Viral gut metagenomics of sympatric wild and domestic canids, and monitoring of viruses: Insights from an endangered wolf population

Abstract Animal host–microbe interactions are a relevant concern for wildlife conservation, particularly regarding generalist pathogens, where domestic host species can play a role in the transmission of infectious agents, such as viruses, to wild animals. Knowledge on viral circulation in wild host species is still scarce and can be improved by the recent advent of modern molecular approaches. We aimed to characterize the fecal virome and identify viruses of potential conservation relevance of diarrheic free‐ranging wolves and sympatric domestic dogs from Central Portugal, where a small and threatened wolf population persists in a highly anthropogenically modified landscape. Using viral metagenomics, we screened diarrheic stools collected from wolves (n = 8), feral dogs (n = 4), and pet dogs (n = 6), all collected within wolf range. We detected novel highly divergent viruses as well as known viral pathogens with established effects on population dynamics, including canine distemper virus, a novel bocavirus, and canine minute virus. Furthermore, we performed a 4‐year survey for the six wolf packs comprising this endangered wolf population, screening 93 fecal samples from 36 genetically identified wolves for canine distemper virus and the novel bocavirus, previously identified using our metagenomics approach. Our novel approach using metagenomics for viral screening in noninvasive samples of wolves and dogs has profound implications on the knowledge of both virology and wildlife diseases, establishing a complementary tool to traditional screening methods for the conservation of threatened species.


| INTRODUCTION
Infectious diseases may have direct effects on demographic patterns of wildlife populations and are seen as a substantial threat to the conservation of global biodiversity (Daszak, Cunningham, & Hyatt, 2000;Gordon et al., 2015). Viruses remain a major cause of disease outbreaks in wildlife, partially due to anthropogenic effects. Disease surveillance is of particular importance for wild canids, as their evolutionary proximity to domestic dogs increases the risk of disease spill over (Bryan et al., 2011;Pedersen, Jones, Nunn, & Altizer, 2007). In fact, domestic dogs are recognized as an important risk factor for disease in sympatric wild canids, including wolves (Canis lupus), as dogs usually attain much higher population size compared to wild canids (Bryan et al., 2011;Cleaveland, Kaare, Knobel, & Laurenson, 2006;Lescureux & Linnell, 2014). Epidemics of diseases, such as canine distemper and canine parvoviruses, can severely affect wolf population dynamics. Canine distemper was suggested to cause high pup mortality (60%-90%) and periodic wolf population declines in the Greater Yellowstone area (Almberg, Cross, Dobson, Smith, & Hudson, 2012;Almberg, Cross, & Smith, 2010;Almberg, Mech, Smith, Sheldon, & Crabtree, 2009). Canine parvovirus caused a catastrophic decline in the wolf population in Isle Royale (Wilmers, Post, Peterson, & Vucetich, 2006) and was shown to reduce wolf recruitment by 40%-60% in Minnesota (Mech, Goyal, Paul, & Newton, 2008). The transmission of infectious diseases between domestic dogs and free-ranging wolves has also been documented in human-dominated landscapes in Portugal (Müller et al., 2011). However, due to lack of appropriate studies, there is only very limited data available on these and other viral pathogens that might be circulating in wild canid populations.
Next generation sequencing (NGS) revolutionized the way we study ecosystems, as it offers a broad, noninvasive tool to detect nucleic acids from different types of pathogens, depending on the sample type and preparation method used. In the virology field, viral detection is traditionally performed using classical methods such as antigen-, antibody-, or nucleic acids-detection assays targeting a specific agent. However, the recent democratization of NGS technologies allows researchers to perform diagnostics without targeting, a priori, specific infectious agents. Its high sensitivity allows the detection of pathogens present at very low abundance in biological samples. This approach is particularly suited to study the occurrence of known and unknown viruses in wildlife populations, coupled with noninvasive sampling methods. Furthermore, this is particularly relevant for free-ranging wolves from which is exceedingly difficult to obtain fresh biological samples, due to their elusive nature.
Wolves in the Iberian Peninsula (Canis lupus signatus) occur in a heterogeneous and anthropogenically modified landscape, coping with intense human-related factors of disturbance, such as settlements, infrastructures, or domestic dogs, but also profiting from food sources such as livestock (Eggermann, da Costa, Guerra, Kirchner, & Petrucci-Fonseca, 2011;Llaneza, López-Bao, & Sazatornil, 2012). In Portugal, wolf populations suffered a drastic decline in range and numbers during the 20th century, with the most critical situation being observed in the region located south of the Douro river. The South Douro wolf population suffered a drastic decline on range and numbers in the last few decades persisting in a highly anthropogenically modified landscape at the very edge of the species distribution (Pimenta et al., 2005). Currently, this population faces several threats such as depleted genetic diversity, low prey availability, habitat fragmentation, and geographic isolation from the main Iberian wolf range (Grilo, Moço, Cândido, Alexandre, & Petrucci-Fonseca, 2002;Hindrikson et al., 2016;Pimenta et al., 2005). Moreover, the few packs comprising this population show low reproduction rates, possibly due to human disturbance in breeding sites, intensive human persecution, and other factors, such as food scarcity and diseases prone to induce pup mortality (Hindrikson et al., 2016;Pimenta et al., 2005;Roque, Bernardo, Godinho, Petrucci-Fonseca, & Álvares, 2013). As a result, this small wolf population is one of the few in Europe considered to be on the verge of extinction (Boitani & Ciucci, 2009). Furthermore, wolves from this region are highly dependent on human-related food resources, with livestock comprising over 90% of their diet (Torres & Fonseca, 2016;Vos, 2000). Anthropogenic food sources including extensively bred livestock (cattle, sheep, and goat) or scavenging on carcasses of intensively farmed animals (horses, pigs, rabbits, and poultry; Roque, Palmegiani, Petrucci-Fonseca, & Álvares, 2012) might increase contacts with domestic dogs and potential for disease spill-over. This human-dominated environment, together with demographic isolation and limited genetic diversity, make the occurrence of disease a particularly important conservation issue in this wolf population (Wilmers et al., 2006).
In this study, we aimed to get insights on the occurrence, sharing, and evolutionary relationship of enteric viruses in sympatric wolves and dogs, by characterizing the fecal virome of diarrheic dogs and wolves from the south of Douro region using viral metagenomics.  (Roque et al., 2013).

| Sample collection and preparation for viral metagenomics
To investigate putative disease causing agents, diarrheic stools attributed to free-ranging wolves (n = 8), feral dogs (n = 4), and pet dogs (n = 6) were collected from the Portuguese South of Douro wolf population in the districts of Viseu and Guarda, Portugal in 2011 (Figures 1 and S1). Fecal wolf-like samples were collected along transects in areas of known occurrence of wolf packs and analyzed for species identification by amplification of a mitochondrial DNA fragment (mtDNA) as previously described (Godinho et al., 2015). Samples of pet dogs were collected at nearby veterinary clinics from owners of urban dogs that reported to occasionally take their animals to areas within wolf range. We use the term "feral" for dog samples collected on the field within wolf range, while "pet" refers to samples of urban dogs collected in veterinary clinics. Diarrheic stools were defined as nonformed, watery feces, after evaluation by a veterinarian. All fecal samples used for viral metagenomics were stored at −20°C before shipping to the laboratory of viral metagenomics (KU Leuven) where they were kept at −80°C until further processing.

| Screening of lupine bocavirus and canine distemper virus in wolf pack feces
We screened 93 fecal samples collected between February 2009 and July 2012, corresponding to 36 individually identified wolves belonging to six packs in South of Douro (Roque et al., 2013), for lupine bocavirus and canine distemper virus. For virus detection, feces were subjected to a DNA extraction following Frantz and colleagues (Frantz et al., 2003). Potential PCR inhibitors were removed after DNA extraction using prerinsed Microcon® YM-30 Filter Units (Millipore). Negative controls were included throughout the process.
Primers were designed for the identified strains, and RT-PCR was carried out for CDV and PCR for bocavirus. Primer sequences for CDV forward is 5′ -ACT TCC GCG ATC TCC ACT GG -3′ and reverse is 5′ -GCT CCA CTG CAT CTG TAT GG -3′. The bocavirus forward primer is 5′ -AGA CCA GAT GCT CCA CAT GG -3′ and reverse 5′ -TGC CTG CCA CGG ATT GTA CC -3′. For CDV amplification, an initial reverse transcription step at 50°C for 30 min was followed by a PCR activation step at 95°C for 15 min, 35 cycles of amplification (30 s at 94°C, 30 s at 57°C, and 1 min at 72°C), and a final extension step for 10 min at 72°C in a Biometra T3000 thermocycler (Biometra).
For bocavirus amplification, the same PCR settings were used except the initial reverse transcription was omitted. Three PCR replicates were performed to decrease the number of negative results due to the low quantity DNA and RNA in samples. All (RT-)PCR reactions were performed using the OneStep RT-PCR kit (Qiagen) in a Biometra T3000 thermocycler (Biometra). The positive samples were purified with ExoSAP-IT (Affymetrix), and Sanger sequenced with the ABI Prism 3130xl Genetic Analyser (Applied Biosystems). The chromatogram sequencing files were inspected with Chromas 2.3.

| Fecal virome of sympatric wolves and domestic dogs using NGS
After quality trimming and removal of bacterial reads, a total of 18,767,912 reads from the five pools were assembled into contigs and  (Buchfink et al., 2015). In total, 2,868,974 sequences (15%) were classified as viral, of which the majority could be attributed to bacteriophage sequences (54%). In fact, members of the order Caudovirales and Microviridae family were detected in all samples (Table 1). Furthermore, sequences from the following families of viruses that infect eukaryotes were detected in the wolf pools: Parvoviridae, Picornaviridae, Circoviridae, Mycodnaviridae, Picobirnaviridae, and Paramyxoviridae (Table 1). In the feral dog pool, sequences from the Nodaviridae, Mimiviridae, Totiviridae, and Parvoviridae were detected. In the pet dog pools, Parvoviridae, Mycodnaviridae, and Virgaviridae sequences were detected (Table 1).

| Identification of canine distemper virus in wolf samples using NGS
The genus Morbillivirus, belonging to the family Paramyxoviridae, includes several known pathogens such as measles and CDV. Viruses

| Identification of two bocaviruses in wolf and dog using NGS
The Parvoviridae is a ssDNA viral family that can be divided into two subfamilies: the Parvovirinae which infect vertebrates and the Densovirinae infecting arthropods. The Parvovirinae subfamily encloses the genus Bocaparvovirus (Cotmore et al., 2014). Bocaviruses have been identified in a wide range of hosts so far, such as humans, dogs, gorillas, cats, sea lions, and pigs (Zhou, Sun, & Wang, 2014).
We identified a near complete canine minute virus (CnMV) genome in one pool of diarrheic pet dogs (Figure 2a) (Sun et al., 2009). Analysis of the VP1 protein shows high conservation and 100% amino acid similarity with isolate SH1 detected in dogs from China ( Figure 2c).
We could retrieve a second complete bocavirus coding sequence from a fecal wolf pool (Figure 2b), representing the first bocavirus isolated from a wolf. We named it Lupine bocavirus isolate South of Douro (5,228 nt). Analysis of the structural coat protein of this lupine bocavirus showed low similarity with known bocaviruses and was most similar (62.5% amino acid similarity) to a porcine bocavirus isolated in China (Cheng et al., 2010), forming a distinct and novel cluster among bocaviruses (Figure 2c).

| Identification of densoviruses in wolf using NGS
We could detect another two viruses from the Parvoviridae family in wolves. This family can be subdivided into two subfamilies:

| Identification of caninovirus, an undescribed nodavirus in a feral dog using NGS
The Nodaviridae family is divided into two distinct genera: the Alphanodavirus and Betanodavirus. These viruses are of segmented single-stranded RNA and the two segments encode for an RdRp (RNA1) of 3.2 kb and a capsid segment of 1.2 kb (RNA2; Ball, Wohlrab, & Li, 1994). Alphanodaviruses have been mostly isolated from insects, opposing to betanodaviruses which infect fish and are responsible for viral nervous necrosis in several species (Bailey & Scott, 1973;Shetty, Maiti, Shivakumar Santhosh, Venugopal, & Karunasagar, 2012

| PCR screening for canine distemper virus and lupine bocavirus in wolf pack feces
Based on our findings using NGS, a further assessment on the prevalence of canine distemper virus (CDV) and lupine bocavirus in fecal samples from the wolf population was carried out (Figure 1b, Table   S1). We selected these two viruses for screening in the endangered South of Douro wolf population as CDV is an established viral pathogen in wolves, and more information of the novel lupine bocavirus is needed to unravel its epidemiology. From the 93 fecal samples corresponding to 36 individual wolves from six packs, nine were PCR positive for CDV and 34 positive for lupine bocavirus, which we further confirmed with Sanger sequencing (Table 2).
For CDV, the nine positive samples belong to eight different wolves genetically identified, as the same individual was positive twice for the virus in October and December 2011 (Table S1). Two of the positive samples were from 2010, six from 2011, and one from 2012 (Table 2 and Figure 3c). Also, seven of the positive samples were from individuals belonging to the Leomil pack, one from Cinfães pack, and one from Trancoso pack (Table S1 and Figure 3a).
Concerning the lupine bocavirus, the screening of these 93 samples revealed that 34 (36.56%) were PCR positive for the undescribed lupine bocavirus (Table 2). These 34 positive samples correspond to 13 different wolves, and since the same animals were sampled along 4 years (2009)(2010)(2011)(2012), the same single wolf harbored the virus in different occasions (Table 2 and S1). Of note, one of the wolves from which nine samples were available, exhibited seven positive samples for bocavirus over a period of 3 years (2009-2012; Table S1).

| DISCUSSION
We applied next-generation sequencing (NGS) technologies to fecal samples of sympatric endangered wolves, feral dogs, and pet dogs with diarrhea, aiming to study their gut virome, and the potential spread of dog viruses to the endangered South of Douro wolf population.
Furthermore, we identified in wolf samples using NGS, canine distemper virus (CDV), a well-established wolf pathogen and a novel lupine bocavirus, from the Parvoviridae family that harbors canine pathogens.
Therefore, we performed a wider PCR screening for these two viruses on fecal samples from genetically identified wolves, belonging to the six packs comprising this endangered wolf population in Portugal.
Using NGS, we were able to detect both undescribed and known viruses in wolves and dogs. CDV was identified in feces of wolf and is known to induce high mortality in wild and domestic carnivores (Almberg et al., 2010;Beineke, Baumgärtner, & Wohlsein, 2015). On the Iberian Peninsula, few studies have investigated exposure of freeranging wolves to CDV by serological methods, reporting from 0 up to 18.7% seroprevalence (Millán et al., 2016;Santos, Almendra, & Tavares, 2009;Sobrino, Arnal, Luco, & Gortázar, 2008). Using molecular methods, Millán et al. (2016) did not detect CDV in feces and tissue samples from 54 free-ranging wolves. Müller and colleagues were able to retrieve a complete H protein gene (encoding the viral hemagglutinin, which is responsible for host recognition) from two wolves whose death was attributed to canine distemper, one of which originated in 2008 in the same South of Douro wolf population sampled in our study. Their H protein analyses revealed the presence of a glycine (G) and a tyrosine (Y) at positions 530 and 549, respectively (Müller et al., 2011). These substitutions were also present in the strain from this study and are typically found in CDV strains identified in dogs, suggesting the occurrence of cross-species transmission events. In other wildlife species, different amino acids have been found at these sites, suggesting a functional role of residue 549 in host switches (McCarthy, Shaw, & Goodman, 2007;von Messling et al., 2005;Sekulin et al., 2011). However, it is possible that mutations at these sites are not required due to the genetic proximity between dogs and wolves (Müller et al., 2011). Our results further support that the CDV strains infecting wolves in Portugal are shared with dogs and may be able to circulate in the wolf population for a prolonged time, confirming that dogs can act as reservoirs of pathogens for wolves, as already suggested previously (Bryan et al., 2011;Müller et al., 2011).
To our knowledge, this is the first survey for CDV in Portuguese wolves using noninvasive fecal sampling, further establishing a reliable new tool for CDV surveillance in this species, as previously carried out in Spain (Millán et al., 2016)  of CDV is thought to occur mainly in oronasal exudates, the virus replicates widely in lymphoid and epithelial tissues, including the gastrointestinal system (Deem, Ph, Dipl, Spelman, & Dipl, 2000;Williams & Barker, 2008), so fecal shedding is not unexpected. However, it cannot be excluded that some wolves harbored the virus but did not shed it in their feces. As viral shedding is limited to up to 90 days postinfection (Deem et al., 2000), the detection of CDV RNA in 9.8% of the fecal samples, corresponding to 21.6% of the sampled individual wolves from 2009 to 2012, suggests a high exposure to the virus in South Douro wolf population. Nevertheless, the small population size estimated in 18-50 individuals; (Pimenta et al., 2005) is probably not capable of maintaining CDV, requiring spill over from sympatric host species such as dogs or other carnivores, as previously reported to occur (Müller et al., 2011). To further support this fact, the wolves sampled for this study show very reduced movements between packs, being the Arada pack the one reported to be the most isolated (Roque et al., 2013).
Moreover, our data show a peak in CDV detection in fecal samples collected in 2011, while in 2009, it was not detected, suggesting an epidemic pattern of infection. This same epidemiological scenario was suggested to occur in the Yellowstone ecosystem (Almberg et al., 2010).
Interestingly, in 2011, none of the wolf packs in this population had evidences of breeding occurrence, in contrast with the previous years Roque et al., 2013), which may suggest an influence of CDV episodes with breeding, by leading to high pup mortality and contributing to the low reproduction rates in this wolf population.
Interestingly, in the same wolf fecal pool where CDV was identified, a complete undescribed lupine bocavirus could be retrieved. This interest, several wolves belonging to the same pack were positive for the virus at the same time (Table 2 and S1). In dogs, the pathogenicity of bocaviruses is better studied for canine minute virus (CnMV), but less elucidated for the more recently identified bocaviruses. However, a recent fatal outbreak of enteritis in dogs was attributed to a canine bocavirus (Bodewes et al., 2014). Overall, this suggests that bocaviruses have the potential to cause disease, but we cannot exclude that under certain circumstances, this virus could be just an asymptomatic transient passenger, or even derived from an infected prey. These data provide novel insights on this agent, which is not standardly screened for. These results underline that this undescribed virus might be circulating among populations, and it would be a candidate to be added to the panel of screening agents for diarrhea.
Canine minute virus (CnMV) has been identified in healthy dogs but also has been associated with mortality of puppies and elderly dogs (Decaro et al., 2012;Jarplid, Johansson, & Carmichael, 1996;Ohshima, Kawakami, Abe, & Mochizuki, 2010). Serological studies indicate that CnMV is rather widespread in domestic dogs (Jang et al., 2003;Mochizuki et al., 2002 This study provides a better understanding on the viral populations in the gut of canids, as well as the cocirculation of various known pathogens in both wolves and dogs. While the genetic similarity between domestic and wild canid hosts may play a role in viral host switching, the intensity and rate of contact are critical in this process (Parrish et al., 2008). Hence, our findings raise awareness for the need for a thorough viral screening among wild canid populations for conservation purposes, particularly the ones occurring in anthropogenically modified environments in close contact with domestic dogs. We detected CDV, a well-known canid pathogen which has been proven to cause mortality and seriously compromise conservation efforts (Di Sabatino et al., 2014;Mech et al., 2008;Müller et al., 2011;Wilmers et al., 2006). Interestingly, we also identified a novel bocavirus complete genome in wolf feces, with unknown pathogenic implications.
In this study, we aimed to study viruses using an estab-

CONFLICT OF INTEREST
None declared.

DATA ACCESSIBILITY
All obtained sequences from the undescribed viruses described in this study were submitted to GenBank under the following accession numbers: KY214439 to KY214448.