overview of parvovirus and vertebrate interactions over the past 100 My (Fig. 3). This revealed that germline incorporation of parvovirus DNA occurred throughout the Cenozoic Era in a broad range of vertebrate species (Fig. 3). Our results reveal that mammals acquired EPVs at a much higher frequency than other vertebrate groups (Table 2). In addition to the previously reported dependoparvovirus-derived elements in “whippomorphs” (cetaceans and hippopotamus) [13], lagomorphs, Old World rodents [7], New World rodents [24, 26], elephants [25], and macropoids [12] we identified numerous other ancient EPV loci diverse range of animal species (Table S1, Fig 3). Orthologous sets of EPV sequences demonstrating were also identified in passerine birds (order Passeriformes), establishing the ancestral presence of these viruses among ancestral members of clade Neoaves >85 Mya [35]. In addition, two ancient EPV loci were identified in snakes - an EPV in pit vipers provides a minimum age of 30 My for the Amdo-Proto lineage.
A time-calibrated evolutionary tree of vertebrate species examined in this study, illustrating the distribution of germline incorporation events over time. Colours indicate parvovirus genera as shown in the key. Diamonds on internal nodes indicate minimum age estimates for EPV loci endogenization (calculated for EPV loci found in >1 host species). Coloured circles adjacent to tree tips indicate the presence of EPVs in host taxa, with the diameter of the circle reflecting the number of EPVs identified. Brackets to the left show taxonomic groups within vertebrates.
The parvovirus family likely has extremely ancient origins, perhaps dating back to the origin of animal species [11], and the independent formation and fixation of EPVs in such a diverse range of vertebrate groups demonstrates that Parvovirinae genera circulated widely among vertebrate fauna throughout the Cenozoic Era. Furthermore, since we know that transmission between distantly related host groups is rare, we can tentatively estimate the age of sublineages within the Parvovirinae, based on the assumption that they reflect broad codivergence of viruses and hosts – at least at higher taxonomic levels. For example, the “Ave-Boca” lineage comprises clearly distinct avian (Ave-) and mammalian (Boca-) lineages, suggesting that ancestral members of this virus lineage circulated among the common ancestors of birds and mammals >300 Mya. Furthermore, we identified EPVs in the genomes of cartilaginous and ray-finned fish, suggesting that the subfamily Parvovirinae may be as old, if not older, than the vertebrate lineage itself (Fig. 2).
Interestingly, roseoloviruses (genus Roseolovirus) – the group of betaherpesviruses (subfamily Betaherpesvirinae) that includes human herpesvirus 6 (HHV6) – have acquired a homolog of the parvovirus rep gene, called U94, presumably when infection of the same host cell led to parvovirus DNA being incorporated the ‘germline’ of an ancient betaherpesvirus [36]. The presence of U94 in an orthologous position in rodent and bat betaherpesviruses, as well as within a betaherpesvirus-derived EVE in the tarsier genome [37], demonstrates that it arose through a horizontal gene transfer event that occurred ancestrally, likely before the divergence of (i) eutherian mammal orders and (ii) betaherpesvirus genera [38]. Phylogenetic reconstructions show that this gene derives from the EDTC sub-lineage within the subfamily Parvovirinae (Fig. 2).
Parvovirus genomes have palindromic ITR sequences at both the 3’ and 5’ ends which can fold back on themselves to form “hairpin” structures that are stabilized by intramolecular base-pairing. These “hairpin” structures are critical for genome replication in all parvoviruses, however, whereas they are heterotelomeric (asymmetrical) in some genera (Amdo-, Proto-, Boca-, and Aveparvovirus) they are homotelomeric (symmetrical) in others [39]. Interestingly, the distribution of this trait (where it has been described) across sub-lineages within the subfamily Parvovirinae suggests that - under the principles of maximum parsimony – the asymmetrical form (which is found across the “Amdo-Proto” and “Ave-Boca” sublineages) would be the ancestral form. Within the “ETDC” lineage, ITRs have only been described for the Dependoparvovirus and Erythroparvovirus genera, both of which have homotelomeric ITRs (Fig 2c), suggesting this that the presence of homotelomeric ITRs is a derived characteristic in subfamily Parvovirinae. Similarly, in all Parvovirinae groups except genus Amdoparvovirus, the N-terminal region of VP1 (the largest of the capsid) contains a phospholipase A2 (PLA2) enzymatic domain that becomes exposed at the particle surface during cell entry and is required for escape from the endosomal compartments. Phylogenetic reconstructions indicate that loss of PLA2 is an acquired characteristic of amdoparvoviruses (Fig. 2).
Another variable characteristic found in the Parvovirinae is the regulation of gene expression strategies, with members of the Proto- and Dependoparvovirus genera using two to three separate transcriptional promoters, whereas in the Amdo-, Erythro-, and Boca- genera all genes are expressed from a single promoter and genus-specific read-through mechanisms are used to produce alternative transcripts [2]. The presence of multiple separate promoters in the distantly related Proto- and Dependoparvovirus genera indicates that this expression strategy is probably ancestral, although the possibility that it evolved convergently in each lineage cannot be formally ruled out.
Mammalian vicariance shaped the evolution of protoparvoviruses
We identified 121 protoparvovirus-related EPV sequences in mammals, which we estimate to represent at least 105 distinct germline incorporation events (Table S1). Several near full-length genomes were identified, and many elements spanned >50% of the genome (Fig 1a). We reconstructed the evolutionary relationships between protoparvovirus-related EPVs and contemporary protoparvoviruses, revealing three major subclades within the Protoparvovirus genus, which we labelled “Archaeo-”, “Meso-” and “Neo-” protoparvovirus) (Fig 4a). The Archaeoprotoparvovirus (ApPV) clade is comprised exclusively of EPVs and is highly represented in the genomes of Australian marsupials (Australidelphia), American marsupials (Ameridelphia) and New World rodents. It includes numerous elements that are near full-length, but none encoding intact open reading frames (ORFs).
(A) Maximum likelihood-based phylogenetic reconstructions of evolutionary relationships between contemporary parvovirus species and the ancient parvovirus species represented by endogenous parvoviral elements (EPVs). The phylogeny was constructed from a multiple sequence alignment spanning 712 amino acid residues in the Rep protein (substitution model=LG likelihood). The tree is midpoint rooted for display purposes. Asterisks indicate nodes with bootstrap support >70% (1000 replicates). The scale bar shows evolutionary distance in substitutions per site. Coloured brackets to the right indicate (i) subgroups within the Protoparvovirus genus (outer set of brackets) and (ii) the host range of each subgroup (inner set of brackets). Terminal nodes are represented by squares (EPVs) and circles (viruses) and are coloured based on the biogeographic distribution of the host species in which they were identified. Coloured diamonds on internal nodes show the inferred ancestral distribution of parvovirus ancestors, using colours that reflect the patterns of continental drift and associated mammalian vicariance shown in the maps in panel (B). **Evidence for the presence of the “Mesoprotoparvovirus” group in Afrotherians is presented in Fig. 2. (B) Mollweide projection maps showing how patterns of continental drift from 200-35 led to periods of biogeographic isolation for terrestrial mammals in Laurasia (Europe and Asia), South America, Australia Africa and Madagascar. The resulting vicariance is thought have contributed to the diversification of mammals, reflected in the mammalian phylogeny as shown in in panel (c). The majority of placental mammals (including rodents, primates, ungulates and bats) evolved in Laurasia. However, these groups later expanded into other continents, and fossil evidence indicates that the ancestors of today’s “New World rodents” had arrived on the South American continent by~35 million years ago (Mya), if not earlier (C) A time-calibrated phylogeny of mammals with annotations indicating the biogeographic associations of the major taxonomic groups of contemporary mammals and ancestral mammalian groups, following panel (b) and key 1. (D) A time-calibrated phylogeny of mammals annotated to indicate the distribution of protoparvovirus subgroups among mammalian groups, following key 2. Question marks indicate where it is unknown whether viral counterparts of the lineages represented by EPVs still circulate among contemporary members of the host species groups in which they are found. Abbreviations: Mya = millions of years ago; NW=New World; NW=New World; (OW); CPV=carnivore parvovirus type 1; PPV=porcine parvovirus; HV=Hamster parvovirus; TuV=Tusavirus.
The Mesoprotoparvovirus (MpPV) clade is also comprised exclusively of EPVs, and was sparsely represented in the EPV fossil record, being detected in the Southern tamandua (Tamandua tetradactyla) – a xenarthran – and the aardvark (Orycteropus afer). The EPV locus found in aardvarks is relatively degraded, but the tamandua EPV sequence is nearly full-length and relatively intact (Table S7) (Fig. 1a). Finally, the Neoprotoparvovirus (NpPV) clade contains EPVs along with all known contemporary protoparvoviruses (Fig. 4a). Of the four NpPV-derived EPVs we identified here, three have been reported previously [9, 40], and all were identified in rodents. The novel representative was identified in the steppe mouse (Mus spicelagus) and comprises a near complete genome (Fig. 1). Notably, the VP gene of this element groups robustly with a bat-derived virus [41] in phylogenetic trees (Fig. S6a), whereas those encoded by other NpPV-derived EPVs group separately, in an entirely different subclade, together with VP sequences derived from carnivore and porcine protoparvoviruses. Notably, phylogenetic reconstructions show that none of the rodent EPVs in the NpPV clade groups with contemporary RoPVs, but instead cluster robustly with pPVs found in other mammalian host groups (e.g., carnivores, artiodactyls). This suggests that horizontal transfer from rodents to other mammalian orders may have been a common feature of parvovirus evolution. Tusavirus, a divergent protoparvovirus of uncertain host origin [42] groups basally in the NpPV clade (Fig. 4a), but could potentially represent an entirely distinct, under-sampled pPV lineage.
Continental drift over the past 150-200 My is widely accepted to have had a dramatic impact on mammalian evolution [43]. Around 200 Mya, all continents were part of an interconnected landmass (Pangaea) that later separated into two subcomponents (Fig. 4b). One (Laurasia) comprised Europe and most of Asia, while the second (Gondwanaland) comprised Africa, South America, Australia, India and Madagascar). Mammalian subpopulations were fragmented by these events, and then fragmented further as Gondwanaland separated into its component continents. The associated genetic isolation due to geographic separation (vicariance) drove the early diversification of major subgroups, including indigenous mammalian lineages in South America (xenarthans and marsupials), Australia (marsupials), and Africa (afrotherians). At points throughout the Cenozoic Era, placental mammal groups that evolved in Laurasia (Boreoeutherians) expanded into other continental regions. For example, the ancestors of contemporary New World rodents (which include capybaras, chinchillas, and guinea pigs among many other, highly diversified species), are thought to have reached the South American continent ~35 Mya [44].
The reconstructed evolutionary relationships between protoparvoviruses and protoparvovirus-derived EPVs strikingly reveal the impact of mammalian vicariance – and later migration – on the emergence and spread of novel protoparvovirus sublineages (Fig. 4). The protoparvovirus phylogeny can readily be mapped onto the phylogeny of mammalian host species so that the three major protoparvovirus lineages emerge in concert with major groups of mammalian hosts. These evolutionary relationships, which are supported by numerous, independently acquired EPV loci and ortholog sets (Table S1), are consistent with a parsimonious evolutionary scenario under which: (i) the ancestors of the contemporary protoparvovirus species were present in the ancient supercontinent of Pangaea prior to its breakup; (ii) the vicariance-driven, deep divergences in the mammalian phylogeny drove the emergence of distinct protoparvovirus lineages in distinct biogeographic regions throughout the course of the Cenozoic Era (from 65 Mya to present); (iii) the founder event associated with migration of rodents into the New World allowed this group to escape infection with NpPVs, but presumably, following their colonisation of the South American continent (estimated to have occurred ~50-30 Mya [44]), they were then exposed to infection with ApPVs, to the extent that numerous ApPV-derived EPVs were independently fixed in the germline. A previously reported ApPV-derived EPV in the common opossum (Monodelphis domestica) [10] groups intermediate between clades comprised of Australian marsupials EPVs and NW rodent EPVs, consistent with this hypothesis. Biogeographic analysis of host species distributions and ancestral range reconstruction support these findings (Fig. S6b).
Ancient origins and inter-order transmission of erythroparvoviruses
Our comprehensive in silico screen of vertebrate genomes identified the first reported examples of EPVs derived from genus Erythroparvovirus. One was identified in the genome of the Patagonian mara (Dolichotis patagonum) - a New World rodent – and another was identified in the genome of the Indri (Indri indri), a Malagasy primate. The mara element spans a complete NS gene, whereas the indri element encodes a complete viral genome with intact NS and VP genes and incorporating putative ITR sequences (Fig. 2), suggesting it integrated relatively recently. As reported in other erPVs, the viral protein 1 unique region (VP1u) of Erythro.1-Indri is relatively long. Neither contained obvious homologs of the accessory proteins reported in contemporary erPV genomes. Both erPV-derived EPVs grouped with erPVs derived isolated from rodents in phylogenetic trees, indicating inter-order transmission from rodents to lemuriforme primates (Fig. S7). Furthermore, when examined in relation to the biogeographic distribution of host species, these phylogenetic relationships provide tentative age calibrations for the Erythroparvovirus genus. based on the parsimonious assumption that the presence of the EPVs derived from rodent erPVs in Madagascar and South America reflects their spread into these isolated geographic regions during the Cenozoic Era (Table 3).
Inter-class transmission and the evolution of non-autonomous dependoparvoviruses
We identified 213 dependoparvovirus-related EPV sequences in mammals, which we estimate to represent at least 80 distinct germline incorporation events (Table S3). A small number of near full-length genomes were identified, but a large share of these elements spanned only small fragments (i.e. >40%) of the dependoparvovirus (dPV) genome (Fig 1b). We reconstructed the evolutionary relationships between dPV-related EPVs and contemporary dPVs (Fig 5, Fig. S9). In trees that included EPVs, support for internal branching order was typically quite low. This reflects the short length of many dPV-related EPV sequences, and the fact that many parts of the viral genome sequence are relatively degraded [13]. However, when only viruses and longer EPV sequences are included, phylogenies based on rep gene sequences disclose several robustly supported subclades within the Dependoparvovirus genus (Fig. 5). They include clades exclusive to reptilian species (Sauria-), Australian marsupials (Oceania-), and Boreoeutherian mammals (Neo-). A fourth clade, which we named “Shirdaldependoparvovirus” (ShdPV), contains dPV taxa derived from both avian and mammalian hosts. Both the composition of this clade in terms of hosts, and its phylogenetic position relative to other dPV groups, implies a role for interclass transmission between mammals and birds in dPV evolution (Fig. 5b). Firstly, the avian viruses in this clade group basally, forming a paraphyletic group relative to a derived subclade - here referred to as ‘Lemuria-’ - of ancient EPVs obtained from a diverse range of mammalian hosts. This topology suggests that clade Lemuria-may have originated via transfer from birds to mammals. Furthermore, in both midpoint-rooted phylogenies, and in phylogenies rooted on the saurian dependoparvoviruses (SdPVs) (as proposed by Penzes et al [45]), the ShdPVs as a whole fall intermediate between two exclusively mammalian groups – the neodependoparvoviruses (NdPVs) found in placental mammals, and the Oceaniadependoparvoviruses (OdPVs) found in Australian marsupials (Fig. 5a). This suggests the ShPVs originated via transmission from mammals to birds.
(A) A maximum likelihood phylogeny showing the reconstructed evolutionary relationships between contemporary dependoparvovirus species and the ancient dependoparvovirus species represented by EPVs. Virus taxa names are shown in bold, EPVs are shown in regular text. The phylogeny was constructed from a multiple sequence alignment spanning (MSA) 330 amino acid residues of the Rep protein and the LG likelihood substitution model and is rooted on the reptilian lineage as proposed by Penzes et al [45]. Brackets to the right indicate proposed taxonomic groupings. Shapes on leaf nodes indicate full-length EPVs and EPVs containing intact/expressed genes. Numbers next to leaf nodes indicate minimum age calibrations for EPV orthologs. Shapes on branches and internal nodes indicate different kinds of minimum age estimates for parvovirus lineages, as shown in the key. Numbers adjacent to node shapes show minimum age estimates in millions of years before present. For taxa that are not associated with mammals, organism silhouettes indicate species associations, following the key. The scale bar shows evolutionary distance in substitutions per site. Asterisks in circles indicate nodes with bootstrap support >70% (1000 replicates) in the tree shown. Plain asterisks next to internal nodes indicate nodes that are not supported in the tree shown here but do have bootstrap support >70% (1000 replicates) in phylogenies based on longer MSA partitions within Rep (but including less taxa). *Age calibrations based on data obtained in separated publications – see references [13] and [25]. **A contemporary virus derived from the marsupial clade has been reported in marsupials, but only transcriptome-based evidence is available [12]. (B) A time-calibrated phylogeny of vertebrate lineages showing proposed patterns of inter-class transmission in the Shirdaldependoparvovirus lineage. Abbreviations: PV=Parvovirus; AAV=Adeno-associated virus; BrdPV=Bearded dragon parvovirus; MdPV=Muscovy duck parvovovirus.
The NdPVs include the non-autonomous parvoviruses (AAVs), which require a helper virus for replication typically a nuclear DNA virus (e.g. herpesvirus, adenovirus [1]). In phylogenies rooted on the RdPVs, the NdPVs emerge as a derived clade with the autonomously replicating avian viruses grouping basal. Fragmentary EPVs found in Cercopithecine primate genomes arose between 23-16 Mya and appear to represent the ancient progenitors of contemporary primate AAVs (Fig. 5a).
DISCUSSION
In this study we recovered the complete repertoire of EPV sequences in WGS data representing 738 vertebrate species. While previous studies have reported a sampling of EPV diversity in vertebrates [7–13, 24, 26, 34, 35, 40, 46], the present study is an order of magnitude larger in scale – we identify 595 sequences representing nearly 200 discrete germline incorporation events (Table S1, Fig 1). Furthermore, we introduced a higher level of order to these data by: (i) discriminating between unique loci and orthologous copies; (ii) hierarchically arranging MSAs so that phylogenetic analysis (and taxonomic classification) of individual EPVs could utilise the maximum amount of available data; (iii) applying to EPVs a standardised nomenclature that captures information about orthology and taxonomy; (iv) inferring ancestral reference sequences for EPV coding domains.
The EPVs reported here are derived from a diverse array of distinct parvovirus groups (Fig. 2). The majority grouped within subfamily Parvovirinae, but we also identify rare examples of EPVs derived from Icthamaparvovirus (a genus in subfamily Hamaparvovirinae) in snakes. Orthologous copies of this element demonstrate that the association between hamaparvoviruses and vertebrates extends to the late Mesozoic Era >100 Mya, reinforces the view that this recently described subfamily is ancient and broadly distributed [46, 47]. Among EPVs derived from subfamily Parvovirinae, the majority derived from two genera – Protoparvovirus and Dependoparvovirus. However, we also identified representatives of other genera (Amdoparvovirus, Erythroparvovirus) as well as several highly divergent EPVs that likely represent novel genera. Among these sequences, those that were identified in mammals may simply represent mutationally degraded members of the established genera, since they are ancient and relatively short (Table 3, Fig 1f). However, those identified in basal vertebrate lineages such as cartilaginous fish (class Chondrichthyes) and lobe-finned fish (clade Sarcopterygii) are likely to represent novel groups. These EPV sequences also demonstrate that the host range of the subfamily Parvovirinae extends to basal vertebrates.
We obtained robust minimum age calibrations based on the identification of orthologous genomic flanking sequences for all parvovirus genera represented in the viral fossil record except Erythroparvovirus. However, for erythroparvoviruses (erPVs) and many other Parvovirinae genera (including those that are not represent in the molecular fossil record) we could infer more tentative calibrations based on the distribution of EPVs and viruses across host groups (Table 3). The most striking example of this occurs in the Protoparvovirus genus, in which the impact of biogeographic isolation throughout the Cenozoic Era is strongly reflected in the phylogenetic relationships between virus subgroups and the distribution of virus subgroups across host taxonomic groups and biogeographic host ranges. A simple, parsimonious explanation of these relationships is presented in Fig. 4, wherein ancestral protoparvoviruses (pPVs) were present in mammalian ancestors inhabiting Pangea ~200 Mya, and distinct pPV lineages emerged as mammalian species were compartmentalised into distinct biogeographic regions by continental drift. Later, the migration of mammalian groups into previously isolated continental regions provided the opportunity for these pPV subgroups to infect new host groups. Thus, the ancient AdPV lineage, which evolved primarily in marsupials spread into placental mammal group (New World rodents) during the later Cenozoic Era (see Fig. 4). This extended evolutionary timeline for pPVs is supported by evidence from orthology (Table 3), lending credibility to similar, biogeography and distribution-based age estimated inferred for viral lineages in which we did not obtain minimum age estimates based on orthologous EPVs (e.g., the Ave-Boca lineage).
Whereas some genera, such as Protoparvovirus and Dependoparvovirus, are highly represented in the genomic ‘fossil record’, others are conspicuously absent. For example, no EPVs derived from the ‘Ave-Boca’ lineage, or from the Tetraparvovirus and Copiparvovirus genera, were identified. However, the ancient calibrations obtained for dPVs and pPVs imply that other Parvovirinae genera have similarly ancient origins, and thus are consistent with the avian and mammalian components of the Ave-Boca lineage emerging via broad codivergence with vertebrate hosts ~400-300 Mya (Table 3). Extending this logic, the identification of Parvovirinae lineages in basal vertebrate lineages such as fish suggests that the subfamily Parvovirinae may have primordial origins within vertebrates.
While inter-class transmission of parvoviruses appears to be rare overall, we obtained compelling evidence that it has occurred in the Dependoparvovirus genus, specifically in the evolution of a lineage that contains parvoviruses and EPVs derived from both avian and mammalian hosts, and which we named “Shirdaldependoparvovirus” (ShDPV). This robustly supported clade contains both the avian autonomous dependoparvoviruses (dPVs) and the lemuriadependoparvoviruses (LdPVs) – a clade of mammalian dPVs that existed >80 million years ago (Table 3) and is so far only represented by EPVs. The topology of NS/Rep phylogenies cannot be reconciled with codivergence and instead implies that both ShDPV and the LdPV subclade it contains arose in separate inter-class transmission events involving mammals and birds (Fig. 5).
The non-autonomous dependoparvoviruses – often referred to as “adeno-associated viruses” (AAVs) – are characterised by the requirement for a helper virus to replicate. All of these viruses group within the ‘neodependoparvovirus’ (NdPV) clade in our analysis. Most recently described AAVs – such as those identified in bats, rodents and carnivores - have only been characterised at sequence-level, and little is known about their phenotypic properties. However, most of these viruses fall within the range of diversity defined by two AAV groups (Dependo-A and Dependo-B) indicating that the requirement for a helper virus (“dependency”) is an ancestral characteristic of the NdPV and likely to be shared among most if not all AAV species. Furthermore, the EPV fossil record supports the view that the host range of NdPVs encompasses all placental mammals. Dependency may have evolved as a means of timing replication - some large DNA viruses, such as herpesviruses, are able to establish latent, persistent infections within which they can ‘sense’ the cellular environment and switch to replicative mode when conditions are optimal [48]. The success of this strategy is reflected in the extremely high prevalence of herpesviruses in mammalian populations (often close to 100%). Possibly, NdPVs can optimise their transmission by tethering their replication cycle to that of these ubiquitous, sophisticated DNA viruses.
It seems extraordinary that so many EPVs have been fixed in the mammalian germline, since the formation of a novel EPVs is almost certainly a rare event - in utero virus infections are often lethal and virus-infected gamete cells are unlikely to be viable under most circumstances. Furthermore, the neutrality principle of population genetics predicts the loss of new alleles occurring at low frequency (unless there are selective advantages from the genotype). In most EPVs, ORFs have been disrupted by indels and contain multiple nonsense mutations rendering the ancestral viral ORFs non-translatable, but some retain long regions of intact coding sequence – both NS/Rep and VP/Capsid sequences are among the longest open regions (Table S7). The Rep protein is structurally and functionally related to the rolling circle replication (RCR) proteins that are among the oldest replicator proteins known [49, 50]. The RCR proteins play a pivotal role for replication of both circular and linear genomes, and therefore, are inextricably linked with single-stranded DNA viruses. With the exception of the mitochondrial DNA polymerase, RCR proteins are restricted to microbial and viral species. Experimental studies have shown that dPV Rep protein (over) expression affects healthy cells through a variety of activities including DNA binding, constitutive ATPase, and inhibiting the (cyclic) cAMP -activated protein kinase A (PKA) and protein kinase X (PrKX) [51, 52]. Rep-mediated inhibition of these kinases not only affects the infected cell, but also diminishes the proliferation of adenovirus helper virus, perhaps attenuating the virulence and virus-induced pathogenesis [53, 54]. Conceivably, it could be these properties that have favoured their capture by herpesviruses and by host species genomes. The selective forces that have favoured the retention of open VP/capsid genes in some EPVs (see Table S7) are unclear.
The extended evolutionary timescale implied by our analysis raises interesting questions about parvovirus evolution. For example, all members of the subfamily Parvovirinae use similar basic mechanisms to achieve specific steps in infection, but the specific details of these processes differ between genera. Our study suggests that these differences could have evolved gradually as distinct parvovirus lineages adapted to distinct ecological niches. It is clear from phylogenetic and genomic analysis that most vertebrate species are infected with multiple distinct parvovirus groups - for example, at least seven distinct genera circulate in mammals. Has each parvovirus genus developed specializations that allow it to occupy a unique ecological niche, or are some or all parvoviruses generalists? Other questions concern the current distribution of parvoviruses – e.g., to what extent does it reflect long-term evolutionary processes versus (possibly) more recent, anthropogenic influences? Also, which parvovirus groups currently only known via EPVs e.g. (Fig. 4, Fig. 5) are still prevalent, and which are extinct? Our study shows that parvovirus host-associations are relatively stable over time, implying that further sampling of parvovirus and EPV diversity will help address these questions.
EPVs allow the present-day biological properties of parvoviruses to be examined in the light of a time-calibrated evolutionary history. They can also be used to investigate the structural properties of ancient capsids based on molecular modelling [34] and even to reconstruct viable versions of capsids from extinct paleoviruses [40]. However, making effective use of EPV data is often challenging, since high levels of sequence divergence preclude straightforward analysis. In this study we introduced a template for computational genomics studies of viruses that focusses on facilitating the reproduction of comparative analyses and re-use of the complex datasets that underpin them (e.g., MSAs) (Fig. S1). This approach can not only scale to accommodate greatly increased quantities of virus species and sequences, but also introduces new levels of reproducibility and re-usability so that researchers working in different areas of parvovirus genomics - but utilizing related data and domain knowledge - can benefit from one another’s work. The resources and approaches developed in this study can thus facilitate the development of a broader understanding of parvovirus biology, covering multiple biological scales, which can be used to mitigate their harmful impacts and inform their development as therapeutic tools.
METHODS
Creation of resources for reproducible comparative analysis of parvovirus genomes
We used the GLUE software environment [30] to create a sequence-oriented resource for comparative genomic analysis of parvoviruses (including extinct paleoviruses). This resource, called ‘Parvovirus-GLUE’, not only contains the data items required for comparative analysis (i.e., virus genome sequences, multiple sequence alignments (MSAs), genome feature annotations, and other sequence-associated data), it also represents the semantic relationships between these data items via a relational database (Fig. S1). A library of parvovirus reference sequences was compiled from the virus reference genomes resource [55] and from recent papers describing novel parvoviruses and parvovirus-derived EVEs (identified via PubMed search). We obtained reference genome sequences for all Parvovirus species recognised by the International Committee for Taxonomy of Viruses (ICTV), as well as a representative set of recently identified parvovirus-related viruses that have not yet been incorporated into official taxonomy. The database is constructed using GLUE’s native command layer (Fig. S3a). The command layer can be used to interact with the database and with bioinformatics software tools required for comparative sequence analysis, thus establishing a platform for implementing comparative analyses in a reproducible, standardised way.
Parvovirus-GLUE incorporates MSAs representing distinct taxonomic levels within the family Parvoviridae and uses a ‘constrained alignment tree’ data structure to represent the hierarchical relationships between them (Fig. S2, Table 1). The root alignment contains all reference sequences, whereas all children of the root inherit at least one reference from their immediate parent. Thus, all alignments are linked to one another via our chosen set of references. This allows a single unified alignment to be used for phylogenetic analyses across distinct taxonomic levels and enables standardised sequence comparisons across the entire parvovirus family. A set of ‘master’ reference sequences - each representing a distinct clade in the parvovirus phylogeny – was defined. Reference sequences were used to define ‘constrained’ alignments (i.e., alignments in which the genomic coordinate spaces are constrained to the chosen reference sequence). For the lower taxonomic levels (i.e., genus and below) we aligned complete coding sequences. For the root we aligned a conserved region of NS protein consistent with the approach used by ICTV [1]. We used GLUE’s ‘constrained alignment tree’ data structure [30] to link MSAs constructed at distinct taxonomic levels, via a set of common reference sequences.
Genome screening in silico
We used the Database-Integrated Genome Screening (DIGS) tool [56] to derive a non-redundant database of EPV loci within published WGS assemblies. The DIGS tool is a Perl-based framework that uses the Basic Local Alignment Search Tool (BLAST) program suite [57] to perform similarity searches and the MySQL relational database management system to coordinate screening and record output data. A user-defined reference sequence library provides (i) a source of ‘probes’ for searching WGS data using the tBLASTn program [57, and (ii) a means of classifying DNA sequences recovered via screening Fig. S4. For the purposes of the present project, we collated a reference library composed of polypeptide sequences derived from representative parvovirus species and previously characterised EPVs. Whole genome sequence (WGS) data of animal species were obtained from the National Center for Biotechnology Information (NCBI) genome database [58]. We obtained all animal genomes available as of March 2020. We extended the core schema of the screening database to incorporate additional tables representing the taxonomic classifications of viruses, EPVs and host species included in our study. This allowed us to interrogate the database by filtering sequences based on properties such as similarity to reference sequences, taxonomy of the closest related reference sequence, and taxonomic distribution of related sequences across hosts. Using this approach, we categorised sequences into: (i) putatively novel EPV elements; (ii) orthologs of previously characterised EPVs (e.g., copies containing large indels); (iii) non-viral sequences that cross-matched to parvovirus probes (e.g., retrotransposons). Sequences that did not match to previously reported EPVs were further investigated by incorporating them into genus-level, genome-length MSAs (see Table 1) with representative parvovirus genomes and reconstructing maximum likelihood phylogenies using RAxML (version 8) [59].
Where phylogenetic analysis supported the existence of a novel EPV insertion, we also attempted to: (i) determine its genomic location relative to annotated genes in reference genomes; and (ii) identify and align EPV-host genome junctions and pre-integration insertion sites (see below). Where these investigations revealed new information (e.g., by confirming the presence of a previously uncharacterised EPV insertion) we updated our reference library accordingly. This in turn allowed us to reclassify putative EPV loci in our database and group sequences more accurately into categories. By iterating this procedure, we progressively resolved the majority of EPV sequences identified in our screen into groups of orthologous sequences derived from the same initial germline incorporation event (Table S1-S6).
We applied standard identifiers (IDs) to all EPV loci, following a convention established for endogenous retroviruses [60] that has more recently been applied to EVEs [19, 20]. Each EVE is assigned a unique identifier (ID) constructed from two components. The first component is the classifier ‘EPV’. The second component is itself a composite of two distinct subcomponents separated by a period; (i) the name of the lowest level taxonomic group (i.e., species, genus, subfamily, or other clade) into which the element can be confidently placed by phylogenetic analysis; (ii) a numeric ID that uniquely identifies the insertion.
Phylogenetic and Phylogeographic analysis
A process for reconstructing evolutionary relationships across the entire Parvoviridae was implemented using GLUE. We used a data structure called a ‘constrained MSA tree’ to coordinate genomic analyses across the large phylogenetic distances found in parvoviruses Fig. S7d. This approach addresses the issue that MSAs constructed at higher taxonomic levels (e.g., above genus-level in the Parvoviridae) typically contain far fewer columns than those constructed at low taxonomic levels (e.g., genus, subgenus), meaning that to fully investigate phylogenetic relationships using all available information it is often necessary to construct several separate MSAs each representing a distinct taxonomic grouping (e.g., see Table 2). The difficulties encountered in maintaining these MSAs in sync with one another while avoiding irreversible data loss are a key factor underlying the low levels of reproducibility and re-use in comparative genomic analyses [27], particularly those that examine more distant evolutionary relationships genomic [29]. To address these issues the ‘constrained MSA tree’ data structure represents the hierarchical links between MSAs constructed at distinct taxonomic levels, creating in effect a single, unified MSA that can be used to reconstruct both shallow and deep evolutionary relationships while making use of the maximum amount of available information at each level (Table 2, Fig. S1d). This approach also has the effect of standardising the genomic coordinate space to the constraining reference sequences selected for each MSA without imposing any limitations on which references are used (e.g., laboratory strains versus wild-type references), since additional, alternative constraining references can be incorporated, and contingencies such as insertions relative to the constraining reference are dealt with in a standardised way [30].
MSAs partitions derived from the constrained MSA tree were used as input for phylogenetic reconstructions. Nucleotide and protein phylogenies were reconstructed using maximum likelihood (ML) as implemented in RAxML (version 8.2.12) [59]. To handle coverage-related issues we generated gene coverage data prior to phylogenetic analysis and used this information to condition the way in which taxa are selected into MSA partitions. Protein substitution models were selected via hierarchical maximum likelihood ratio test using the PROTAUTOGAMMA option in RAxML. For multicopy EPV lineages we constructed MSAs and phylogenetic trees to confirm that branching relationships follow those of host species (Fig S4b, [31]).
Time-calibrated vertebrate phylogenies were obtained via TimeTree [61]. We used a time-calibrated phylogeny of protoparvovirus host species and the present continental distribution of host organisms to model ancestral biogeographical range of protoparvovirus host [62] (Fig. 6b). Country-level distribution information for each host species was obtained via the occ_search function of the rgbif library in R [63]. Country records were consolidated in continent entries with the continents function of the countrycode library and manually curated to ensure accuracy. Within continents, North Africa and Sub-Saharan Africa were considered distinct distributions and coded separately. The Dispersal-Extinction-Cladogenesis (DEC) model implemented in the program Lagrange C++ was applied without constraining the number of ancestral states nor limiting connectivity between biogeographic units [64]. Ancestral states at all nodes in the tree were inferred and the tree visualized in R with the ggtree and ggplot libraries [65, 66]. Input data and configuration files for Lagrange along with the time tree and Lagrange output are provided in the Data Supplement [31].
Genomic analysis of EPVs
ORFs were inferred by manual comparison of sequences to those of reference viruses. For phylogenetic analysis, the putative peptide sequences of EVEs (i.e., the virtually translated sequences of EVE ORFs, repaired to remove frameshifting indels) were aligned with polypeptide sequences encoded by reference genomes. We used PAL2NAL [67] to generate in-frame, DNA alignments of virus coding domains from alignments of polypeptide gene products. Phylogenies were reconstructed using maximum likelihood (ML) as implemented in RAxML [59] and GTR model of nucleotide selection as selected using the likelihood ratio test. The putative peptide sequences of EPVs were aligned with NS and VP polypeptides of representative exogenous parvoviruses using MUSCLE.
Expression and intactness of EPVs
We identified open coding regions of coding sequence in EPVs by using PERL scripts, (included with Parvovirus-GLUE [31]) to process EPV sequence data. To determine if there was evidence of expression of EPVs in host species, we searched the NCBI Reference RNA Sequences (refseq_rna) with Dependoparvovirus VP and Rep sequences (NC_002077). We searched a translated nucleotide query and a translated database using tBLASTx [57] and evaluated alignments found between refseq_rna sequences and Dependoparvovirus VP and Rep sequences. To further verify expression, we determined if the annotations were solely based on computational prediction or if there is RNAseq data alignment to the annotation in support of the feature. For those host species with evidence of expression, we conducted blastn searches within refseq_rna to identify expressed EPVs.
COMPETING INTERESTS
R.M.K. is a co-founder of Synteny Therapeutics, Inc., which is a co-assignee of a patent application filed on behalf of University of Massachusetts Medical School and Synteny Therapeutics, Inc.
ACKNOWLEDGEMENTS
This work was supported by funding from the Association Monégasque Contre les Myopathies, and the Bill & Melinda Gates Foundation (OPP1202116).
References
