REPEAT SPILLOVER OF BEAK AND FEATHER DISEASE VIRUS INTO AN ENDANGERED PARROT HIGHLIGHTS THE RISK ASSOCIATED WITH ENDEMIC PATHOGEN LOSS IN ENDANGERED SPECIES

In disease ecology, emerging pathogens have received attention as theoretical causes of species decline, extinction, or host endangerment, but epidemiological theory suggests that highly host-adapted pathogens are unlikely to extirpate a species since disease fadeout becomes more likely as the size of the susceptible host population declines (Smith et al. 2006; Frick et al. 2010; Langwig et al. 2016). However, this same principle infers that once a host population falls below the minimum population size capable of maintaining an endemic infection, it becomes susceptible to repeated spillover of related pathogens from reservoirs, which are potentially closely related more abundant host species (Pedersen et al. 2007; Pedersen and Babayan 2011).

Under conditions where multiple host species are equally available, a balance between forces of host–virus codivergence and spillover infection should result from episodes of selection and counterselection (Dennehy et al. 2007; Shin and MacCarthy 2016). Selective forces at play within the host (e.g., to improve resistance or better cope with infection) as well as the pathogen (e.g., to attenuate its effect) also drive the evolution of innate and adaptive immune responses expressed dynamically as protective population-based herd or flock immunity. Host radiation may also allow a virus to expand its ecological niche; however, a virus's ability to infect a diverse range of hosts may decrease its fitness because of different replication rates according to host physiology and tissue tropism, altered virulence, and concurrent adaptation to multiple hosts and microenvironments (Redman et al. 2016). Nevertheless, empirical studies have shown that the evolution and adaptation of viruses in fluctuating host environments can lead to fitness improvements for both host and virus, with alternating host transmission cycles not necessarily affecting interhost fitness (Weaver et al. 1999; Turner and Elena 2000).

Stable virus–host codivergence sits in marked contrast to the transmission between distantly related host species that characterizes many emerging infectious viral diseases (Longdon et al. 2011, 2014). The latter is the likely method by which beak and feather disease virus (BFDV), the dominant pathogen of Australasian psittacine birds, has adversely affected the critically endangered orange bellied parrot (Neophema chrysogaster), which, with the swift parrot (Lathamus discolor), are the only migratory parrot species in existence (Peters et al. 2014; Sarker et al. 2014b). Both species breed in the island state of Tasmania and migrate to mainland Australia over winter.

As a member of the family Circoviridae, BFDV possesses relatively simple but compact circular, single-stranded DNA genomes of approximately 2,000 nucleotides that encode two genes, a replicase-associated gene (Rep) and a capsid gene (Cap). The disease it causes, psittacine beak and feather disease (PBFD), was recognized in Australia as a disease of national concern because the establishment of a captive-breeding program for the orange-bellied parrot was hindered by PBFD in 1985. Whereas BFDV infection may be subclinical, it can result in disease associated with damaged plumage, in particular primary remiges and rectrices, resulting in critical loss of flight and mobility. Large viral titers may be excreted in the feces before feather lesions develop. The incubation period of PBFD may be as short as 21 d depending on the dose of virus, age of the bird, stage of feather development, and absence or level of immunity (Raidal and Peters 2018). Primary virus replication occurs in the bursa of Fabricius and lymphoid tissue of the alimentary tract, with secondary replication in the liver, thymus, bone marrow, and other tissues. The maximum incubation period is unknown since the manifestation of skin disease requires a molt and disease expression may skip a molt; fledgling birds that become infected after feather development has completed may not develop clinical signs until their first or second molt. This could take 6 mo or longer, with many years of latency possible. Severely affected birds develop chronic progressive disease, advanced plumage deficits, and immunosuppression, which can ultimately result in beak and claw deformities, as well as predisposing affected birds to secondary infections (Raidal et al. 2015). Thus, BFDV infection may act as a keystone pathogen with a compound negative effect on infected birds and naïve populations (Raidal et al. 2015).

The last remaining wild population of the orange-bellied parrot is now fewer than 20 birds and, although captive breeding in Tasmania, Victoria, and New South Wales has boosted annual recruitment into the wild, the management of captive populations of orange-bellied parrots continues to be troubled by BFDV since its re-emergence in captivity in 2006 (Peters et al. 2014; Raidal et al. 2015). Before this event, routine high-density testing of the captive population for many years demonstrated freedom from infection with BFDV. The original reservoir of BFDV transmission in this population of birds was considered almost certainly to have been from other wild parrot species in Tasmania (Sarker et al. 2014b). In 2015, 19 of 23 wild nestling orange-bellied parrots tested positive for BFDV during routine disease surveillance monitoring at the species' last remaining stronghold in Tasmania, which raised the possibility of either a separate spillover infection occurring in the wild population or accidental introduction of BFDV genotypes by captive-bred released birds. Previous studies have demonstrated high BFDV mutation and recombination rates as well as BFDV quasispecies dynamics among infected orange-bellied parrot flocks (Sarker et al. 2014b), but have not characterized the likely origin of spillover genotypes.

Regardless of the primary cause of species loss, decline to extinction may be driven by synergistic processes disconnected from the original cause of decline (Brook et al. 2008). Degraded immunocompetence due to reduced genetic diversity and proximate stressors that further degrade population health can act as self-reinforcing processes, amplifying negative feedbacks such as infectious disease. The wild population of orange-bellied parrots has likely been in decline for more than a century (Ashby 1923; Dove 1924; Martin et al. 2012), but the main drivers for the initial and ongoing decline of the species are not well characterized, although habitat loss and degradation, weed invasion, introduced competitors, and altered fire regimes probably were and remain primary contributing factors (Martin et al. 2012; Stojanovic et al. 2018). Species that are adversely affected by major threats such as habitat loss, competition from invasive species, or hunting are more susceptible to infectious diseases (Heard et al. 2013). Conservation actions must consider the multifactorial and synergistic action of threatening processes and how primary and opportunistic pathogen reservoirs are needed to predict and mitigate the re-emergence of pathogens in naïve subpopulations. Otherwise, conservation actions that target single-threat drivers risk failure because of the potential cascading effects that can result from unmanaged threats (Brook et al. 2008). To assess the potential host-based BFDV reservoir threats to the orange-bellied parrot captive breeding program, we characterized BFDV genotypes in re-emergent infections of captive and wild orange-bellied parrot subpopulations and assessed the likely geographic or taxonomic source of these infections from potential reservoirs of BFDV spillover in Australia. The results help us to understand the risks and relationship between BFDV–host specificity and likelihood of future BFDV spillover to wild orange-bellied parrots and other threatened Australasian parrot species with dwindling population sizes.

Samples used in the analyses described soon were obtained during the course of normal health monitoring and veterinary checks of orange-bellied parrots held in captive breeding flocks controlled by the Tasmanian Government Biodiversity Conservation Branch and the Victoria Government-controlled Healesville Sanctuary (Healesville, Victoria), Zoos Victoria (Parkville, Victoria) and two independent facilities (Priam Psittaculture Center, Bungendore, New South Wales and Moonlit Sanctuary Wildlife Conservation Park, Pearcedale, Victoria) in 2014 and 2015. A batch of 23 samples collected from wild nestling orange-bellied parrots at Melaleuca, Tasmania was also analyzed during this period. Samples collected (n=23) opportunistically from common species of wild parrots in proximity to Healesville Sanctuary were also available to detect potential spillover events of BFDV infection at the wild–captive bird interface and to evaluate the spatial divergence of the BFDV population. This included samples obtained from sulphur-crested cockatoo (Cacatua galerita), musk lorikeet (Glossopsitta concinna), rainbow lorikeet (Trichoglossus haematodus), little corella (Cacatua sanguinea), eastern rosella (Platycercus eximius), crimson rosella (Platycercus elegans), and Australian king parrot (Alisterus scapularis). Animal sampling was authorized by the Charles Sturt University Animal Care and Ethics Committee (permit 09/046). Total genomic DNA was extracted from archived dried blood samples collected onto filter paper from all birds, including 35 orange-bellied parrots, following a previously described protocol (Bonne et al. 2008). The extracted DNA samples were screened for BFDV infection using routine BFDV diagnostic regimen at the Veterinary Diagnostic Laboratory, Charles Sturt University using established protocols (Ypelaar et al. 1999) and initially genotyped using high-resolution melt analysis of the Cap gene (Das et al. 2016a).

From BFDV-positive samples, full-length BFDV genomes were amplified in multiple overlapping PCR fragments using previously developed primers and thermal conditions (Sarker et al. 2014b) and commercially Sanger sequenced (Australian Genome Research Facility Ltd., Sydney, Australia). In situations where full-length amplification failed, a partial Rep gene segment of the genome was sequenced to provide a lower-resolution genotypic identity for all isolates (Sarker et al. 2014a). All unambiguous full-length and partial Rep sequences from this and previous studies (totaling 160 full-length genome and 319 partial Rep BFDV sequences) were used to determine BFDV population structure in orange-bellied parrots and in multiple other parrot species across Australia.

Individual sequences were annotated with accession number, geographic (state-based) origin, host species, host taxonomic tribe, and sampling year. Genomes were aligned in Geneious (Geneious version 9.2, Biomatters, Auckland, New Zealand) with MAFFT version 7.017 using the G-INS-i (gap open penalty 1.53; offset value 0.123) alignment algorithm (Katoh et al. 2002). For the BFDV phylogeny a general-time-reversible model with gamma distribution rate variation and a proportion of invariable sites (GTR+I+G4) provided the lowest Akaike information criterion using the program jModelTest 2.1.3 (Darriba et al. 2012). Maximum-likelihood phylogenies of full length BFDV genomes sampled across Australia (n=160) were inferred using the program PhyML version 3.1 (Guindon and Gascuel 2003). Raven circovirus was used as an outgroup for the full genome phylogenetic reconstruction.

For median joining network reconstruction, partial Rep sequences were aligned using the DNA alignment software program DNA Alignment version 1.3.3.2 (Fluxus Technology Ltd., Colchester, UK) and networks were calculated in Network version 5.0 (Bandelt et al. 1999). Maximum-parsimony purging was used to remove unnecessary median vectors and links.

A series of sequence-based multivariate analyses was performed to investigate stratification in the BFDV population across Australia. Principal component analysis was used in the analysis of genetic haplotype data as an unsupervised clustering method to discern underlying population structure and because it can summarize highly multivariate genetic data into a few synthetic variables that capture variation observed across the data set. We implemented principal component analysis in ClustVis (Metsalu and Vilo 2015), which produces synthetic variables showing the differences between predefined sample groups, in this case sampling region and host taxonomy. A nonhierarchical statistical parsimony network was also used to explore genealogical relationships between haplotypes and their respective host population using the R script, TempNet (Prost and Anderson 2011).

We used an unsupervised Bayesian genetic stratification approach to discriminate the Australian BFDV subpopulation structures to derive an unbiased estimation of probable genetic clustering of BFDV sequences across the Australian landscape. This was achieved by analyzing full-length BFDV sequences using STRUCTURE version 2.3.4 (Pritchard et al. 2000). We used the program xmfa2struct (Didelot and Falush 2007) to convert the genomic sequences to STRUCTURE-compatible files. To estimate K, the number of subpopulations, the BFDV data set was analyzed allowing the value of K to vary from 1 to 12, with an initial burn-in of 10,000 iterations followed by 50,000 iterations. Five independent runs were carried out for each K value (equating to 60 runs in total). Default parameters and an admixture model with the option of correlated allele frequencies between populations were used since the model could account for BFDV recombination, producing some individuals with mixed ancestry or for allele frequencies in subpopulations being similar because of admixture or shared ancestry. The program STRUCTURE HARVESTER was used to detect the optimum number of population structures (Earle and von Holdt 2012) by inferring the appropriate ΔK (highest change of likelihood function). After obtaining the optimum K value, at least 30 independent reruns in STRUCTURE were performed to obtain the final result. Genetic stratification was visualized using the program CLUMPAK (Kopelman et al. 2015) with a suitable distract setting and default color parameters. The number of genetic subpopulations thus obtained by unsupervised clustering was subsequently coupled with spatial or host-dependent analysis and subjected for molecular variance analysis implemented in the program Arlequin version 3.1 (Excoffier et al. 2005) using both standard haplotypic format and locus by locus with 1,000 permutations for each (Excoffier et al. 1992). Gene flow between the inferred genetic clusters (subpopulations) was determined by comparing pairwise Fst values using both Arlequin and DnaSP version 5 (Librado and Rozas 2009).

Genetic differentiation parameters such as segregating sites, haplotype frequency, haplotype diversity, and nucleotide diversity for each putative population were determined using DnaSP version 5. Tajima's D neutrality test in DnaSP version 5 was also performed in each population. Here, subpopulations were initially denoted on the basis of either geography using state borders or taxonomic nomenclature (tribe) of host species. This approach was subsequently followed in STRUCTURE-defined subpopulations.

Nineteen of 23 (83%) samples obtained from wild orange-bellied parrots, including nestlings (n=19), juveniles (n=3), and one adult (n=1) bird during the 2015–16 breeding season at Melaleuca, Tasmania, Australia, were PCR positive for BFDV. High-resolution melt curve analysis targeting Cap indicated that these were genetically different from previously known BFDV genotypes circulating in captive orange-bellied parrot populations (Sarker et al. 2014b) and more closely related to genotypes found in cockatoos (GenBank accession no. KF385417; Supplementary Material Fig. S1). Twelve full-length BFDV genomes were obtained from these samples. In addition, 23 whole-genome BFDV sequences were obtained from wild parrots presented to Healesville Sanctuary and diagnosed with PBFD, including Australian king parrots (n=7), little corellas (n=4), sulphur-crested cockatoos (n=2), rainbow lorikeets (n=2), crimson rosellas (n=4), eastern rosellas (n=3), and a musk lorikeet (n=1). A further four new BFDV genomes were obtained from archived samples collected in 2015 during a previous BFDV outbreak (Sarker et al. 2014b) in captive orange-bellied parrots from Tasmania. A maximum-likelihood phylogeny (Fig. S2) was reconstructed using all publicly available full-length BFDV genomes isolated in Australia and the full-length genomes (n=39) generated from this study (Table 1). This phylogeny revealed that the BFDV genotypes associated with the 2015 outbreak of infection in wild orange-bellied parrots at Melaleuca, Australia formed a monophyletic clade with weak bootstrap support (39%) as sister to a BFDV clade mostly infecting cockatoos. A previously known BFDV lineage from captive orange-bellied parrots in Tasmania was found to be the closest relative of, but not nested within, these other lineages, but with only relatively weak bootstrap support (36%). This suggests that BFDV genotypes present in the 2015 outbreak in wild orange-bellied parrots were distantly related to that circulating in the captive population. Median joining haplotype network analysis further supported the presence of genotypes of diverse origin in orange-bellied parrots (Fig. S3).

For BFDV genomes from other parrot species, there was no spatial segregation and different geographic locations were distributed throughout the phylogenetic tree. The BFDV genomes from Australian king parrots clustered together to form a separate monophyletic clade close to a clade dominated by Platycercini hosts. Similarly, genomes from rainbow lorikeets clustered with those from Loriini. These findings were consistent with host-species structuring of BFDV.

Weak spatial clustering was detected among extant BFDV genotypes from across Australia (Fig. 1A) alongside a strong host tribe-dependent population structure with variable degrees of admixture (Figs. 1B, S4 and Tables S1–S3). An unsupervised Bayesian clustering approach revealed nine (K=9) statistically supported BFDV genetic population structures across Australia (Fig. 1B). Each individual genome was given a membership probability to the respective population using a distinct color, whereas multiple color patterns represented probability of genetic admixture between populations. Consistent with other analyses, the population lacked structure using spatial variables, but host-dependent groups could be inferred. Therefore, on the basis of STRUCTURE membership probability, three distinct populations within cockatoo hosts (Cacatuidae 1, 2, and 3), one Platycercini, one Polytelini, two distinct Pezoporini (Pezoporini-1 and Pezoporini-2), one Melopsittacinae, and one Loriini BFDV populations, were identified (Fig. 1). Pezoporini groups consisted mostly of genotypes found in orange-bellied parrots and those genotypes responsible for the recent outbreak in wild orange-bellied parrots formed its own distinct population (Pezoporini-2). However, these genomes shared membership probability with other populations from Cacatuidae-2, Pezoporini, and Platycercini, indicating genetic admixture (Fig. 1B). A third, separate genotype in orange-bellied parrots does not correspond to either of the Pezoporini groups and instead shares significant membership probability with Cacatuidae populations.

It is exceptionally rare in Australia, where BFDV genotypes are widely circulating in most wild psittacine populations, to find a parrot species whose entire population was once demonstrably free of BFDV infection that could act as a model by which to understand mechanisms of spillover and reservoir maintenance of BFDV in the environment. The recent re-emergence of BFDV in captive and wild orange-bellied parrots is a reminder that assessment of disease threats to wildlife populations should consider known pathogen reservoirs in related, more abundant host species as potential sources of infectious disease transmission. We sequenced 39 new BFDV genomes from psittacines including wild and captive orange-bellied parrots and hitherto unrepresented species such as wild Australian king parrots, little corellas, eastern rosellas, and a musk lorikeet and we analyzed these alongside all available Australian BFDV genomes. The results identify at least three separate BFDV spillover events into the remaining populations of orange-bellied parrots over the past 10 yr. We uncovered significant host-based (but not geographic) BFDV stratification across the Australian landscape, highlighting processes of cryptic host adaptation and competing forces of host codivergence and cross-species transmission. These likely reflect ancient and ongoing associations between BFDV genotypes and particular host tribes, while the opportunistic ability of BFDV genotypes to spill over into new psittacine hosts has been retained. Endemic circulation of BFDV in most wild parrot populations is evidently cross-protective against such spillover, and this is most likely because of BFDV antigenic homogeneity associated with capsid structural conservation. More important, these characteristics have allowed us to infer the likely host of origin for BFDV spillover into wild orange-bellied parrots in two disease events (2012 and 2015) and have highlighted the continuing susceptibility of orange-bellied parrots and other threatened parrot species to disease precipitated by such spillover events.

Natural populations of pathogens are frequently composed of numerous interacting strains (Redman et al. 2016), with host–pathogen dynamics having a strong impact on both infection outcome and the evolution of pathogen virulence. The distribution of organisms in space is also an important arbiter of host–pathogen interactions. Previous work has shown that spatial structure can dominate the shape of host–pathogen dynamics and influence transmission–virulence trade-offs (Messinger and Ostling 2009). Spatially oriented gene flow within BFDV has been demonstrated with BFDV infecting the crimson rosella species complex along the great rivers of southeastern Australia (Eastwood et al. 2014), a distance of approximately 1,000 km. However, our results for BFDV show that spatial structuring across the Australian continent is less important than structuring according to host taxonomy. This has implications for understanding the distribution and evolutionary timescale of BFDV in Australia. Phylogeographic studies have highlighted Australia as the origin of BFDV (Harkins et al. 2014; Das et al. 2016b), with an association in Psittaciformes that is likely to have existed for at least 10 million yr (Das et al. 2016b), long enough to permit host-based stratification despite BFDV retaining an ability to jump host species.

In conservation management of endangered species, the decision to use captive breeding programs is a high-risk and costly approach because a self-sustaining captive population that can be used for releasing individuals into the wild is not always possible and there is a risk that the acquisition of founder animals could itself contribute to loss of genetic diversity in the wild (Snyder et al. 1996). Progeny destined for release should be free of nonendemic pathogens and genetically diverse to enhance wild populations. Initial concerns that the unexpectedly high prevalence of BFDV infection in wild orange-bellied parrots in 2015 might have followed transmission from released BFDV-infected captive-bred birds were allayed by the genotypic position of the BFDV lineage detected in the wild. Our demonstration that this genotype is related to genotypes associated with the host tribe Platycercini (Fig. 1) indicates that some other host psittacine species in the environment of Melaleuca, likely a green rosella (Platycercus caledonicus), was the source of spillover transmission. Nest hollows have previously been implicated in the transmission of BFDV and green rosellas are known to use nest hollows interchangeably with orange-bellied parrots at Melaleuca. The high prevalence of BFDV in wild orange-bellied parrots in 2015 highlights both a strong background force of infection in the environment at Melaleuca that the relic population of wild orange-bellied parrots and any captive-bred birds released at this location have to contend with, and population-level susceptibility to BFDV infection in orange-bellied parrots after a spillover event. The latter reflects lack of the adaptive immunity typical of parrot populations with endemic circulating BFDV.

Epidemic disease associated with BFDV is rare in wild Australian parrots, despite its presence in wild populations of most psittacine species across the continent. Evidence for vertical transmission of protective humoral immunity in nestling parrots and the acquisition of such immunity in adult parrots after infection suggests that in a typical psittacine population with circulating endemic BFDV, the manifestation of disease in individuals is unlikely, affecting less than 10% of recruitment (McOrist et al. 1984), and as a result overall population effects are negligible. Structural conservation of BFDV capsid protein explains observed antigenic homogeneity and cross-protection of individuals against subsequent infection from other BFDV genotypes. The loss of endemic infection thus opens an opportunity for BFDV circulating in reservoirs to spill over into naïve populations, eliciting disease in juvenile birds especially. The absence of circulating BFDV in an Australian parrot species or population should be considered atypical, and the conspicuous absence of this virus in previous diagnostic surveillance of orange-bellied parrots suggests that their population had fallen below a viable threshold to maintain endemic BFDV infection (Peters et al. 2014; Sarker et al. 2014b; Raidal et al. 2015; Raidal and Peters 2018). For this reason, the pre-existing and repeated loss of BFDV infection in orange-bellied parrots appears to have been the ultimate cause of repeated spillover and disease. The presence of widespread and abundant natural reservoirs for BFDV in Australia poses a significant threat to threatened parrot species that have lost endemic BFDV, most likely including western ground parrots (Pezoporus flaviventris) and swift parrots (Lathamus discolor). For such species, alternative strategies including vaccination need to be considered for providing artificially stimulated flock immunity.

Our finding that spillover of BFDV into orange-bellied parrots from other wild parrots has occurred repeatedly highlights how loss of an endemic pathogen in threatened species can allow spillover from reservoirs and the emergence of disease. The presence of infectious organisms in threatened species is typically viewed as deleterious, but this case demonstrates that a more holistic understanding of the relationship between hosts and their pathogens is needed in conservation. Threatened species have declining and often relatively small population sizes and these characteristics correlate with the loss of endemic pathogens. Rather than being advantageous, impoverishment of the microbiome might actually be an acquired risk. Healthy wild populations have endemic pathogens and our approach to conserving threatened species should recognize the significance of these and proactively manage the repercussions of their loss.

The authors thank the contributions of Peter Copley, Sheryl Hamilton, Jocelyn Hockley, Kristy Penrose, Judy Clark, and Annika Everaardt of the Orange-Bellied Parrot Recovery Team, and the staff of the Australian Wildlife Health Centre (Healesville Sanctuary). S.D. received an Australian Government Research Training Program Scholarship for this project.

Supplementary material for this article is online at http://dx.doi.org/10.7589/2018-06-154.