INTRODUCTION
Animal diseases compatible clinically with rabies have been described in the Arctic and sub-Arctic areas for over a century. At least 150 years ago extensive epizootics among Arctic foxes (Alopex lagopus) and sled dogs were documented [Reference Gondatti1, Reference Ognev2]. The disease was frequently referred to as ‘Polar madness’, ‘Eskimo dog disease’, ‘Arctic dog disease’ [Reference Plummer3, 4], and dikovanie or dikusha in Russian [Reference Kantorovich5]. As Negri bodies were not detected in Seller-stained brain impressions, and human rabies was rarely reported from the Arctic and sub-Arctic territories, the identification of the aetiological agent as rabies virus (RABV) was not confirmed until the 1940s, when serological relatedness was demonstrated [Reference Turevich and Tebiakina6, Reference Jenkins and Wamberg7]. A variety of methods, such as the fluorescent antibody test, electron microscopy, and typing with monoclonal antibodies (mAbs), confirmed the agent as RABV [4].
The use of mAbs for antigenic typing has significantly improved the differentiation of RABV variants. For example, mAb P-41, obtained as the result of immunization of mice with a RABV isolate from an Arctic fox in Yakutia, reacts selectively with the nucleocapsid of Arctic RABV isolates [Reference Schneider, Kuwert, Merieux, Koprowski and Bögel8]. The P-41 reactive viruses were found in Arctic and sub-Arctic areas circumpolarly, and antigenic patterns of isolates from Alaska and Canada are identical or similar to those of viruses circulating in Siberia. The P-41-reactive viruses have also been identified in raccoon dogs (Nyctereutes procyonoides) and red foxes (Vulpes vulpes) in Baltic regions. It has been proposed that Arctic viruses have been translocated to this territory, and thereafter established circulation in new host species [Reference Selimov9].
Nucleotide sequencing has facilitated a more precise differentiation and subsequent phylogenetic placement of viruses isolated during the last several decades [Reference Smith10, Reference Kissi, Tordo and Bourhy11]. Three phylogenetic groups of Arctic RABV have been described [Reference Mansfield12]: one of these groups was identified in North America, and includes viruses circulating in Ontario, Maine and Greenland (at least formerly); a second group includes viruses circulating in Siberia and Alaska; the third group includes Arctic viruses with apparent circumpolar circulation patterns. The first of these three groups was described in earlier studies as having been introduced into Ontario with a wave of Arctic rabies during the 1950s, and further established independent circulation among red foxes and striped skunks (Mephitis mephitis) [Reference Nadin-Davis, Casey and Wandeler13, Reference Nadin-Davis14].
Recent studies have shown that several RABV lineages related phylogenetically to the Arctic viruses are present in the Middle East, and southern and eastern Asia. These viruses were referred to as either the Arctic or Arctic-like RABVs [Reference Kissi, Tordo and Bourhy11, Reference Nadin-Davis15–Reference Nadin-Davis19]. The most western isolation point of these viruses was described in northern Iran [Reference Nadin-Davis15].
In contrast, Baltic isolates do not belong phylogenetically to the Arctic or Arctic-like lineages, but form a clade within the European fox lineage, which is part of the ‘cosmopolitan’ canine RABV lineage [Reference Bourhy20]. Reactivity of MAb P-41 with these viruses remains incompatible with their phylogeny. One common feature of both Arctic and Baltic viruses is the amino-acid substitution (Asp to Gly) at position 115 of the nucleoprotein (N) gene. It has been suggested that this substitution could have facilitated switching of RABV from Arctic foxes to raccoon dogs, and further to red foxes [Reference Bourhy20]. However, further comparisons have demonstrated that raccoon dogs in eastern Asia maintain circulation of Arctic-like viruses with the presence of Asp115, whereas Gly115 was detected in some isolates from corsac foxes (Vulpes corsac) in eastern Siberia. The substitution with Gly115 may facilitate a positive reaction with mAb P-41, but the evolutionary or functional significance of this substitution, if any, is unclear [Reference Botvinkin21].
The objective of this study is presentation of new data on the distribution and phylogenetics of Arctic and Arctic-like viruses. These include the oldest viruses available, isolated in north-eastern Siberia (Yakutia) during 1950–1960. In addition, the implementation of a time-based evolutionary analysis, using a relaxed molecular clock, has enabled a deeper insight into the time-scale of RABV spread within Arctic regions.
METHODS
Viral isolates
Viruses isolated from Arctic foxes in Yakutia during 1950–1960 were obtained from the Russian State Collection of Viruses (Ivanovsky Institute of Virology, Moscow, Russia) as frozen mouse brain suspensions. Other viruses described here for the first time were isolated during surveillance in Siberia, Alaska and southern Asia during the period 1980–2006. A positive sample from an Iraqi dog was kindly provided by Dr L. Fuhrmann (Veterinary Laboratory Europe, Kirchberg, Germany). Initial diagnosis was performed using the standard direct fluorescent antibody test (protocol available at http://www.cdc.gov/ncidod/dvrd/rabies/professional/publications/DFA_diagnosis/DFA_protocol-b.htm). Virus isolation by intracerebral inoculation of inbred laboratory mice was performed in some instances.
RT–PCR, gene sequencing and phylogenetic analysis
Total RNA was extracted from infected material (either original host brain or mouse brain following limited passages), using TRIzol™ (Gibco-BRL Inc., Gaithersburg, MD, USA) according to the manufacturer's recommendations. The RT–PCR was performed as described previously [Reference Kuzmin22] with primers for amplification of the entire N gene (1350 bp). The PCR products were purified and subjected to direct sequencing using the ABI Prism™ 377 DNA Sequencer (Applied Biosystems, Foster City, CA, USA).
Primary assembly, alignment, consensus generation and DNA translation were performed using BioEdit [Reference Hall23]. Neighbour joining (NJ) tree analysis was performed using either the Kimura-2-parameter or Jukes–Cantor model of nucleotide substitution and performed using MEGA version 2.1 [Reference Kumar24]. Bootstrap support was estimated for 1000 replicates.
Previously published gene sequences were retrieved from GenBank for comparison (Table 1).
Table 1. Rabies viruses used in the present study
Estimation of the evolutionary rate and application of a relaxed molecular clock
Estimates of the rate of molecular evolution (μ; substitutions per site per year) for complete N gene alignments were obtained using a relaxed molecular clock [Reference Drummond25] implemented using a Bayesian Markov Chain Monte Carlo (MCMC) method within the beast program (available from http://www.beast.bio.ed.ac.uk). Sequences were dated with the year of isolation and identical sequences with the same year of isolation excluded. For sequences with an imprecise isolation date, a random subset was selected and sequences dated with the midpoint of the isolation range. For each dataset, the maximum-likelihood model of nucleotide substitution was selected using the Modeltest software [Reference Posada and Crandall26] and the selected model used as the basis for beast analysis. A lognormal distribution of rates was used with Jeffrey's priors on μ and population size [Reference Drummond27]. Two models of population dynamics were used and their likelihood compared [Reference Davis28]. MCMCs were run for a minimum length of 5×106 with a 1% burn-in ensuring all effective sample sizes were >100. A minimum of 10 000 trees was used to produce a time-scaled tree with TreeAnnotator which was displayed and edited using FigTree (both available from http://www.beast.bio.ed.ac.uk). Details of all models used are available from the authors on request.
Three datasets were used for substitution rate analysis: (1) a combined analysis of Arctic/Arctic-like RABV isolates using SG1, SG15 and SG91 dated as 1955 (n=32); (2) 17 European red fox isolates (GenBank accession nos.: AF033905, U22474-6, U22480, U42605-7, U42700-2, U42704, U42706, U42707, U43432-4); and (3) the Arctic/Arctic-like and European red fox datasets combined with the addition of 11 further isolates from the same lineage [Reference David29] (GenBank accession nos.: DQ837383, DQ837385, DQ837446, DQ837448, DQ837463, U22481, U22482, U22484, U22627, U22629, U22852).
RESULTS
Phylogenetic relationships
The NJ trees show 42 viruses as members of the Arctic and Arctic-like phylogenetic clades. The majority of the Arctic RABV isolates belonged to the same phylogenetic groups to which they have been previously assigned [Reference Mansfield12]. Group Arctic-1 (Fig. 1) was comprised of viruses circulating in Ontario among red foxes and striped skunks [Reference Nadin-Davis, Casey and Wandeler13, Reference Nadin-Davis14]. Group Arctic-2 consisted of viruses circulating in north-eastern Siberia and Alaska, predominantly among Arctic foxes. Group Arctic-3 included viruses that circulate circumpolarly in Siberia, Alaska and Canada. Despite the circumpolar distribution, the percentage identity of these viruses is remarkable. For example, isolate RVHK obtained in Norilsk (north-central Siberia), shared 98·5–99·3% nucleotide identity with isolates from Alaska and Canada. The oldest available viruses, isolated between 1950 and 1960 in north-eastern Siberia (Yakutia), all belonged to the Arctic-2 and Arctic-3 groups, and shared 98·6–99·2% identity with the more recent isolates, including those obtained during 2007 in Alaska. Mansfield et al. [Reference Mansfield12] considered groups Arctic-2 and Arctic-3 as (a) subgroups and (b) of one group, Arctic-2. However, bootstrap support for this node is low in this and the previous study. Since these viruses also exhibit different circulation patterns, we designate them as distinct groups.
Fig. 1. Neighbour-joining phylogenetic tree of Arctic and Arctic-like rabies viruses (entire N gene) used in the present study (with a European fox virus RV1590 as outgroup). Bootstrap support for 1000 replicates is provided for nodes. Branch lengths are drawn to scale.
Several viruses from Alaska, isolated from foxes and dogs during 2006–2007, were not included in the Arctic-2 and Arctic-3 groups, but were placed ancestrally, within a well-formed new group that we designated Arctic-4. In fact, viruses isolated in Alaska during the last 2 years represent all three major phylogenetic lineages except Arctic-1. However, distribution patterns are different: Arctic-3 viruses were isolated only along the northern coast, supporting their circumpolar circulation with animals migrating along the pack ice. Arctic-2 viruses were isolated in the west, and are probably maintained by Arctic foxes migrating between Siberia and Alaska across the Bering Strait which is frozen in winter. Arctic-4 viruses were found in the south-western area, and probably circulate within the local population of Arctic foxes (Fig. 2).
Fig. 2. Isolation points of Arctic-2 (●), Arctic-3 (■) and Arctic-4 (▲) rabies viruses in Alaska during 2005–2007.
Arctic-like viruses formed two clades. The Arctic-like-1 clade contained viruses circulating in the Middle East and southern Asia (Iraq, Iran, Pakistan, India), and the Arctic-like-2 clade contained viruses circulating in eastern Asia (south-eastern Siberia, Russian Far East, Japan (formerly) and Korea. The same topology was obtained for the limited gene sequences available in GenBank from a previous study [Reference Nadin-Davis19]. Interestingly, the RABV isolate ‘994 dog’, from Chita Province (south-eastern Siberia) isolated during an extensive dog outbreak in the late 1970s to early 1980s, was phylogenetically more closely related to viruses circulating in the Far East than to viruses circulating in south-eastern Siberia in wild canids.
Substitution rates and time-scaled trees
For all datasets the exponential model of population size was significantly favoured above the constant one (results not shown). Although the mean substitution rate for the European red fox dataset is higher than that of the other two, substitution rates for all three datasets were shown to have widely overlapping confidence intervals (Table 2) and are comparable to those previously estimated for the RABV N gene using the same method [Reference David29, Reference Davis, Bourhy and Holmes30]. The low value of the coefficient of variation of rates (σr) for the Arctic/Arctic-like dataset suggests a more clock-like evolution than for the other datasets [Reference Drummond25]. The use of the larger, lineage-based dataset, selected on the basis of the global time-scaled phylogeny of RABV [Reference David29], has enabled an estimation of the divergence times of the branches which led to current diversity of European/Middle-Eastern RABV and Arctic/Arctic-like RABV (Fig. 3). The time of the most recent common ancestor (TMRCA) of all these variants is estimated as 652 years [1353, 95% highest probability distribution (HPD) 1000–1663]. The TMRCA of Arctic lineages is similar when the variants are analysed alone (1524, HPD 1255–1763) or as part of the larger lineage dataset (1610, HPD 1394–1786). Moreover, both analyses suggest that Arctic-like RABV variants diverged slightly earlier than Arctic RABV variants (tree for Arctic/Arctic-like dataset not shown, available from the authors upon request). Of the Arctic variants, the divergence of the Arctic-1 lineage (RABV circulating in Ontario) appears to have occurred first (TMRCA of current Arctic-1 diversity: 1921, HPD 1874–1959) and represents a distinct lineage from other Arctic variants.
Fig. 3. Time-scaled phylogenetic tree of the rabies virus N gene dataset estimated using a relaxed molecular clock.
Table 2. Relaxed molecular clock testing
HPD, Highest probability distribution.
* Substitution rate per site per year.
† Coefficient of rate variation.
In agreement with a previous study [Reference David29], we show that the emergence of European red fox rabies (currently distributed in moderate latitudes of Eurasia) is a recent event. The lower mean substitution rate for the lineage dataset leads to a larger estimate of the root height of the European red fox variant (1834, HPD 1748–1910) than when the variant is analysed alone (1921, HPD 1839–1974).
DISCUSSION
This phylogenetic study is the first to include the oldest known Arctic RABV isolates. Analysis of these isolates demonstrates that Arctic-2 and Arctic-3 groups were already well established over 50 years ago, and that these viruses show a high degree of similarity to those currently circulating in Arctic regions. The fact that the Arctic-2 viruses appear to be restricted to Siberia and Alaska, whereas lineage Arctic-3 has a circumpolar distribution, may be connected to migratory activity of the main wildlife virus reservoir, the Arctic fox, or intercontinental movements of humans and dogs, as suggested in Figure 2. Indeed, phylogenetically these groups are closely related, and their separation could have occurred about 100 years ago. Given the previous estimation that Asian (dog) RABV variants appear at the root of a time-scaled tree [Reference David29], Arctic RABV variants probably evolved via a northerly spread, as recently proposed [Reference Nadin-Davis19]. Furthermore, Arctic-like viruses appear to have compartmentalized earlier than Arctic viruses.
Phylogenetic methods which apply a molecular clock must include the uncertainty inherent in estimation of substitution rates, and in turn, reflected in the confidence limits of dating divergence times. The recent development of methods allowing for a relaxed molecular clock to be used has obvious advantages to those which use a strict molecular clock which assumes a single substitution rate across a phylogeny [Reference Drummond25]. The TMRCA estimates here (and the dating of the divergence of the Arctic/Arctic-like lineage) are greater than those previously estimated using a strict molecular clock [Reference David29], although the topology of both time-scaled trees is congruent. In this analysis, the high posterior probabilities associated with major nodes (1 for all major branching events) allows a high degree of confidence to be placed on the direction and relative timing of divergence events in a tree, even when the confidence limits of a timed event (i.e. a TMRCA) are quite large.
According to the available epidemiological data, Arctic-1 RABV entered Ontario from Arctic regions during the 1950s [Reference Tabel31]. Unfortunately, the progenitor viruses for this event are not available for comparison, but observations within the Arctic-2 and Arctic-3 groups does suggest that substitutions which differentiate Arctic-1 RABV from these groups are unlikely to have accumulated within this time-frame; a fact very much supported by our time-scaled phylogeny. Our age estimate suggests that the Arctic-1 group is the oldest within the Arctic RABV lineages. In this context it is interesting to note that the 8486GRO, isolated from Greenland in 1981 [Reference Kissi, Tordo and Bourhy11], belongs to the Arctic-1 group, whereas six Greenland RABV isolated more recently (1990–2002) all belong to the Arctic-3 group (designated Arctic-2b; see ref. [Reference Mansfield12]). Mansfield et al. [Reference Mansfield12] proposed that this could be a result of the short-term incursion of Arctic-1 viruses from Canada into Greenland, without sustained circulation. Alternatively, these viruses may have co-circulated undetected within that (and potentially other) region for many years. As the density of the human population in northern Canada and Greenland is quite low, many epizootic events are undoubtedly not detected, such that only a limited number of isolates are available from those territories for comparison.
Whether genetic markers exist which provide evidence for RABV adaptation to arctic host species is currently unclear. The RABV genes analyzed to date appear to be under a high degree of purifying selection [Reference Davis28, Reference Davis, Bourhy and Holmes30]. Several substitutions have been suggested to be specific for Arctic-1 viruses, such as the T/A379 in the nucleoprotein; L/V183, Q/P244 and A/S483 in the glycoprotein [Reference Nadin-Davis, Casey and Wandeler13]. However, they are not conserved within the large dataset of Arctic viruses from this study.
Similarly, the isolates from Alaska, that form the Arctic-4 group in our study, could represent a lineage of RABV that has circulated in Alaska for a long time (the divergence date according to our estimations goes to the beginning of 20th century) but were not present in previous studies, due to the lack of adequate surveillance. Viruses of this group were isolated during our study only from red foxes and dogs in the south-western Alaska. Further observation is needed to establish their distribution and circulation patterns.
The geographical area of Arctic RABV variants circulation is separated from that of Arctic-like RABV variants by a wide band of conifer taiga forests, lying to the south of the tundra-forest zone in Siberia. These forests are considered largely free of rabies, perhaps because the population density of wild canids appears too low to maintain active virus circulation [Reference Kuzmin16]. Among the Arctic-like viruses, those genetically most closely related to the Arctic viruses currently circulate in the Middle East and southern Asia (Arctic-like-1). However, where the progenitor virus circulated before the incursion to the Arctic is unknown. The incursion probably happened several hundred years ago, and no isolates are available from that time. Our finding of the related virus in Iraq provides the most westerly isolation point to date. However, it is uncertain whether this represents a recent emergence, a sporadic incursion, or part of sustained circulation. Extensive rabies surveillance in Israel did not show the presence of Arctic-like viruses [Reference David29].
Circulation areas of Arctic-like-1 and Arctic-like-2 viruses are separated from each other, as well as being separated phylogenetically and historically. Mountainous areas of the Himalayas and Tibet may serve as natural barriers for viral populations circulating in wild animals. Political and cultural distinctions of human society might prevent translocations of these viruses with companion dogs. Unfortunately, no RABV sequences are available from northern China. We can expect that Arctic-like-2 viruses might be found there, as they circulate in the bordering parts of Russia and Korea. In central and southern China, no Arctic-like viruses have been recovered to date [Reference Meng32].
Numerous human migrations into northern territories have occurred over the last centuries, bringing with them the possible spread of rabies. Arctic (and perhaps the initial pre-Arctic) RABV appear best adapted to the principal Arctic host, the Arctic fox, whereas other variants may not be able to establish long-term circulation, due perhaps to a relative reduction in fitness. The mechanisms of such adaptation, which may occur on the organism or population level, are unclear. For example, we do not know why the ‘cosmopolitan’ canine RABV, broadly disseminated in moderate latitudes of Eurasia, the Americas and Africa, does not circulate in the Arctic. Further, we do not understand why the ‘cosmopolitan’ lineage could not establish circulation in southern and eastern Asia (except for several isolates described in China, which were similar to the vaccine RABV strains, suggesting a probability that they could derive from poorly attenuated veterinary vaccines [Reference Meng32]). Similarly, why Arctic viruses do not circulate in areas besides the Arctic is unresolved. Multiple situations may have arisen for Arctic RABV to be translocated with rabid dogs.
Sporadic rabies outbreaks, which could be caused by viruses other than those of the Arctic lineage, have been described in the Far North. For example, an epizootic among dogs in the delta of the Anadyr River (Chukotka Peninsula) occurred between 1953 and 1956, causing four cases of human rabies. Thereafter, no such epizootics or human mortality was reported from Chukotka, whereas rabies among the Arctic foxes, with limited spillover into dogs, was repeatedly reported [Reference Savitsky33].
Arctic RABV variants may have a lower pathogenicity than other canine RABV variants [Reference Crandell and Baer34], although no truly robust study has been performed to date. Among 99 trapped Arctic foxes in Alaska, only one had RABV in neuronal tissue, but five had virus-neutralizing antibody in serum, suggesting a previous exposure or abortive infection [Reference Ballard35]. In an experimental study, the mortality of Arctic foxes inoculated peripherally with the homologous virus strain was less than 100% [Reference Follmann, Ritter and Donald36].
The reduced pathogenicity of Arctic RABV to humans has also been repeatedly proposed, but never proven. For example, residents of the Ustinsky settlement in delta of Lena River lost all of their dogs in August, 1855 as the result of a disease clinically compatible to rabies. However, no cases of human rabies were recorded, even after bites from rabid dogs [Reference Ognev2]. The limited evidence for a reduced virulence of Arctic RABV includes the presence of RABV neutralizing antibody in the serum of an Alaskan trapper who, although having never been vaccinated against RABV, had trapped animals for 47 years [Reference Follmann, Ritter and Beller37]. In general, very few cases of human rabies have been reported from the north. Most reported cases have occurred after severe, multiple animal bites (reviewed in ref. [Reference Mansfield12]) but one case was reported as the result of the skinning of a dead Arctic fox [Reference Shmit38]. Unfortunately, most of the viruses that have caused human rabies in the Arctic are unavailable for typing. However, one isolate (RVHK; Fig. 1) was proven to be a typical Arctic RABV belonging to the Arctic-3 group [Reference Kuzmin16]. In that case, an adult man who was severely bitten on the head, shin and hand by a wolf developed rabies after an incubation period of 24 days, despite initiation of post-exposure prophylaxis using a commercial rabies vaccine (immunoglobulin was unavailable). Therefore, at least severe exposure to the Arctic virus does result in rabies in humans. Besides reduced pathogenicity, other potential explanations of the infrequent reports of human rabies from the north include sparse human populations and protective outer garments used in the cold climate that protect against bites. Moreover, in many parts of the Far North surveillance is very limited, and human cases may go underreported.
In contrast, there is no evidence to suggest that Arctic-like viruses circulating in southern and eastern Asia show an altered pathogenicity for humans. In India, more than 20 000 human rabies cases occur every year. Arctic-like viruses are broadly distributed in this country [Reference Nagarajan18, Reference Nadin-Davis19] and two human isolates of Indian origin are presented in our study, RV61 and AY956319. The last virus was responsible for three human rabies cases following organ transplantation in Germany.
Another interesting example is the extensive dog rabies outbreak that occurred in the Chita Province of eastern Siberia during the late 1970s–early 1980s, that caused multiple human deaths. The responsible RABV variant had a different reactivity pattern to mAbs than those viruses circulating locally among the red fox and corsac fox populations. It was proposed that the new variant may have been introduced to Chita from some other enzootic territory [Reference Selimov9]. After the outbreak was eliminated, no other dog or human cases were reported from that area for decades. From phylogenetic analysis, both virus lineages, the locally circulating (304c and 248c) and the newly introduced one (994 dog), belong to the Arctic-like-2 group (Fig. 1). Further, the introduced strain is mostly related to isolates from the Far East (857r and Komatsugawa) rather than to the viruses from eastern Siberia. In contrast to eastern Siberia, human rabies occurs in the Far East quite frequently. In this example, we again encounter a potential indirect suggestion for distinct pathogenicity of these RABV variants.
Unfortunately, specific substitutions which may contribute to altered infection dynamics and consequent epidemiology of Arctic RABV variants are unknown. Even single amino-acid substitutions may be critical for pathogenicity of RABV (such as amino acid at position 333 in the glycoprotein ectodomain [Reference Dietzschold39]). Climate change due to global warming raises further questions on future prospectives regarding Arctic rabies. For example, one possible repercussion is the potential extension of the geographic range of the red fox. The red fox is a well-established reservoir for RABV, and the possible northerly extension of its range may influence the epidemiology of Arctic rabies. Will Arctic RABV variants circulate in this species as readily as they circulate among Arctic foxes? If so, should we expect adaptive changes that will affect either the circulation properties of the virus and its pathogenicity for animals and humans? Additional surveillance, greater number of specimens, and extensive evolutionary analyses are necessary to address these questions.
ACKNOWLEDGEMENTS
The findings and conclusions in this report are those of the authors and do not necessarily represent the views of the funding agency (CDC). We are grateful to Drs L. Castrodale (Alaska State Department of Health) and M. Westcott (Alaska State Virology Laboratory) for providing the rabies-positive samples from Alaska. We also thank Dr. L. Fuhrmann Veterinary Laboratory Europe (Kirchberg, Germany) for providing the dog RABV isolate from Iraq.
DECLARATION OF INTEREST
None.
