Recent US bluetongue virus serotype 3 isolates found outside of Florida indicate evidence of reassortment with co-circulating endemic serotypes

Since 1999, 11 serotypes of bluetongue virus (BTV) similar to Central American or Caribbean strains have been isolated in the southeastern United States, predominantly in Florida. The majority of the incursive serotypes have remained restricted to the southeastern US. In recent years, BTV serotype 3 (BTV-3) has been isolated in areas increasingly distant from Florida. The current study uses whole genome sequencing of recent and historical BTV-3 isolates from the US, Central America and the Caribbean with additional sequences from GenBank to conduct phylogenetic analyses. The individual segments of the BTV genome were analysed to determine if recent BTV-3 isolates are reassortants containing genomic segments from endemic US serotypes or if they retain a majority of Central American/Caribbean genotypes. The analyses indicate that BTV-3 isolates Mississippi 2006, Arkansas 2008 and Mississippi 2009 are closely related reassortants that contain five to six genomic segments that are of US origin and two to three segments of Central American/Caribbean origin. In contrast, the BTV-3 South Dakota 2012 isolate contains seven genomic segments that are more similar to isolates from Central American and the Caribbean. These different evolutionary histories of the BTV-3 isolates suggest that there are at least two different lineages of BTV-3 that are currently circulating in the US.

Bluetongue virus is transmitted by several species of biting midge of the genus Culicoides [2,3]. BTV is the etiological agent of bluetongue disease (BTD), an economically important disease of domestic and wild ruminants. Impacts of BTD on the livestock industry are not limited to the production losses associated with the mortality/morbidity of BTD but also include international restrictions on the trade of animals from areas with BTD or specific BTV serotypes [4,5] BTD was first described in South Africa in the early 1900s [6,7]. Initially, it was believed that BTV would emerge from Africa and devastate the world's sheep population. However, as additional serotypes of BTV were identified on other continents without the presence of severe disease it was realized that BTV emergence was not a recent event [8][9][10]. Currently, at least 29 serotypes of BTV exist worldwide [11,12]. In tropical and subtropical regions that support continuous vector populations and circulation of endemic BTV serotypes disease outbreaks are uncommon [10]. In these areas, outbreaks of disease are generally associated with the introduction of a new serotype often from a neighbouring region.
BTV was first isolated in the United States in the early 1950s, although BTD, known as 'sore muzzle', had been described earlier. By the early 1980s, four serotypes of BTV (10, 11, 13 and 17) were known to be endemic throughout the western and southern United States [8][9][10]13]. BTV serotype 2 (BTV-2) was first detected in Florida in 1982 and has since become endemic in the southeastern United States [14,15]. Only one presumably imported isolate of BTV-2 has been reported in California [16,17]. In Central America and the Caribbean, serological typing of BTV isolates from the 1980s identified serotypes 1, 3, 4, 6, 8, 12 and 17 as endemic. More recently, sequencing of the serotype-specific segment of 1990s isolates from Central America and the Caribbean added six BTV serotypes (10, 11, 13, 14, 19 and 22) to this list [18]. Culicoides insignis is considered to be the primary vector of BTV in Central America and the Caribbean [2,8,10,19]. C. insignis is also found throughout southern Florida, while Culicoides sonorensis is believed to be the primary vector of BTV in the rest of the United States. Data compiled from USDA, APHIS, National Veterinary Service Laboratories annual reports show that 11 invasive BTV serotypes were first isolated in the US between 1999 and 2015 [18,[20][21][22][23][24][25][26][27][28][29][30][31][32][33][34][35][36][37][38][39] (see Table S1, available in the online version of this article). Although, some of these isolations were from sick animals, many came from healthy animals being tested for export purposes. Nine of the 11 invasive serotypes were first isolated in Florida (3,5,6,9,14,18,19,22,24). Two additional serotypes, BTV-1, first isolated in 2004 in Louisiana [40] and BTV-12, first isolated in Texas in 2008 [32], were later isolated in Florida. While the US does not conduct active surveillance of circulating BTV serotypes in all areas, the available data suggests that Florida may be a common point of entry for invasive BTV serotypes. Many of the exotic serotypes continue to be sporadically isolated only in Florida, suggesting that either their persistence is due to the presence of a competent Florida vector or that the same serotypes are repeatedly introduced and then die out. In contrast, BTV-3 was first detected in Florida in 1999 and was repeatedly isolated over the next several years. However, since 2006 BTV-3 has been isolated in Mississippi, Arkansas, South Dakota and most recently in Texas [31,33,34,39].
One reason for this increase in distribution of BTV-3 may be due to reassortment. During co-infection of the same animal or cell by two genetically different viral strains or serotypes of a segmented virus can create novel combinations of genetic segments during packaging of progeny viruses. These new reassortant viruses may display different phenotypic characteristics than either parental virus [41][42][43]. These new combinations of genetic elements can result in an increase or decrease in pathogenicity or transmissibility, in the ability to infect a new host or vector species, or in the persistence of an exotic strain [43].
This study uses whole genome sequencing and phylogenetic analyses of BTV isolates from the US, Central America and the Caribbean to investigate the hypothesis that BTV-3 has reassorted with co-circulating endemic strains enabling BTV-3 to utilize additional vectors with a larger geographic distribution and to extend from Florida into the Northern Plains.

The outer capsid
The phylogenetic tree for segment 2, the determinant of serotype, shows sequences grouping according to serotype, as expected (Fig. 1a). All BTV-3 isolates form a single, wellsupported clade with 84-86 % nucleotide identity (NI). The MAM clade groups with other BTV-3 sequences from Florida, Central America and the Caribbean, as well as South Dakota 2012. The nearest well-supported relative of the MAM clade for VP2 is BTV-3 Panama 1989. BTV-3 South Dakota 2012 is most closely related to sequences from BTV-3 Martinique 2010 and Barbados 1988 for segment 2 which forms the sister group to the BTV-3 Florida strains. For segment 6, the four isolates of interest cluster together with 95-96 % NI to BTV14 Tobago 1989 (Fig. 1b). In trees for both segments nearly all of the BTV-3 Florida isolates are found in a well-supported clade separate from the recent isolates. This suggests that for the isolates of interest, BTV-3 segments 2 and 6 are the descendants of a Central American/ Caribbean BTV-3 ancestor.

The capsid
The geographic origin of the segment 3 (VP3) sequences for the recent BTV-3 isolates is somewhat equivocal due to the majority of BTV-3 sequences grouping together (see  South Dakota 2012 is most closely related to BTV-18 Guatemala (99 % NI) and BTV-14 Guatemala (98 % NI) for segment 9.

Non-structural proteins
The non-structural proteins are encoded by segments 5 (NS1), 8 (NS2) and 10 (NS3/3A). In the phylogenetic tree for segment 5, (Fig. 4a), the four isolates of interest are all found within the same large, well-supported clade that includes US BTV-10, 11, 17, four isolates from Central is also included in the same larger group of sequences, it is part of a polytomy and its closest relative is undetermined. The phylogeny for segment 8, shown in Fig. 4b, identifies the MAM clade as the well-supported sister group to all other BTV strains in the study except those from Australia and India. The placement of this clade suggests that the most recent ancestor for these isolates has not been sequenced. BTV-3 South Dakota 2012, however, shows a close relationship to sequences from Florida, Central America and the Caribbean with the sequence of BTV-9 Honduras 1991 being its closest relative (97 % NI). The segment 10 analysis places the MAM clade as the sister group to a large clade of US BTV-10, 11, 13, 17 isolates with 95-97 % NI (Fig. 4c). BTV-3 South Dakota 2012, in contrast, falls within a clade consisting of BTV-3, BTV-11 and BTV-6 isolates from

DISCUSSION
The current study uses whole genome sequencing and phylogenetic analyses of recent and historical isolates from the US, Central America and the Caribbean to determine if recent US BTV-3 isolates are reassortants with endemic US serotypes or if they retain a Central American/Caribbean signature. The analyses show that BTV-3 isolates from Mississippi 2006, Arkansas 2008 and Mississippi 2009 have very similar evolutionary histories that have resulted in the acquisition of a majority of genomic segments of US endemic serotype origin. In contrast, the BTV-3 South Dakota 2012 isolate has a majority of genomic segments that are more similar to BTV-3 sequences from Florida, Central America and the Caribbean. These different histories of the BTV-3 isolates suggest that there are at least two different lineages of reassortant BTV-3 currently circulating in the US (see Table 1).
Early studies of BTV isolates used electropherotypes or the pattern of RNA segments run on a polyacrylamide gel to show that significant variation existed between isolates of the same and different serotypes, as well as within isolates from the same area, outbreak or animal [44,45]. Oligonucleotide fingerprinting of BTV RNA from US prototype isolates demonstrated that similarities between segment fragments from different serotypes were the result of reassortment in the field [46,47]. Further research on field isolates of BTV has shown that reassortment is common where multiple serotypes circulate and/or live attenuated vaccines have been used [47,48].
Reassortment of segmented viruses in cell culture has been shown to be essentially at random when the parental viruses have equal fitness [43,49]. In studies where viruses were not matched for fitness, one constellation of viral segments became dominant with a small number of combinations being found at low frequencies [42,43,50]. Often when two serotypes were co-infected, only one serotype was found among the progeny [41,42,50,51]. These data suggest that while all segments may be able to reassort, selection for fitness in the animal or vector will determine which reassortants are passed on [43,52,53]. In the current study, the BTV-3 Mississippi and Arkansas isolates only retained the BTV-3 segment that confers serotype, segment 2. In the field, previous exposure of livestock to endemic serotypes induces a protective response against subsequent exposures to these serotypes. Since serotype 3 is novel in the US and livestock outside of Florida are naïve, the segments that Segments that code for non-structural proteins tend to be conserved due to negative selection that may be linked to functional constraint of these proteins. However, slight differences in fitness may lead to selection of specific segments. Ramig et al. [50] suggest that reassortant progeny viruses from co-infection of cell cultures contain more segments from the parental virus that was infected at a higher multiplicity of infection. This previous study, however, analysed reassortants by electropherotype. By repeating this former study using current sequencing methods, it may become possible to identify the sequence of parent virus infections in reassortant viruses. In the current study, the differences in evolutionary histories of the BTV-3 Mississippi and Arkansas isolates and the South Dakota isolate may reveal that in South Dakota only BTV-3 was circulating in the affected animals while multiple serotypes were circulating further south. Future studies using the BTV reverse genetics system and sequences from recent BTV-3 and endemic isolates may allow us to tease apart the influence of the different segments on virulence, replication and transmissibility in animals and vectors.
Unfortunately, due to a lack of continual surveillance of BTV serotypes throughout the US, isolations are only made during an outbreak of BTD or when testing for export purposes and usually only serotype is determined at the time. This does not allow us to determine with any precision where and when a virus has undergone reassortment or which parental strains were involved. Whole genome sequencing of new BTV-3 isolates is needed in order to   continue tracking its evolution and persistence. Monitoring serotypes circulating in vectors, livestock and wild ruminants in areas where BTV-3 has previously been isolated will provide the information necessary to determine if recent isolates are the result of transient incursions from the southeastern US or if BTV-3 will become the next North American endemic serotype.

METHODS
A total of 27 bluetongue isolates of multiple serotypes (17 from BTV-3) from the United States, Central America and the Caribbean were sequenced for this study (see Table 2). BTVs not isolated in the United States were obtained from the Inter-American Bluetongue Project and the Onderstepoort Veterinary Institute virus library. Isolates from the United States were obtained from the National Veterinary Services Laboratories or the Arthropod-borne Animal Disease Research Unit reference collection. Viruses were typically isolated in embryonated chicken eggs or cattle pulmonary artery endothelial (CPAE) cells (ATCC CCL-209), followed by one to as many as seven passages in baby hamster kidney (BHK-21) (ATCC CCL-10) or CPAE cells. Total RNA was extracted from cells as previously described [54]. Viral double-stranded RNA was then purified by lithium chloride differential precipitation as described in [55] and subjected to whole genome sequencing using the sequence-independent amplification procedure described by [56] with modifications as described previously [57]. Standard Flow-gram Format files were imported into Geneious 6.0 (Biomatters) for contig creation. Partial contigs were assembled and BLASTed against the NCBI nucleotide database to determine reference sequences that were then used for reference-based assemblies.

Phylogenetic analysis
In order to encompass the currently known diversity of isolates from the United States, Central America and Caribbean regions additional BTV segment sequences were downloaded from GenBank. The genome segments from serotypes 1 and 2 of the closely related orbivirus, epizootic haemorrhagic disease virus (EHDV) were used to root the trees. The newly sequenced BTV genomes have been deposited in GenBank, KY091901-KY092170 (Table 2). GenBank accession numbers for all viral segments included in the study are located in Table S2.  congruence between phylogenetic trees of individual genome segments. When all of the segments for a viral strain share the same ancestor-descendent relationships the topologies of the phylogenetic trees of each segment will be very similar or identical. A segment from a co-infecting strain will have different ancestor-descendent relationships and therefore the closest relatives will differ in the phylogenetic tree for that segment.