Emerging infectious diseases are increasingly reported in wildlife, often in association with changing environmental conditions (Daszak et al., 2000). Environmental change may alter the community structure of hosts or pathogens, including species diversity, composition and contact patterns, and may also influence host-parasite interactions and disease dynamics (Gottdenker et al., 2014; Van Hemert et al., 2015). Northern ecosystems are changing especially rapidly, where unprecedented rates of climate change are influencing a wide range of ecological processes (Post et al., 2013; IPCC, 2014). Several recent emergence events of macroparasites in northern wildlife have been linked to climate warming (Kashivakura, 2013; Kutz et al., 2013; Laaksonen et al., 2015), whereas the role of climate change in emerging diseases caused by microparasites, including viruses, bacteria and protozoa, is not yet as clearly established (e.g., Goldstein et al., 2009; Kutz et al., 2009; Harms, 2012; Loiseau et al., 2012; Ducrocq et al., 2013; Hansen et al., 2015; but see Ytrehus et al., 2015). Pathogen emergence poses a risk to the health and sustainability of wildlife populations which, in addition to their intrinsic value and ecosystem services, continue to be an important component of community health and food security for many indigenous communities in temperate and arctic regions of North America (Inuit Circumpolar Council- Canada, 2012; Council of Canadian Academies, 2014).

The mechanisms behind the emergence of pathogens or novel disease presentations in previously unaffected ecosystems are often difficult to establish. One of the key questions associated with emergence events is whether they are caused by new introductions, or whether they are outbreaks of endemic pathogens. The ability to make such distinctions is often hampered by a paucity of baseline data on pathogen biodiversity, especially in ecosystems poorly characterized in this respect, such as those at northern latitudes (Hoberg et al., 2008; Kutz et al., 2009, 2014; Van Hemert et al., 2014).

Recently, widespread mortality events of muskoxen (Ovibos moschatus wardi) in the Canadian Arctic Archipelago associated with Erysipelothrix rhusiopathiae were reported (Kutz et al., 2015). This was the first report of E. rhusiopathiae in the Arctic, and the first time E. rhusiopathiae was associated with large-scale ungulate mortalities. Erysipelothrix rhusiopathiae is a Gram positive, facultative intracellular bacterium, best known as an opportunistic pathogen of swine (Opriessnig and Wood, 2012), poultry (Bricker and Saif, 2013), and humans (Veraldi et al., 2009). It is a ubiquitous, globally-distributed multi-host opportunistic pathogen and commensal which has been found in terrestrial and aquatic mammals, birds, fishes and arthropods (Brooke and Riley, 1999). Reports of E. rhusiopathiae causing disease in wildlife have been sporadic (Leighton, 2008). Rare outbreaks with a large number of mortalities have involved avian hosts (Wolcott, 2007), or mice (Wayson, 1927), whereas mortalities in wild ungulates were previously limited to small numbers (Bruner et al., 1984; Campbell et al., 1994). Both cattle (Hassanein et al., 2003; Zambrano-Nava et al., 2011) and swine (Stephenson and Berman, 1978) can act as asymptomatic carriers, with E. rhusiopathiae isolated from tonsils, bone marrow (Spears, 1955), and other tissues at slaughter of apparently healthy animals, and it is also part of the normal vaginal flora of cattle (Zambrano-Nava et al., 2011). Erysipelothrix rhusiopathiae appears to have a particular tropism for tonsils; tonsillar crypts are a likely port of entry and site for persistent infection in swine (Harada et al., 2013). It is probable that this bacterium is carried by various additional host species, including wildlife, with no clinical signs (Wang et al., 2010), although there is little empirical evidence available to support this hypothesis. Transmission of E. rhusiopathiae is thought to occur primarily through indirect contact with bacteria shed in feces, urine, saliva and nasal secretions; it is also hypothesized that arthropods may occasionally act as vectors (Brooke and Riley, 1999). Although there is currently no evidence that E. rhusiopathiae can replicate in the environment (Wood, 1973; Chandler and Craven, 1980), the bacterium can survive for up to several months outside of a host depending on the ambient conditions (Mitscherlich and Marth, 1984). It has been hypothesized that this ability to persist in the environment for extended periods, in addition to E. rhusiopathiae's wide host range, are the most important factors facilitating its widespread distribution (Brooke and Riley, 1999). Several virulence factors have been demonstrated or suggested for this bacterium, including neuraminidase, surface protective antigen A and the capsular polysaccharide (Ogawa et al., 2011; Shi et al., 2012), the majority of which appear to be commonly present (Janßen et al., 2015). Isolates of E. rhusiopathiae sequenced to date fall into three distinct clades, all of which appear to have retained some degree of host generalism (Forde et al., 2016), suggesting that most strains can likely be transmitted between species.

Erysipelothrix rhusiopathiae septicemia was determined to be the proximate cause of acute deaths among muskoxen of all ages and both sexes on Banks and Victoria Islands, Northwest Territories and Nunavut, Canada, first noted in 2010 (Kutz et al., 2015). These two islands, which encompass an area of over 300,000 km², were previously home to over 70% of muskoxen in Canada (Ferguson and Gauthier, 1991). Their combined population increased from around 75,000 in the late 1980's to approximately 140,000 animals in 2001 (Dumond, 2006). Recent surveys on both islands suggest that these populations have since declined by more than 50% (Davison et al., 2014; Dumond and Leclerc, Government of Nunavut, unpublished data 2014); only on the northwest corner of Victoria Island has the muskox population remained relatively stable since 2005 (Davison and Williams, 2015). The extent to which the E. rhusiopathiae-associated mortalities have contributed to these population declines is unknown, however, high mortality rates of adult animals are likely to have had an impact at the population scale.

Over a similar timeframe to the muskox population declines in Nunavut and Northwest Territories, higher than expected mortality rates or unexplained mortalities were observed in other northern ungulate populations. On the Northern Seward Peninsula of Alaska, US, mortality rates of 14–33% were recorded among adult radio-collared muskoxen June through August 2009–12 (L. G. Adams, U.S. Geological Survey, unpublished data). In northeastern British Columbia (BC), Canada, during a broad study on the health of boreal caribou populations (Rangifer tarandus caribou) the annual survival rate among radio-collared adult females between May 2013 and April 2014 was lower than expected (Culling and Culling, 2014). Multiple deaths among caribou were observed in this study area between December 2012 and July 2013 that could not be attributed to predation (B. Macbeth, University of Calgary, unpublished data). Meanwhile, in central BC, population declines in moose (Alces alces) of 50–70% were noted over a 7–15 year period (BC FLNRO, 2012, 2015; Kuzyk and Heard, 2014). These observations of higher than expected mortality of unknown etiology in muskoxen in Alaska and boreal caribou in BC similar to that observed among muskoxen on Banks and Victoria Islands, along with declining moose populations in parts of BC—all across a similar timeframe—led us to hypothesize that these mortalities might be part of a common emerging disease event associated with E. rhusiopathiae. Given its previously documented effects on muskoxen (Kutz et al., 2015), the potential role of E. rhusiopathiae in these other wild ungulate mortalities warrants investigation.

The aims of this study were to (1) test for the presence of E. rhusiopathiae in geographically disparate ungulate populations across northern latitudes of North America, and (2) quantify molecular diversity and evolutionary relationships among isolates to assess if recent disease events and population declines are associated with endemic or recently introduced pathogens, and whether a single strain was involved in the disparate mortalities. We took a bacterial genomic approach to achieve the maximum level of resolution for investigating genetic diversity and relatedness among strains (Robinson et al., 2013; Kao et al., 2014) in order to gain insights into the processes underlying the possible emergence of E. rhusiopathiae in northern wild ungulates.

Prior to the muskox die-offs on Banks and Victoria Islands (Kutz et al., 2015), E. rhusiopathiae had not been implicated in large-scale wildlife mortality events, although individual cases or small clusters of deaths associated with this bacterium had been reported in wild ungulates (Bruner et al., 1984; Campbell et al., 1994). In this study, we isolated E. rhusiopathiae from tissues collected from muskoxen, caribou and moose across a broad geographic range where unusual acute mortalities and/or population declines occurred over a common timeframe. This led us to hypothesize that the emergence of a single strain might be responsible for the mortalities. Results from whole genome sequencing of isolates from these different populations, however, refuted this hypothesis, indicating that the geographically disparate ungulate mortalities were not caused by a shared strain of E. rhusiopathiae.

We observed three distinct levels of genomic heterogeneity of E. rhusiopathiae among the populations in this study. (1) There was minimal heterogeneity among and within most individual muskoxen sampled on Banks and Victoria Islands in the Canadian Arctic Archipelago (Sample Group 1), with only one of 45 isolates differing by more than 26 SNPs, representing a case of polyclonal diversity within an individual. (2) Among Alaska muskoxen (Sample Group 2) there was moderate heterogeneity, with isolates from different individuals all falling within Clade 3 but differing by several hundred SNPs. (3) Caribou and moose in Alberta and BC (Sample Group 3) harbored high diversity, including isolates from multiple clades. This diversity is on a similar level to that seen throughout the entire global collection of E. rhusiopathiae isolates from multiple hosts and continents (Figure 4). In this latter group, polyclonal diversity was present within several individual caribou, and we detected some cases wherein a single clone was shared among animals. The varying levels of E. rhusiopathiae heterogeneity among the sample groups might reflect differences in (i) the amount of time since the bacterium was introduced into the population, (ii) the underlying transmission dynamics of E. rhusiopathiae in these populations, or (iii) the level of existing diversity of this bacterium among the geographically disparate areas, with these explanations not being mutually exclusive.

At least two potential, non-exclusive explanations for the limited diversity observed on Banks and Victoria Islands can be initially considered: exposure to a common and possibly widespread source, or between-animal transmission (Robinson et al., 2013). The use of whole-genome sequencing eliminates poor discriminatory power as a possible explanation for the finding of a common strain. Considering the high degree of mortality and the predominance of a single clonal lineage of E. rhusiopathiae across the vast area encompassed by these islands, it is possible that these patterns are associated with a recent introduction of the bacterium into a previously naïve population. We did not find any evidence that this clone was linked to livestock sources or strains circulating at more southern latitudes (Figure 4). Previous research failed to identify a clock-like signal in E. rhusiopathiae isolates collected over a timeframe of several decades, so a global substitution rate has not been established for this bacterium (Forde et al., 2016). Therefore, it was not possible to estimate the date of the most recent common ancestor of the isolates from Banks and Victoria Islands using phylogenetic inference. Ongoing testing of archived muskox serum samples may help elucidate historic temporal patterns of exposure to this bacterium on these islands.

The presence of a clonal lineage across both Banks and Victoria Islands suggests that muskoxen on these islands are somehow epidemiologically connected. However, since muskoxen are non-migratory and their movement between these islands and to and from the mainland is uncommon (van Coeverden de Groot, 2001), the spread of E. rhusiopathiae across these islands is unlikely to have occurred exclusively through direct transmission. In contrast to swine, where carrier status is believed to be common (20–40%) (Stephenson and Berman, 1978), we found limited evidence that muskoxen act as carriers of the bacterium: of 261 hunter-killed muskoxen sampled, only four were qPCR positive and one subsequently culture positive for E. rhusiopathiae. While it is possible that this animal could have subsequently developed disease, this adult male was in good body condition with no gross abnormalities noted during government slaughter inspection associated with the commercial harvest. Environmental survival of E. rhusiopathiae is possible for more than 6 months, but its amplification outside of a host is unlikely; early studies monitoring the viability of E. rhusiopathiae under different environmental conditions failed to show any evidence of population growth or maintenance (Wood, 1973; Chandler and Craven, 1980). Additionally, the E. rhusiopathiae genome lacks several genes for the biosynthesis of essential nutrients for growth (e.g., many amino acids, cofactors, and vitamins) (Ogawa et al., 2011). Given the broad host range of E. rhusiopathiae, the maintenance and spread of this bacterium in the Arctic may involve multiple species (Kutz et al., 2015). Species such as polar bears, arctic fox, fish, and migratory birds all range very broadly and could serve as reservoirs and transport hosts, linking the islands and facilitating spread. Similarly, lemmings are cyclically abundant and sympatric with muskoxen (Larter, 1998), and may play a role in maintaining and amplifying the bacteria (Kutz et al., 2015). Although we failed to detect E. rhusiopathiae in opportunistic samples tested from non-ungulate species on Banks and Victoria Islands (Supplementary Table 1), the sample size was limited and with the exception of wolf samples, relatively geographically restricted.

The finding of a clonal lineage of E. rhusiopathiae within muskoxen on Banks and Victoria islands (Sample Group 1) is in stark contrast with the extensive diversity seen in the other ungulate populations sampled. The high degree of E. rhusiopathiae strain diversity found within the smaller geographic areas occupied by Sample Groups 2 and 3 indicates heterogeneous sources, and suggests that E. rhusiopathiae may be endemic in these populations. Hence, it is unlikely that ungulate-to-ungulate transmission or a single point source introduction of E. rhusiopathiae were major drivers in these mortalities. However, a few nearly identical isolates (differing by a single SNP) occurred among caribou or moose, so exposure to a common source or sporadic cases of between-animal transmission cannot be excluded.

The detection of polyclonal infections has important implications for molecular epidemiological studies. Closely related strains are commonly taken as evidence for potential transmission links (Wylie et al., 2005); however, this study confirmed that such connections may be overlooked if only a single isolate is selected per individual (Döpfer et al., 2008; Ågren et al., 2014). The challenge of detecting mixed infections may soon be partially overcome through the use of novel bioinformatics approaches that do not require multiple isolates to be sequenced (Eyre et al., 2013).

While the mortalities among the different populations, and most mortalities within Sample Groups 2 and 3, do not appear to be epidemiologically linked based on shared E. rhusiopathiae strains, they may instead be connected by common changes in exposure or susceptibility to this bacterium. If considered within the framework of the “epidemiological triangle” of agent, host and environment (Wobeser, 2006), all three factors may contribute to the disease outcome. Cumulative stressors, both biotic and abiotic, could be sufficient for tipping the balance in the interaction between host and pathogen from a bacterium that is otherwise tolerated, to one capable of contributing to negative health outcomes (Beldomenico and Begon, 2010; Verbrugghe et al., 2012). Because E. rhusiopathiae is a generalist and opportunist bacterium (Brooke and Riley, 1999), common stressors among or within these different populations could allow unrelated strains to opportunistically cause disease and facilitate transmission. If this is true, differences in bacterial heterogeneity observed among the study populations, rather than being associated with different transmission patterns or being the result of newly introduced vs. endemic strains, could instead reflect pre-existing bacterial diversity in these disparate environments, thus representing a third plausible explanation for the limited diversity seen on Banks and Victoria Islands. These islands were glaciated during past glacial maxima and currently harbor substantially lower biodiversity than western Alaska, which was part of the species-rich Beringia during the last ice age (Hoberg et al., 2003), and Alberta and BC, which have comparatively higher species diversity (Kays and Wilson, 2009). It follows that a generalist bacterial species would have considerably more diversity in an environment where a greater diversity and abundance of potential host species exists, compared to an area with a relatively recent history of colonization by a lower diversity and abundance of vertebrate hosts. Thus, although whole genome sequencing enabled us to determine the diversity of E. rhusiopathiae in the different ungulate populations, as a stand-alone method it did not allow us to distinguish among the multiple possible explanations for the limited diversity found on Banks and Victoria Islands, nor did it allow us to differentiate between a recently introduced vs. a previously existing pathogen.

It is worth noting that within the global collection of isolates, one branch of Clade 3 which comprises about half of the isolates from caribou, moose and muskoxen, including the predominant Banks and Victoria islands clonal lineage (Figure 4, shown by red asterisk), has not been detected in domestic animals, nor at lower latitudes, possibly suggesting a high latitude or Beringean lineage. Although challenging to predict, both changes in climate and in human and animal contact patterns could result in an expansion of the range of this particular lineage. Meanwhile, the fact that several isolates from northern ungulates, including the single divergent isolate from Victoria Island, belong to lineages of E. rhusiopathiae that contain a mix of isolates from wildlife and livestock, suggests that there might be the potential for long-range introductions from distant sources.

The possible role of infectious diseases in widespread wildlife population declines, such as those in barren ground caribou (Rangifer tarandus groenlandicus), often remains enigmatic (Vors and Boyce, 2009). There is strong evidence to suggest that septicemia with E. rhusiopathiae was the proximate cause of death among the muskoxen on Banks and Victoria Islands (Kutz et al., 2015), and acute death due to septicemia, with few premonitory signs, is among the common disease presentations in swine and other species (Leighton, 2008; Opriessnig and Wood, 2012). It is, therefore, reasonable to hypothesize that the isolation of E. rhusiopathiae from the other study populations is associated with the death of the animals, particularly those where there was evidence of acute mortality such as in some of the BC caribou. There is evidence of regular ante-mortem exposure to E. rhusiopathiae in muskoxen and caribou based on seroconversion (Kutz, unpublished data). Moreover, of 59 genes we examined that are associated with E. rhusiopathiae pathogenicity, 51 were determined to be core genes, supporting the likelihood that the strains isolated from wild ungulates are indeed capable of causing disease (Supplementary Table 3). However, we cannot rule out the possibility that the detection of E. rhusiopathiae in tissues sampled from older carcasses could be the result of post-mortem bacterial transmigration, wherein bacteria enter tissues after death (Morris et al., 2006), or contamination from the external environment. Anecdotal reports in some of the older literature suggest that E. rhusiopathiae is commonly found in dead and decomposing matter (Klauder, 1938; Shuman, 1970). Longitudinal epidemiological studies that involve targeted surveillance and timely sample collection and testing will be necessary to better understand the role of E. rhusiopathiae in wild ungulate mortalities and investigate what other factors may be contributing, including viral infections, poisoning, and/or environmental stressors. A further limitation of this study, although a reality for many wildlife studies, was the fact that samples were collected opportunistically through different health surveillance initiatives, resulting in inconsistencies in the type of samples collected and in storage conditions. We found the agreement between culture and qPCR to be inconsistent for samples tested using both methods (Table 1). Factors such as long-term sample storage (Eriksson et al., 2014), suboptimal transport conditions, bacterial concentration and viability, or the presence of inhibitory substances for qPCR could all differentially influence the sensitivity of culture and qPCR. Finally, the sensitivity of qPCR and culture for detecting asymptomatic carriers is unknown. Where possible, we based the choice of tissues for sampling on literature pertaining to asymptomatic carriage/infection of E. rhusiopathiae in livestock; however, these may not be the most relevant samples for other potential carriers, including different wildlife species.

In conclusion, the use of bacterial genomics allowed us to determine that geographically widespread mortalities associated with E. rhusiopathiae in northern wild ungulates were not linked by a common bacterial strain. Instead, the underlying epidemiology was found to be complex, with three distinct patterns of bacterial genetic heterogeneity, which may reflect either different underlying transmission dynamics, the amount of time the bacterium has been present in these environments, or geographic differences in the diversity of this bacterium. While the limited diversity of E. rhusiopathiae on Banks and Victoria Islands initially suggested the introduction of a new bacterium with subsequent host-to-host transmission, the availability of genomic data from other ungulate populations provided a valuable point of comparison, leading us to consider alternative explanations. The detection of various strains of E. rhusiopathiae in similar northern wild ungulate mortalities over a common timeframe may point to a broader issue of reduced population resilience or environmental changes favoring bacterial amplification. This finding warrants a return to the classical epidemiological triangle to examine underlying stressors, including climate change, that could make hosts more permissive to opportunistic pathogens (Kutz et al., 2015). Understanding pathogen emergence and disease dynamics in boreal and arctic environments will require investigations that look beyond single factors and simple answers.

Raw sequence reads are available on GenBank as part of BioProject PRJNA288715.

Conceived and designed the study: TF, SK, KO, RZ, RB, JD, BM, and FV. Coordinated the study: SK, KO, and TF. Performed sample collections and provided associated metadata: LA, SC, TD, MD, BE, LF, BM, CN, AN, SS, HS and SK. Performed laboratory work and data analysis: TF. Drafted the paper: TF. Edited and contributed to the manuscript: SK, KO, RZ, RB, JD, FV, LA, SC, MD, LF, CN, and HS. All authors read and approved the final manuscript.

Funding for this research was provided by the Natural Sciences and Engineering Research Council of Canada Discovery Grant and Northern Supplement and Canada Graduate Scholarship, University of Calgary Eyes High, Izaak Walton Killam Pre-Doctoral Scholarship, Canada North Outfitters, Nunavut General Monitoring Program, Nunavut Harvesters Association, Inuvialuit Implementation, BC Oil and Gas Research and Innovation Society (OGRIS), Weyerhaeuser Co. Ltd, Alberta Environment and Parks Governments of Northwest Territories and Nunavut, and fRIResearch. Any opinions, findings and conclusions expressed herein are those of the authors and do not necessarily reflect the views of funding entities. Funding entities had no involvement in the design of the study or in the interpretation of results.