Metatranscriptomic Analysis of Virus Diversity in Urban Wild Birds with Paretic Disease

Wildlife naturally harbor a diverse array of infectious microorganisms and can be a source of novel diseases in domestic animals and human populations. Using unbiased RNA sequencing, we identified highly diverse viruses in native birds from Australian urban environments presenting with paresis. This research included the clinical investigation and description of poorly understood recurring syndromes of unknown etiology: clenched claw syndrome and black and white bird disease. As well as identifying a range of potentially disease-causing viral pathogens, this study describes methods that can effectively and efficiently characterize emergent disease syndromes in free-ranging wildlife and promotes further surveillance for specific pathogens of potential conservation and zoonotic concern.

excluded from the analysis due to incongruity of histological lesions compared with those of all other birds (n ϭ 49) examined. The signalments, clinical signs, and histological lesions from these birds are summarized chronologically in Table S2.
Although BWBD was first recognized during an epizootic extending across Victoria, NSW, and into QLD in 2006, records from the Australian Registry of Wildlife Health identified the index cases as magpies and currawongs found in Sydney and the NSW central coast in August 2003. The majority of birds examined were magpies (n ϭ 26), currawongs (n ϭ 15), and ravens (n ϭ 6), with adults, juveniles, males, and females evenly represented. A solitary adult female figbird and one adult female magpie lark were also among the affected birds. Concurrent mortality was observed in crested pigeons (Ocyphaps olphotes), common koels (Eudynamys scolopacea), silver gulls (Larus novaehollandiae), and Indian mynahs (Acridotheres tristis). Presentation of affected birds varied, occurring as nine mass morbidity and mortality events and 25 sporadic cases. The birds examined from outbreaks in 2003,2006,2013, and 2015 represent a small fraction of affected birds and accumulatively total 105 animals reported formally and more than 250 reported informally. Suspected intoxication was a common history in those birds presenting en masse. Although seven birds died prior to veterinary examination, clinical signs among the remaining birds included paresis (Fig. 2a), recumbency or profound weakness (n ϭ 27), loss of righting reflex or difficulty righting (n ϭ 12), and, less commonly, dyspnea (n ϭ 4), watery or bloody diarrhea (n ϭ 4), clenched claws (n ϭ 1), and a moribund state (n ϭ 1). Although most birds were recumbent, they were not paralyzed, as they had good cloacal tone, withdrawal reflexes, and could stand and shuffle, flap, or walk across the room when stimulated. Most ravens had additional clinical signs indicative of central nervous system dysfunction, including ataxia, head tilt, circling, and unusual head posture (4/6). Other species were alert and could accurately grasp items with their beaks, despite being recumbent. The body condition of affected birds varied, with 21 in good body condition, while 27 were classified as thin, very thin, or emaciated.
Gross lesions in affected birds were rare, but included hydropericardium, epicardial hemorrhage, pulmonary congestion to hemorrhage, hemorrhage into the gastrointestinal lumen, and fibrinous serositis. Microscopic lesions characteristic of BWBD include the unifying feature of degeneration of cardiac myocytes with or without nonsuppurative perivascular and interstitial cardiac inflammation ( Fig. 2b to e). Myocyte degeneration or nonsuppurative inflammation within skeletal muscle was observed in 24 of the 45 birds examined (four animals were unavailable for examination). Nonsuppurative inflammation of the central nervous system was observed in 25 birds. Lesions were most severe in the central cerebellar white matter and included one-to four-cell-deep perivascular cuffs, spongiotic change of the neuropil, degeneration to necrosis of nerve cell bodies, and multifocal gliosis. Other common histological features of the syndrome included acute hepatic necrosis or inflammation (25/49), necrosis or nonsuppurative gastrointestinal inflammation (21/49), nonsuppurative interstitial nephritis (8/49), mild vasculopathy to marked fibrino-necrotizing vasculitis (19/49), nonsuppurative perivascular infiltrates throughout serosal surfaces (20/49), and nonsuppurative interstitial pancreatitis (11/49). Less common lesions included perivascular pulmonary, cardiac and neural hemorrhage, meningitis, and peripheral neuritis. Ravens generally had more severe and widespread central nervous system lesions and lacked skeletal muscle, renal, and serosal inflammation that was evident in the other species. In birds other than ravens, paresis was most likely associated with lesions in striated muscle and the peripheral nervous system. Although leucocytozoons and microfilaria are common hemoparasites of the bird species examined here, and were found in earlier metagenomic data generated from these cases (32), these organisms did not accord with the lesions present. Ancillary diagnostic testing for known bacterial and viral pathogens, Chlamydophila species, and intoxication revealed no significant findings.
Metatranscriptomic virus identification. Archived samples, including brain, liver, heart, and kidneys from representative cases of diseased birds, were pooled and used to generate nine RNA-sequencing libraries that resulted in 7,725,034 to 26,555,569 paired reads per pool. From these data, we discovered eight viruses from the RNA virus families Astroviridae, Paramyxoviridae, and Picornaviridae and the DNA virus families Adenoviridae, Circoviridae, Parvoviridae, and Polyomaviridae.
Avian avulavirus in brain samples of lorikeets with clenched claw syndrome. The Paramyxoviridae are a large group of enveloped linear negative-sense RNA viruses that range from 15.2 to 15.9 kb in length. We identified paramyxovirus-like contigs in the libraries of the brain from rainbow lorikeets. A reverse transcription (RT)-PCR designed to amplify the paramyxovirus L protein (301 nucleotides [nt]) identified matching RNA in 4 of 5 brain tissues, corresponding to the four birds presenting with neurological symptoms. Ten overlapping RT-PCRs were then performed to confirm the genomic sequence (Table S4), revealing a typical 3=-N(455 amino acids [aa])-P(446 aa)-M(366 aa)-F(619 aa)-HN(528 aa)-L(2,263 aa)-5= organization. Based on the phylogenetic analysis of the L and M proteins, the novel paramyxovirus identified here, termed avian avulavirus 5/Lorikeet/Australia/2016, fell into the clade comprising members of the genus Avulavirus, exhibiting 86.7% and 88.5% amino acid pairwise identity to its closest relative, avian paramyxovirus 5 (APMV-5)/budgerigar/Kuntachi/74 (GenBank accession no. NC_025361.1/YP_009094157).
To provide a provisional assessment of the virulence of this novel virus, we compared F protein cleavage sites between our lorikeet avulavirus and known pathotypes of NDV (avulavirus 1) that commonly causes disease in avian species, including the velogenic, lentogenic, mesogenic, and asymptomatic vaccine types. The precursor F glycoprotein (F0) is cleaved into F1 and F2 subunits. The F protein cleavage position for virulent or mesogenic strains contain a furin recognition site comprising multiple basic amino acids. Notably, the lorikeet avulavirus had an F cleavage RRRKKRF motif identical to pathogenic NDV strains, suggesting its potential virulence, although this remains to be confirmed (Fig. 3).
DNA viruses identified in sporadic cases or mortality events in rainbow lorikeets. We identified three DNA viruses in the context of sporadic mortality events of rainbow lorikeets: avian chapparvovirus, beak and feather disease virus, and lorikeet adenovirus. In addition to the RNA-sequencing results, we extracted DNA from corresponding tissues for DNA virus validation through PCR and rolling-circle amplification (RCA) assays.
Lorikeet chapparvovirus. The Parvoviridae are a family of small, nonenveloped, dsDNA animal viruses with linear genomes of ϳ5 kb in length. We identified parvoviruslike transcripts in both transcriptome sequencing (RNA-seq) libraries (RBL-L-2 and RBL-L-3) of the liver from diseased lorikeets (Table 1). Using RCA to enrich for circular DNAs, we recovered the complete genome of a novel lorikeet chapparvovirus, comprising 4,271 nt with two distinct open reading frames (ORFs) that encoded the nonstructural protein NS1 (670 aa) and the structural protein VP (542 aa). We designed specific primers to amplify the targeted VP region (ϳ248 nt) for screening both the RNA (positive in two liver samples) and DNA (all positive) products. In addition, we inferred two separate phylogenetic trees based on the complete NS1 and VP proteins to determine the evolutionary relationships between the lorikeet virus and other parvoviruses. These revealed that the closest relative to the novel lorikeet chapparvovirus identified here was an avian-associated red-crown crane parvovirus, yc-9 (GenBank accession no. KY312548.1), although this shares only 48.2% amino acid identity in NS1 and 46.9% identity in VP (Fig. 4).
Beak and feather disease virus. Beak and feather disease virus (BFDV), a member of Circoviridae, is highly prevalent in Australian wild birds, particularly psittacine species (33). We identified BFDV-like contigs in all lorikeet RNA libraries, sharing 95% nucleotide identity to known strains of BFDV. Remapping the raw reads to the closest BFDV strain (GenBank accession no. KM887928) gave coverage of the whole virus genome, comprising 2,014 nt and encoding a replication-associated protein (290 aa) and a capsid protein (247 aa) (Fig. 5A). We then screened all the RNA samples from rainbow lorikeets using PCR primers targeting the capsid protein (596 nt) (GenBank accession no. KM887928), with the PCR products then Sanger sequenced. The BFDVs identified from individual birds carried distinctive single nucleotide variants, and sequences from each case formed strong phylogenetic clusters, suggesting that these results are not due to cross-sample contamination (Fig. 5B). Lorikeet adenovirus. Adenovirus-like contigs were identified from libraries of both the brain and liver of rainbow lorikeets. Phylogenetic analyses were performed based on the products of PCR primers designed based on the obtained sequences of polymerase (Pol, 843 aa) and hexon (580 aa) proteins. The virus identified, termed lorikeet adenovirus, exhibited 76% amino acid identity to the most closely related skua adenovirus (GenBank accession no. NC_016437.1) (Fig. 6).
A novel lorikeet hepatovirus in lorikeets with clenched claw. A complete hepatovirus-like (Picornavirales) contig was identified in one of the libraries of the liver from the diseased lorikeets, exhibiting relatively high read abundance (ϳ10,000-fold coverage depth). The genome of this novel lorikeet hepatovirus, termed Garigal virus, is 7,339 nt in length and encodes a polyprotein of 2,070 amino acids, excluding poly(A) tails, that is similar in structure to most avihepatoviruses (Fig. 6). Interestingly, Garigal virus was most closely related to Hepatovirus sp. isolate HepV-bat3206/Hipposideros_ armiger/2011 identified in a round-leaf bat, although these two viruses exhibit only 46.3% amino acid identity across the polyprotein. Recently, a lorikeet picornavirus, LoPV-1 (GenBank accession no. MK443503), was identified in fecal samples from rainbow lorikeets in China (34), although this exhibited only 17.4% amino acid identity to Garigal virus (Fig. 7).
Novel picornavirus identified in black and white bird disease cases. The Picornaviridae are a large family of single-strand positive-sense RNA viruses with genomes ranging from 6.7 to 10.1 kb in length. We identified a highly abundant complete picornavirus-like contig in the library of archived BWBD cases. Remapping raw reads to the picornavirus-like transcript showed complete genome coverage (ϳ300-fold depth). We named this novel black and white bird picornavirus Cowan virus. A RT-PCR targeting this picornaviral contig (ϳ220 bp) was positive in 2/6 of the brain tissues tested from  an Australian magpie (Cracticus tibicen) and an Australian raven (Corvus coronoides).
Notably, the two positive birds were those presenting with nonsuppurative encephalomyelitis. Cowan virus exhibits classical picornavirus features, with a genome length of 7,011 nt excluding the poly(A) tails. This genome encodes a polyprotein of 2,070 amino acid residues, as well as a 5= untranslated region (UTR) of 508 nt that includes the putative internal ribosome entry site (IRES). The genome organization of Cowan virus is similar to that of other picornaviruses, with the following characteristic gene order: 5=-VP4, VP2, VP3, VP1, 2A, 2B, 2C, 3A, 3B, 3Cpro, 3Dpol-3= (Fig. 7). Cowan virus also exhibited such typical picornavirus features as rhv-like domains (aa 111 to 222 and aa 307 to 424), an RNA helicase (aa 1172 to 1271), and an RNA-dependent RNA polymerase (RdRp; aa 2903 to 2154). Phylogenetic analysis revealed that Cowan virus was most closely related to seal picornavirus (GenBank accession no. NC_009891.1), a member of the genus Aquamavirus (Fig. 7), although with very low amino acid similarity across the polyprotein: only 29.5% with Kunsagivirus A (GenBank accession no. YP_009505615.1) and 30.8% with seal picornavirus type 1 (GenBank accession no. YP_001497152.1). According to the International Committee on Taxonomy of Viruses (ICTV), members of a genus within the Picornaviridae should share at least 40% amino acid sequence identity in the polyprotein region. As such, Cowan virus may represent a new virus genus.
A divergent black and white bird astrovirus in a BWBD outbreak from Nowra, NSW. Several astrovirus-like reads were found in the RNA-seq library of the suspected BWBD (i.e., nonsuppurative encephalitis of undetermined etiology) outbreak from Nowra, NSW (Table 1). This novel virus, termed Nowra virus, was highly divergent in  sequence and contained only 25 reads that covered the partial RdRp of astrovirus Pygoscelis/DT/2012 (GenBank accession no. KM587711.1) (see below). Reads from Nowra virus were used to develop PCR assays to bridge the gap in the RdRp. Using RT-PCR amplification of a 195-nt region targeting the RdRp, we obtained positive virus hits in 4 of 5 brain samples. Based on RdRp protein identity and phylogenetic analysis, there was 69.8% amino acid sequence similarity between Nowra virus and its closest relative, astrovirus Pygoscelis/DT/2012 (GenBank accession no. KM587711) sampled from Adelie penguins (Pygoscelis adeliae) in Antarctica (Fig. 8).
Combining the PCR results with clinical and histopathology patterns, we suggest that Nowra virus might be the cause of the BWBD-suspected outbreak in Nowra (see Fig. 10B).  Novel magpie polyomavirus in the Nowra outbreak. In addition to the novel astrovirus, our RNA-seq analysis identified a novel circular avian polyomavirus in brain and heart tissue from one nondiseased Australian magpie from the Nowra outbreak. Of note, the histopathology of BWBD in this case was inconsistent with polyomavirus infection, which is not known to be associated with myositis or encephalitis in birds. The genome of the novel magpie polyomavirus, termed magpolyV, comprised 5,115 bp with an overall GC content of 46.5%, and we were able to recover the full-length genome of this virus using RCA and PCR assays. Based on sequence similarity with existing reference polyomaviruses, we predict that the genome of magpolyV encodes capsid proteins VP1, VP2, and VP3 and open reading frame X (ORF-X) from the late region, as well as two alternative transcripts, large T antigen (LT) and small T antigen (ST) (Fig. 9). Consistent with most known polyomaviruses, magpolyV retains the typical conserved motifs and splicing sites in LT, including HPDKGG (DnaJ domain), LRELL and LLGLL (LXXLL-CR1 motif), LFCDE (LXCXE, a pRB1-binding motif). and GAVPEVNLE (ATPase motif). However, the consensus sequence CXCXXC for protein phosphatase 2A binding, mostly found in mammalian polyomaviruses, was absent. Phylogenetic analysis revealed that magpolyV consistently clustered within avian lineages, exhibiting 90.3% amino acid similarity to its closest relative, butcherbird polyomavirus isolate AWH19840 (GenBank accession no. KF360862), in the LT protein and 92% amino acid similarity in VP1.

DISCUSSION
We characterized the clinical and histological features of CCS and BWBD and applied a metatranscriptomic approach to identify potential viral etiological agents in wild Australian birds presenting with hepatic inclusion bodies, paresis, and nonsuppurative myocarditis or encephalomyelitis. Through this combined approach, we were able to elucidate the complex nature of viral disease investigation amid multifactorial infections. In particular, metagenomic next-generation sequencing, in combination with traditional gross and histological examination, identified several candidate pathogens  for disease syndromes that had been reported as undiagnosed for decades. Application of this methodology in the face of emergent disease syndromes or mass mortality events will facilitate the early detection of wildlife viruses, in turn contributing to disease control and wildlife conservation.

CCS.
A key element of our paper was the investigation of clenched claw syndrome (CCS), a paretic disease of unknown etiology in rainbow lorikeets (Trichoglossus haematodus) and a single red-collared lorikeet (Trichoglossus rubritorquis) in Australia. The

Virome of Urban Wild Birds with Neurological Signs
Journal of Virology diseased lorikeets presented with paresis and were often found recumbent or sitting on their hocks with feet clenched, the passive position for the avian foot. Affected birds kept in care progressively developed head tilt, ataxia, and paralysis. The disease syndrome has occurred primarily along coastal New South Wales and Southeast Queensland since the early 1980s. Although previous studies report 5% to 10% of free-living rainbow lorikeets rescued annually presenting with this syndrome, it seems likely that the clinical syndrome in fact encompasses two or more disease etiologies (35). For example, lead poisoning and plant-based intoxication have been proposed as contributing to syndromes described as clenched claw or drunken lorikeet syndrome (36). We focused on 55 cases of CCS characterized by nonsuppurative encephalomyelitis and often ganglioneuritis in which a viral etiology was suspected. The metatranscriptomic results, in conjunction with the clinical signs, histopathologic data, and confirmation through PCR assays, suggested that an avulavirus (APMV-5) potentially played a leading role in CCS pathogenesis in these diseased birds (Fig. 10A). An increasing incidence of clinical APMV infection in wild birds has been reported worldwide (37). APMVs have been categorized into 12 serotypes, with APMV-1 (NDV) having a significant economic impact on the poultry industry and population-level impacts on wild birds (37). Both NDV and APMV-5 have been associated with disease outbreaks in which mortality rates approach 100%, although the pathogenicity of APMV-5 varies  GenBank accession numbers are given in the taxon labels. Trees were midpoint rooted for clarity only, and branch support was estimated using 1,000 bootstrap replicates, with those of Ͼ70% shown at the major nodes.

RdRp
substantially among aviary species. APMV-5 was first isolated from a fatal outbreak of caged budgerigars in Japan in 1974, followed by epizootic outbreaks in budgerigars in the United Kingdom (38) and Australia (39). Interestingly, APMV-5 does not replicate in chicken embryonated eggs, lacks a virion hemagglutinin, and is not pathogenic to chickens (40).
The gross and histological pathologies of avulaviruses vary depending on the viral pathotype and the susceptibility of the host. Although APMV-5 infection in budgerigars caused enteritis, APMV-3 in Australian Neophema species parrots has been associated with fatal neurological disease (41). The histological changes in rainbow lorikeets with CCS were most similar to neural lesions described for velogenic APMV-1 in poultry, free-ranging double crested cormorants (Phalacrocorax auritus), and pigeons (Columba livia domestica) (42)(43)(44)(45). Collectively, Newcastle disease and CCS are characterized by gliosis, neuronal necrosis, nonsuppurative perivascular infiltrates, and white matter vacuolation. Notably, CCS-affected birds were devoid of the lesions in the respiratory and gastrointestinal tracts, lymphoid tissues, pancreas, liver, and kidney that are  described for some APMVs and pigeon paramyxovirus (PPMV-1). Because of the possibility of cross-species transmission, the prevalence and pathogenicity of avian paramyxoviruses like APMV-5 circulating among wild birds in Australia clearly merit further investigation.
Detection of avian chapparvovirus, avian adenovirus, and beak and feather disease viruses in rainbow lorikeets. Chapparvoviruses are being increasingly identified in metagenomic studies and seemingly have a wide host range, including bats (46) and the feces of turkeys (47), red-crowned cranes (48), wild rats (49), and pigs (50). Importantly, a murine kidney parvovirus was recently found in immunodeficient laboratory mice with renal failure and kidney fibrosis (51), indicating that some chapparvoviruses result in highly virulent infections.
Avian polyomavirus, beak and feather disease virus, and avian adenovirus appear to be prevalent in psittacine species in Australia (52), compatible with the data presented here. As multiviral infection is common, our analysis of viral communities associated with neglected avian diseases enabled a comprehensive understanding beyond the one᎑host, one᎑virus system (13). Many known viral infections represent clinical issues in psittacine species due to their association with acute death, disease, the capacity to induce immunocompromise, and difficulties in treatment and control (53). These have been associated with a variety of RNA viruses, including avian paramyxovirus (37), avian influenza virus (54,55), and avian bornavirus (56), and with DNA viruses, such as psittacine beak and feather disease virus (PBFD) (57), avian polyomavirus (APV) (58), psittacid herpesvirus (PsHV) (59), psittacine adenovirus (PsAdV) (60), poxviruses (61), and papillomaviruses (62).

Role of astroviruses, polyomaviruses, and picornaviruses in black and white bird disease.
Viruses of the family Astroviridae infect multiple mammalian species, including humans, and also avian species and are most commonly associated with gastroenteritis and, rarely, neurological symptoms. A number of novel divergent astroviruses in wild birds have recently been discovered through metagenomic surveillance (63)(64)(65). Many avian astroviruses, including avian nephritis virus, duck and turkey astroviruses, and duck hepatitis virus 2, have considerable economic impact on the poultry industry, and these viruses are widely distributed in wild birds (66). Despite this, the diversity and ecology of these viruses, including their potential interspecies transmission events, are largely unknown. Previously, astrovirus-associated central nervous system impairment had been reported in mink, human, bovine, ovine, and swine (67)(68)(69)(70).
In mammals such as dogs and cats, astroviruses are commonly isolated with enteric bacterial pathogens in sporadic gastroenteritis outbreaks (71), although encephalitis has been described in cattle and humans (72). In contrast, avian astrovirus infection manifests as a broader disease spectrum, including enteritis, hepatitis, and nephritis, but rarely neurological symptoms. The novel Nowra virus identified here was present in each bird with nonsuppurative encephalomyelitis, concurrent with myositis, coelomitis, and myocarditis from a single outbreak. This suite of histological lesions is unusual for avian astroviruses. A considerable proportion of BWBD-affected animals also had enteric and hepatic lesions similar to those described in astrovirus-infected poultry but also commonly associated with a variety of other viral pathogens. Further surveillance and in situ diagnostic modalities are required to reveal the association between Nowra virus, the clinical illness described, and the histological lesions observed.
A novel polyomavirus, magpolyV, was identified from the brain and heart of one of the nondiseased Australian magpies (Cracticus tibicen) that died in the same BWBD outbreak. An avian polyomavirus (APV), namely budgerigar fledgling disease polyomavirus, has been documented in young budgerigars and other psittacine species, causing feather abnormalities, abdominal distension, head tremors, and skin hemorrhages, reaching infection rates of up to 100% in aviaries worldwide (73). The clinical presentation and degree of susceptibility to APV infection range from skin diseases to acute death. In the case of nonpsittacine species, goose polyomavirus has been characterized as the etiologic agent of fatal hemorrhagic nephritis and enteritis of European geese (HNEG) (74). Other species of polyomaviruses have been identified in finches (75), crows (76), and butcherbirds (77), and a novel APV was recently associated with a fatal outbreak in canaries (78). Moreover, a recent study revealed an APV isolated from pigeon feces in China showing almost 99% identity with previously identified psittacine strains, suggesting a much broader host range of APVs and undetermined cross-species transmission (79).
Picornaviruses are commonly identified infecting a variety of avian species in metagenomic studies. In the context of overt disease, notable avian picornaviruses include avian encephalomyelitis viruses in chickens, pheasants, and turkeys (80), duck hepatitis A in ducklings (81), and a novel poecivirus that has been identified strongly associated with avian keratin disorder (AKD) in Alaskan birds (82). We identified two novel picornaviruses in this study, Garigal virus and Cowan virus, the pathogenic potentials of which merit further investigation. Interestingly, a novel lorikeet picornavirus (LoPv-1; GenBank accession no. MK443503) was identified in the feces of healthy rainbow lorikeets (Trichoglossus haematodus) in China (34), although the polyprotein shared only 12.6% similarity with the lorikeet hepatovirus identified here.
The high frequency of microbial coinfections suggests that rather than being the primary cause of the disease outbreak, some novel avian viruses might augment another pathogen, causing immunosuppression in immunocompromised hosts (83). Although we were successful in elucidating seven candidate viral pathogens that previously eluded detection, it is clear that additional experimental evidence such as in vivo inoculation experiments and in situ methodologies to identify pathogens within lesions are critical for clarifying the relationship between infection and the emergent disease syndromes we consider. Notably, the metagenomic approach utilized here is not limited to viruses and can be useful in the detection of bacteria, fungi, and protozoan and metazoan parasites (32). Previous studies have shown that the coinfection of pathogens like APMV-1 and avian influenza virus may greatly complicate disease outbreaks, for example, by modulating host immune responses, impacting the prevalence of other viral infections (84), and masking the presence of other pathogens (85). Thus, broader virome-scale studies such as that undertaken here provide a greater opportunity to understand virus-virus and virus-host interactions in more detail.
Due to the limited collection from natural outbreaks, more extensive sampling in the future will clearly help produce better case definitions and disease associations. Admittedly, the retrospective investigation may not have identified every potential case, particularly as viral infections may present differently in various avian species. Despite the relatively small number of disease cases sampled here, the utility of metatranscriptomics in the context of a thorough and multidisciplinary diagnostic approach is that it is able to identify viruses and other microbial species, including those that are highly divergent, with greater efficiency and scope than conventional diagnostic tests alone. The sequencing data and PCR assays developed here enhance diagnostic capabilities for avian disease and enable epidemiological surveillance to better understand the ecology and impacts of the viruses described. The investigation focused on identifying potential pathogens within specimens emanating from undiagnosed, but welldescribed clinical conditions in birds. These clinical syndromes and associated distinctive pathological changes have been previously described, enabling us to generate search terms to interrogate the archived collections. More broadly, our study highlights the value of proactive viral discovery in wildlife and targeted surveillance in response to emerging infectious disease events that might be associated with veterinary and public health. Sample collection. Samples were collected between 2002 and 2013 from 18 birds, predominantly from within the Sydney basin and the coastal center of NSW. Fresh portions of brain, liver, heart, and kidney were collected aseptically and frozen at -80°C. Additionally, a range of tissues were fixed in 10% neutral buffered formalin, processed in ethanol, embedded with paraffin, sectioned, stained with hematoxylin and eosin, and mounted with a cover slip prior to examination by light microscopy. Giemsa stains were applied to a subset of tissue samples to determine whether protozoa were present within lesions.

MATERIALS AND METHODS
Historical case review. The records of the Australian Registry of Wildlife Health dating back to 1981 were interrogated to identify rainbow lorikeets with clinical signs relating to the nervous system and those with unexplained nonsuppurative inflammation in the central nervous system. An additional search was conducted to identify passerine birds with unexplained nonsuppurative inflammation within cardiac or skeletal muscle. Retrieved records were investigated to determine the signalment, clinical signs, and histological lesions of affected animals.
The severity of nonsuppurative inflammation was graded on a scale of 0 to 4, where 0 indicated no discernible lesions and 4 represented severe and extensive mononuclear cell inflammation. Degeneration and necrosis were graded on a scale of 0 to 4, where 1 represented mild, multifocal single-cell degeneration or necrosis and 4 characterized extensive coagulative necrosis or malacia. Wallerian degeneration of the white matter tracts within the brain, spinal cord, and central and peripheral nerves was noted when present. In rainbow lorikeets, the lesions were graded in the cerebrum, brain stem, cerebellum, spinal cord, ganglia, and peripheral nerves. In the passerines, degeneration or necrosis and nonsuppurative inflammation were graded in skeletal and cardiac muscle, brain, liver, gastrointestinal tract, pancreas, blood vessel walls, and renal interstitium.
Avian influenza and NDV were excluded using real-time PCR on oropharyngeal, conjunctival, and cloacal swabs from 22 magpies and 12 other wild birds (including currawongs, magpies, and ravens). Serological testing of plasma from the same birds comprised hemagglutination inhibition to identify NDV antibodies and competitive ELISAs targeting avian influenza virus and flavivirus antibodies. Sixty-eight tissue samples from a subgroup of 13 magpies, currawongs, and ravens identified as having acute histological lesions suggestive of viral infection were subjected to further avian influenza virus-, NDV-, and WNV-specific RT-PCR and virus isolation. Virus isolation was attempted in both mammalian and insect cell lines (Peter Kirkland, personal communication.). Additional viral culture was attempted on the same samples inoculated into chicken embryos while assessing hemagglutinating agents and in Vero cells assessing cytopathic effect consistent with WNV infection. Clearview antigen capture ELISA (Unipath, Mountain View, CA) targeting Chlamydophila species was conducted on splenic tissue from eight birds with BWBD (magpies, currawongs, a raven, and a magpie lark). Liver and lung tissues from eight magpies and currawongs were subjected to routine aerobic and anaerobic bacterial and fungal culture. Finally, liver tissue from eight magpies and currawongs, representing the 2006 and 2015 BWBD epizootics, were subjected to toxicological testing to exclude the presence of a variety of acaricides, fungicides, organophosphates, carbamates, synthetic pyrethroids, and organochlorines and their metabolites.
RNA extraction, library construction, and sequencing. Viral RNA was extracted from the brain and liver, heart, and kidney samples of animals using the RNeasy Plus minikit (Qiagen, Germany). RNA concentration and integrity were determined using a NanoDrop spectrophotometer (Thermo) and TapeStation (Agilent). RNA samples were then pooled in equal proportions based on animal tissue type and syndrome. Illumina RNA libraries were prepared on the pooled samples following rRNA depletion using a RiboZero Gold kit (Epidemiology) at the Australian Genome Research Facility (AGRF), Melbourne, Australia. The rRNA depleted libraries were then sequenced on an Illumina HiSeq 2500 system (paired 100-nt reads).
Virome metatranscriptomics. Unbiased sequencing of RNA aliquots extracted from diseased animals in each outbreak were pooled based on host species, clinical syndrome, and histological findings as shown in Table 1. The RNA sequencing reads were trimmed of low-quality bases and any adapter sequences before de novo assembly using Trinity 2.1.1 (86). The assembled sequence contigs were annotated using both nucleotide and protein BLAST searches against the NCBI nonredundant sequence database. To identify low-abundance organisms, the sequence reads were also annotated directly using a BLASTX search against the NCBI RefSeq viral protein database using Diamond (87) with an E value cutoff of Ͻ10 Ϫ5 . Open reading frames were then predicted from the viral contigs in Geneious v11.1.2 (88), with gene annotation and functional predictions made against the Conserved Domain Database (CDD) (89). All sequences were aligned using the E-INS-i algorithm in MAFFT version 7 (90). Virus abundance was assessed using a read mapping approach in which reads were mapped to the assembled viral contigs using the BBmap program (91). Viruses were named in accordance with the ICTV guidance for each group in question.
RCA assays for circular DNA viruses. To recover the complete genomes of the polyomaviruses and circoviruses identified through our RNA-seq analysis, we enriched for circular DNA using rolling-circle amplification (RCA) (92). Briefly, genomic DNA was extracted from animal tissue and combined with a reaction buffer containing random primers, 1ϫ phi29 buffer, dithiothreitol (DTT), bovine serum albumin, deoxynucleoside triphosphates (dNTPs), and phi29 DNA polymerase (Thermo Scientific, Australia) and then incubated at 30°C for 16 h. The RCA products were then purified using the Monarch PCR & DNA cleanup kit (NEB) and then quantified using the Qubit dsDNA broad-range assay. Nextera XT DNA libraries were then prepared and sequenced on an Illumina MiSeq to a depth of ϳ2 million reads (2 ϫ 150 nt).
RT-PCR assays and Sanger sequencing. Total liver and brain RNA from individual birds was reverse-transcribed with SuperScript IV VILO mastermix (Invitrogen). The cDNA generated from the sampled tissues was used for virus-specific PCRs targeting regions identified by RNA-seq. RT-PCR primers were designed to bridge any genomic gaps based on detected transcripts of the paramyxoviruses, polyomaviruses, adenoviruses, and astroviruses identified in the RNA-seq data (see Tables S3 and S4 in  the supplemental material). Accordingly, all PCRs were performed using Platinum SuperFi DNA polymerase (Invitrogen) with a final concentration of 0.2 M for both forward and reverse primers. All PCR products were visualized by agarose gel electrophoresis and Sanger sequencing. Long RT-PCR products were also sequenced using Nextera XT and the MiSeq platform, per the RCA assays described above.
Phylogenetic analysis. Conserved protein domains were used to determine the evolutionary relationships of the viruses identified here. In the case of RNA viruses, we used amino acid sequences of the RNA-dependent RNA polymerase (RdRp) that is the most conserved protein among this group. After removing all ambiguously aligned regions using TrimAl (93), phylogenetic trees were inferred using the maximum likelihood method (ML) implemented in PhyML version 3.0 (94), employing a Subtree Pruning and Regrafting topology searching algorithm and the LG model of amino acid substitution. Bootstrap resampling with 1,000 replications under the same substitution model was used to assess nodal support. All phylogenetic trees were then visualized using FigTree v1.4.3 (http://tree.bio.ed.ac.uk/software/ figtree).
Data availability. The RNA sequencing data in this study have been deposited in the NCBI Sequence Read Archive under BioProject accession no. PRJNA631876 and PRJNA626677. All genome sequences of the viruses identified have been uploaded to GenBank under accession no. MT457853 to MT457860.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 0.3 MB.