Identification of Novel Rosavirus Species That Infects Diverse Rodent Species and Causes Multisystemic Dissemination in Mouse Model

While novel picornaviruses are being discovered in rodents, their host range and pathogenicity are largely unknown. We identified two novel picornaviruses, rosavirus B from the street rat, Norway rat, and rosavirus C from five different wild rat species (chestnut spiny rat, greater bandicoot rat, Indochinese forest rat, roof rat and Coxing's white-bellied rat) in China. Analysis of 13 complete genome sequences showed that “Rosavirus B” and “Rosavirus C” represent two potentially novel picornavirus species infecting different rodents. Though being most closely related to rosavirus A, rosavirus B and C possessed distinct protease cleavage sites and variations in Yn-Xm-AUG sequence in 5’UTR and myristylation site in VP4. Anti-rosavirus B VP1 antibodies were detected in Norway rats, whereas anti-rosavirus C VP1 and neutralizing antibodies were detected in Indochinese forest rats and Coxing's white-bellied rats. While the highest prevalence was observed in Coxing's white-bellied rats by RT-PCR, the detection of rosavirus C from different rat species suggests potential interspecies transmission. Rosavirus C isolated from 3T3 cells causes multisystemic diseases in a mouse model, with high viral loads and positive viral antigen expression in organs of infected mice after oral or intracerebral inoculation. Histological examination revealed alveolar fluid exudation, interstitial infiltration, alveolar fluid exudate and wall thickening in lungs, and hepatocyte degeneration and lymphocytic/monocytic inflammatory infiltrates with giant cell formation in liver sections of sacrificed mice. Since rosavirus A2 has been detected in fecal samples of children, further studies should elucidate the pathogenicity and emergence potential of different rosaviruses.


Introduction
Picornaviruses are positive-sense, single-stranded RNA viruses with icosahedral capsids. They infect various animals and human, causing various respiratory, cardiac, hepatic, neurological, mucocutaneous and systemic diseases [1,2]. Based on genotypic and serological characterization, the family Picornaviridae is currently divided into 29 genera with at least 50 species. Among the various picornaviruses belonging to nine genera that are able to infect humans, poliovirus and human enterovirus A71 are best known for their neurotropism and ability to cause mass epidemics with high morbidities and mortalities [3,4]. Picornaviruses are also known for their potential for mutations and recombination, which may allow the generation of new variants to emerge [5][6][7][8][9][10].
Emerging infectious diseases like avian influenza and coronaviruses have highlighted the impact of animal viruses after overcoming the inter-species barrier [11][12][13][14][15]. As a result, there has been growing interest to understand the diversity and evolution of animal and zoonotic viruses. For picornaviruses, numerous novel human and animal picornaviruses have been discovered in the past decade [1,[16][17][18][19][20][21][22][23][24][25][26][27]. We have also discovered a novel picornavirus, canine picodicistrovirus (CPDV), with two internal ribosome entry site (IRES) elements, which represents a unique feature among Picornaviridae [28]. Moreover, novel picronaviruses were identified in previously unknown animal hosts such as cats, bats and camels [29][30][31], reflecting our slim knowledge on the diversity and host range of picornaviruses. The discovery and characterization of novel picornaviruses is important for better understanding of their evolution, pathogenicity and emergence potential.
Although rodents can be infected by several picornaviruses, the picornaviral diversity is probably underestimated, given the enormous species diversity of rodents. Moreover, little is known about the pathogenicity of the recently discovered rodent pricornaviruses, such as rodent stool-associated picornavirus (rosavirus) A1, mouse stool-associated picornavirus (mosavirus) A1, Norway rat hunnivirus and rat-borne virus (rabovirus A) [32,33]. In this report, we explored the diversity of picornaviruses among rodents in China and discovered two potentially novel picornaviruses, "Rosavirus B" and "Rosavirus C". While rosavirus B was detected in the street rat, Norway rats, rosavirus C was detected in five different wild rat species, suggesting potential interspecies transmission. Their complete genome sequences were determined, which showed that "Rosavirus B" and "Rosavirus C" represent two novel picornavirus species distinct from Rosavirus A. Rosavirus C isolated from cell culture causes multisystemic diseases in a mouse model, with histopathological changes and positive viral antigen expression in lungs and liver of infected mice.

Genome organization and coding potential of rosavirus B and C
A total of 13 complete genomes from samples of four different wild rodent species (chestnut spiny rat, Coxing's white-bellied rat, roof rat and Indochinese forest rat) positive for "Rosavirus C" and one street rodent species (Norway rat) positive for "Rosavirus B" were sequenced directly from the positive respiratory or alimentary samples and characterized. These 13 strains were selected because they were detected from different rodent species or geographical locations (Hong Kong, Hunan and Guangxi) to allow comparison between host species-or geographically distinct strains. The G + C contents of the three rosavirus B and 10 rosavirus C genomes range from 50 to 53%, with genome size 8639 to 9094 bases, after excluding the polyadenylated tract (Table 2). However, the genome sizes of some strains may be larger, as further sequencing of the ends may have been hampered by secondary structures. They share similar genome organization typical of picornaviruses, with UTR at both 5' (470-752 bases) and 3' (693-948 bases) ends, and a large open reading frame of 7449-7476 bases, which encodes   potential polyprotein precursors of 2482-2491 aa known to be cleaved by virus-encoded proteases. The predicted protease cleavage sites at P1 (encoding capsid proteins) P2 and P3 (both encoding non-structural proteins) are shown in Table 3. Notably, "Rosavirus B" differed from

Phylogenetic analyses
Phylogenetic trees constructed using the aa sequences of P1, P2 (excluding 2A) and P3 (excluding 3A) of rosavirus B and C are shown in Fig 2 and the corresponding pairwise aa identities are shown in Table 2. 2A and 3A regions were excluded to avoid bias due to poor sequence

Genome features of rosavirus B and C
Comparison of genome features of rosavirus B and C to those of rosavirus A is summarized in Table 4. The conserved sequence Yn-Xm-AUG is present in the 5'UTR of rosavirus B and C. While Y7-X19-AUG is found in rosavirus A, the number of Y (6 or 7) and X (19-21) varies among rosavirus B and C. The putative translation initiation sites were contained by an optimal Kozak context (RNNAUGG), with in frame AUG at position 471 to 753. The 5'UTR of many picornaviruses possesses an internal ribosomal entry site/segment (IRES) which is responsible for directing the initiation of translation in a cap-independent manner, and requires both canonical translation initiation and IRES trans-acting factors [3,34]. Similar to rosavirus A [32], rosavirus B and C also contained a type II-like IRES with stem loops, major domains [19,27,[35][36][37][38][39][40] and conserved motifs (Fig 3). However, domain E was only present in rosavirus C strains, RASK8F, RATLC11A, NFSM6F and RASM14A, and rosavirus B strain RNYL1109081R, but not other strains. The pyrimidine-rich region was located near 3' end of 5'UTR. One to three stem-loop structures were present upstream of the start codon and/or between the pyrimidine-rich region and start codon of the polyprotein [41,42]. The predicted "VP0" of rosavirus B and C are probably cleaved into VP4 and VP2 based on sequence alignment [32]. In contrast to rosavirus A of which the VP4 possessed the myristylation site, GXXX [ST], involved in capsid assembly or virus entry [43], such myristylation site is absent Table 3. Coding potential and putative proteins of "Rosavirus B" and "Rosavirus C" compared to Rosavirus A. in rosavirus C and variably present in rosavirus B (present in strains RNCW0602091R and RNCW1002091R but absent in strain RNYL1109081R). The predicted 2A of rosavirus B and C exhibited 45.1% aa identities to that of rosavirus A, possessed the conserved H-box/NC involved in cell proliferation control, but not Asn-Pro-Gly-Pro (NPGP) motifs [44,45]. Their predicted 2C possessed GXXGXGKS motif for NTP-binding [46] and DDLXQ motif for putative helicase activity [47]. Their predicted 3C pro contained the catalytic triad H-D-C [48], conserved GXCG motif in the protease active site and GXH motif [49,50]. Their predicted 3D pol contained conserved KDE[LI]R, GG[LMN]PSG, YGDD and FLKR motifs [51], although the second Gly was replaced by Ala in GG[LMN]PSG.    Although rosavirus C were detected in five different rodent species from Hong Kong, Hunan and Guangxi, no major distinct genome features were identified between strains from different rodent species or geographical locations. Yet, the two strains from Coxing's white-bellied rat from Hunan (NCHN06IO) and Guangxi (NCGX12IN), were always clustered together in the P1, P2 and P3 trees, suggesting that geographically distinct strains may be genetically closely related (Fig 2). These two strains from mainland China possessed a total of 432 unique nucleotide substitutions over the entire genomes compared to the other eight rosavirus C strains from Hong Kong.

Estimation of synonymous and non-synonymous substitution rates
Using available rosavirus B and C genome sequences for analysis, the Ka/Ks ratios for various coding regions were estimated ( Table 5). The Ka/Ks ratios for most coding regions were low, supporting purifying selection.

Isolation of rosavirus C in 3T3 cells
Of the various cell lines inoculated with the 11 rodent samples positive for rosavirus B (three samples) or rosavirus C (eight samples), viral replication was detected by RT-PCR in the lysates of 3T3 cells infected by rosavirus C strain RASM14A, with viral load of 4.5×10 8 copies/ml (3.2 × 10 3 TCID 50 ) at day 7. Cytopathic effect (CPE), mainly in the form of rounded and refractile cells rapidly detaching from the monolayer, was also observed in infected 3T3 cells five  days after inoculation, which showed viral VP1 expression by immunofluorescence in 40% of cells (Fig 4). Electron microscopy of ultracentrifuged cell culture extracts from infected 3T3 cells showed the presence of non-enveloped viral particles of around 25-30 nm in diameter compatible with those described for members of the family of Picornaviridae (Fig 4).

Seroprevalence of rosavirus B and C
To determine the seroprevalence of rosavirus B and C among different rodent species, western blot analysis was performed on available rodent serum samples to test for specific antibodies against rosavirus B or C recombinant VP1 protein. The purity of the recombinant VP1 proteins was confirmed by the dominant band observed at the predicted size of 40 kDa upon SDS polyacrylamide gel electrophoresis. Anti-rosavirus B antibodies were detected in two (3.3%) of 61 Norway rats from Hong Kong whereas anti-rosavirus C antibodies were detected in three (4.3%) of 70 Indochinese forest rats from Hong Kong and three (7.5%) of 40 Coxing's whitebellied rats from Hunan Province. However, the antibodies from Norway rats against rosavirus B were likely of low levels, as reflected by the relatively weak band observed (Table 1 and Fig 5).
Using sera with anti-rosavirus B antibodies against rosavirus C recombinant VP1 protein and Novel Rosavirus in Rodents sera with anti-rosavirus C antibodies against rosavirus B recombinant VP1 protein, no cross reactivities were observed between the two proteins. Neutralization assays showed that five of the six rats with anti-rosavirus C antibodies by western blot analyses were positive for neutralizing antibodies against rosavirus C RASM14A with titer 1:10 to 1:40.

Pathogenicity of rosavirus C in challenged mice
We attempted to study the pathogenicity in mice challenged with rosavirus C RASM14A isolated from infected 3T3 cells. To mimick the fecal-oral route of transmission typical of many picornavirus infections, oral inoculation of rosavirus C RASM14A was performed on 21 fourday-old suckling mice. One of the suckling mice was eaten by its mother on day one post-challenge. All the remaining 20 suckling mice survived after viral challenge till sacrifice, but some mice exhibit transient roughening of hair two to three days after challenge. Among the nine mice sacrificed on day 3 post-challenge, rosavirus C RASM14A was detected in the intestine and lung of all nine mice, kidney of one mouse, and spleen and liver of three and four mice respectively by RT-PCR. Among the five mice sacrificed on day 7 post-challenge, rosavirus C RASM14A was detected in the intestine of all five mice, liver and lung of four mice, and spleen and kidney of two and one mice respectively by RT-PCR. Among the three mice sacrificed on day 14 post-challenge, rosavirus C RASM14A was detected in the lung of one mouse by RT-PCR. Among the three mice sacrificed on day 21 post-challenge, rosavirus C RASM14A was detected in the lung of one mouse by RT-PCR. Anti-rosavirus C VP1 antibody was detected in none of the mice sacrificed on day 3 and 14, two of the five mice sacrificed on day 7, and one of the three mice sacrificed on day 21 by Western blot assay (Table 6). qRT-PCR of tissues positive by RT-PCR showed high levels of mean viral RNA copies in lung (2.1 ×10 6 copies/g) and intestine (2.8 ×10 5 copies/g) tissues of mice sacrificed on day 3 (Fig 6).
Since some picornaviruses have been associated with neurovirulence, intracerebral inoculation was also performed on another group of 21 one-day old suckling mice. One of the suckling mice was eaten by its mother on day one post-challenge. All the remaining 20 suckling mice survived after viral challenge till sacrifice. Among the nine mice sacrificed on day 3 post- Novel Rosavirus in Rodents challenge, rosavirus C RASM14A was detected in the lung, liver, brain and spleen of eight mice, and intestine and kidney of four mice by RT-PCR. Among the five mice sacrificed on day 7 post-challenge, rosavirus C RASM14A was detected in the lung of three mice, brain and liver of five mice, intestine of two mice, and spleen and kidney of four mice by RT-PCR. Among the three mice sacrificed on day 14 post-challenge, rosavirus C RASM14A was detected in the brain, intestine, liver and lung of one mouse by RT-PCR. Among the three mice sacrificed on day 21 post-challenge, rosavirus C RASM14A was detected in the lung of one mouse by RT-PCR. Anti-rosavirus C VP1 antibody was detected in none of mice sacrificed on day 3, four of the five mice sacrificed on day 7, and two of three mice sacrificed on day 14 and 21 by Western blot assay (Table 6). qRT-PCR of tissues positive by RT-PCR showed high levels of mean viral RNA copies in various tissues (0.8 to 9.6 ×10 5 copies/g) of mice sacrificed on day 3 (Fig 6).
Histological examination of various organs revealed alveolar fluid exudation, interstitial infiltration, alveolar fluid exudate and wall thickening in lung sections of mice sacrificed on day 3 after oral or intracerebral inoculation. Moreover, hepatocyte degeneration and lymphocytic/monocytic inflammatory infiltrates with giant cell formation were observed in liver sections of mice sacrificed on day 3 after oral or intracerebral inoculation (Fig 7). Immunohistochemical staining with guinea pig anti-serum against rosavirus C VP1 protein antibody revealed viral antigen expression in bronchiolar and bronchial epithelial cells in lung sections, and hepatocytes in liver sections (Fig 7).

Discussion
We report the discovery of two novel rodent picornaviruses, rosavirus B and C, from six rodent species in southern China. Though being phylogenetically most closely related to Rosavirus A, "Rosavirus B" and "Rosavirus C" should represent two novel species distinct from Rosavirus A under the genus Rosavirus, according to the criteria for International Committee on Taxonomy of Viruses species demarcation for different members of Picornaviridae [52]. Rosavirus B and C also exhibited different genome features when compared to rosavirus A. Notably, the absence of myristylation site in VP4 of rosavirus C and its varying presence in rosavirus B is intriguing. The VP4 myristylation site has been shown to play a role in localization of the capsid protein Table 6. RT-PCR and western blot analysis of mice challenged with rosavirus C RASM14A.

Route of infection
Day at which mice were sacrificed (total no. of mice) for cellular entry and permeability [43,53]. Its absence in some rosavirus strains suggests that rosaviruses may utilize alternative strategies for capsid localization on cellular targets. In addition, the variations in Yn-Xm-AUG sequence in 5'UTR may also suggest different translational dynamics among rosaviruses. Besides phylogenetic and genomic evidence, the two viruses infect different host, with "Rosavirus B" infecting Norway rats (a street rat) and "Rosavirus C"  known to infect different rodents [32,33,54]. Rosavirus A was first discovered in a wild canyon mouse (Peromyscus crinintus) in California [32]. Subsequently, a variant, rosavirus 2, was detected in the fecal specimens of children in the Gambia [26], which prompted further studies to investigate potential transmission of rosavirus from rodents to humans. In this study, the observed low Ka/Ks ratios in various coding regions supported that Norway rats (street rats) in Hong Kong, and wild rats across Hong Kong and mainland China, are natural reservoirs of rosavirus B and C respectively. In particular, Coxing's white-bellied rats appeared to be an important host for rosavirus C. The relatively low seroprevalence of rosavirus C, as compared to RT-PCR detection rate, among tested Coxing's white-bellied rats (7.5%) may be due to delayed antibody response during acute infection when the animals were still shedding viruses. Together with the ability of rosavirus C in infecting house mouse (Mus musculus), our findings provided evidence for the interspecies transmission potential of rosavirus C among different rodent species. This is in line with the ability of bat picornaviruses group 1 to 3 in infecting bats of different genera or species [29]. Encephalomyocarditis virus (EMCV), of which swine is the major reservoir, can also infect different animals including rodents, elephants, boars, macaques and humans [55]. Further investigations are warranted to elucidate the ability of rosaviruses to cross species barrier and emerge in other animals or human.

No. of tissues/sera (%) from sacrificed mice tested positive for rosavirus C RASM14A by RT-PCR/western blot analysis
The present results suggest that rosaviruses can be pathogenic to their hosts. Although rosavirus A has been detected in rodents and human previously [26,32], no virus isolate was available for pathogenicity studies. A few picornaviruses, such as EMCV (cardiovirus A) and

Theilovirus (cardiovirus B) under Cardiovirus and Ljungan virus (LV) (parechovirus B) under
Parechovirus, were also known to cause systemic infections in infected rodents. In particular, Theiler's murine encephalomyelitis virus (TMEV), which primarily causes asymptomatic enteric infections in mice, has been intensively studied because of its ability to cause myocarditis, type 1 diabetes and acute or persistent demyelinating infections mimicking multiple sclerosis [35,56]. On the other hand, rodents experimentally infected with EMCV may develop type 1 diabetes mellitus, encephalomyelitis, myocarditis, orchitis and sialodacryoadenitis [57]. Interestingly, LV, which may cause type 1 diabetes and fetal deaths in infected rodents, has been recently found in human intrauterine fetal death and sudden infant death syndrome [58][59][60][61]. However, the pathogenicity of other rodent picornaviruses, such as rosavirus A1, mosavirus A1, murine kobuvirus 1, Norway rat hunnivirus and rabovirus A, was less clear [32,33]. The ability of rosavirus C in causing multisystemic infection in mice with high viral loads in infected organs suggested that rosaviruses may cause severe infections in their host. Further experimental studies using other rosaviruses and rodent species may help to better understand the pathogenicity of members of the genus Rosavirus in different rodents and human.
Rodents are the largest order of mammals on earth, accounting for 43% of the approximately 4,800 living mammalian species. They are widely distributed, being found in all habitats except the oceans. The order, Rodentia, with around 2050 species under 28 families, is further classified into five suborders: Anomaluromorpha, Castorimorpha, Hystricomorpha, Myomorpha and Sciuromorpha. [62,63]. Viruses of at least 22 families, including Adenoviridae, Arenaviridae, Arteriviridae, Astroviridae, Bunyaviridae, Caliciviridae, Circoviridae, Coronaviridae, Flaviviridae, Hepadnaviridae, Hepeviridae, Herpesviridae, Papillomaviridae, Paramyxoviridae, Parvoviridae, Picobirnaviridae, Polyomaviridae, Reoviridae, Rhabdoviridae, Togaviridae, Picornaviridae and Poxviridae, are known to infect rodents [64,65]. Rodent pathogens may infect human either by direct contact such as bites and inhalation of aerosolized animal excreta, or indirectly through vectors such as ticks and fleas. Urban rodents may pose particular risk to human health, as in the case of Hantavirus and lymphocytic choriomeningitis virus infections. More epidemiological studies should be performed to explore the diversity of rodent picornaviruses and their potential risks to human.

Collection of rodent samples
A total of 1232 wild and street rodents, belonging to eight different species, were captured from various locations in both rural and urban areas of Hong Kong, Hunan Province and Guangxi Province of China over a five-year period (September 2008 to August 2013) ( Table 1). Samples from Hong Kong were provided by the Agriculture, Fisheries and Conservation Department (AFCD) and Food, Environment and Hygiene Department (FEHD), the government of the Hong Kong Special Administrative region (HKSAR), as part of a surveillance program on local rodents. All rodents were individually trapped and samples were collected from each rodent using procedures described previously [11,66]. To prevent cross contamination, collection of samples were performed using disposable swabs with protective gloves changed for each rodent. Wild rodents in rural areas of Hong Kong were released back to nature after sample collection. Samples from street rodents in urban areas of Hong Kong and rodents from China were collected immediately after euthanasia as routine policies for disposal of captured rodents. All samples were placed in viral transport medium (Earle's balanced salt solution, 0.09% glucose, 0.03% sodium bicarbonate, 0.45% bovine serum albumin, 50 mg/ml amikacin, 50 mg/ml vancomycin, 40 U/ml nystatin) to inhibit bacterial and fungal overgrowth, and stored at -80°C before RNA extraction.

RNA extraction
Viral RNA was directly extracted from the respiratory and alimentary samples in viral transport medium using Viral RNA mini kit (QIAgen, Hilden, Germany). The RNA was eluted in 60 μl of RNase-free water and was used as the template for RT-PCR.
RT-PCR of 3D pol gene of picornaviruses using conserved primers and DNA sequencing Initial picornavirus screening was performed by amplifying a 159-bp fragment of the 3D pol gene of picornaviruses by RT-PCR using conserved primers (5'-GGCGGYTNGAYGGYGCS ATGCCGT-3' and 5'-CCGACCARCACRTCRTCRCCRTA-3') and previously described protocols [6,24,29]. The primers were designed by multiple alignment of the nucleotide sequences of the 3D pol genes of all known picornaviruses, based on the conserved 3D pol motifs, GG[LMN]PSG and YGDD. All samples positive by RT-PCR were confirmed by sequencing. Briefly, reverse transcription was performed using the SuperScript III kit (Invitrogen, San Diego, CA, USA) and the reaction mixture (10 μl) contained RNA, first-strand buffer (50 mM Tris-HCl pH 8.3, 75 mM KCl, 3 mM MgCl 2 ), 5 mM DTT, 50 ng random hexamers, 500 μM of each dNTPs and 100 U Superscript III reverse transcriptase. The mixtures were incubated at 25°C for 5 min, followed by 50°C for 60 min and 70°C for 15 min. The PCR mixture (25 μl) contained cDNA, PCR buffer (10 mM Tris-HCl pH 8.3, 50 mM KCl, 2 mM MgCl 2 and 0.01% gelatin), 200 μM of each dNTPs and 1.0 U Taq polymerase (Applied Biosystem, Foster City, CA, USA). The mixtures were amplified in 40 cycles of 94°C for 1 min, 50°C for 1 min and 72°C for 1 min and a final extension at 72°C for 10 min in an automated thermal cycler (Applied Biosystem, Foster City, CA, USA). Standard precautions were taken to avoid PCR contamination and no false-positive was observed in negative controls.
All PCR products were gel-purified using the QIAquick gel extraction kit (QIAgen, Hilden, Germany). Both strands of the PCR products were sequenced twice with an ABI Prism 3130xl DNA Analyzer (Applied Biosystems, Foster City, CA, USA), using the two PCR primers. The sequences of the PCR products were compared with known sequences of the 3D pol genes of picornaviruses in the GenBank database.
RT-PCR of 3D pol gene of novel rodent picornaviruses using specific primers and DNA sequencing As initial RT-PCR of the 3D pol gene revealed at least two potential novel picornavirus species in 18 respiratory and 24 alimentary tract samples, all the 2450 respiratory and alimentary tract samples were re-tested using specific RT-PCR assays to enhance the sensitivities for detection of these novel picornaviruses. Primers were designed by multiple alignment of the 3D pol gene sequences obtained during genome sequencing from the initial positive samples. The PCR assays were targeted to a 253 bp (5'-ATGCTCCTGTTCTCATGCTTTT -3' and 5'-GAAAA TCTGGGTCAGGGGTGAA -3') fragment and a 243-bp (5'-TGTTCTCTTGYTTYTCCCAG AT -3' and 5'-AAYTGCGGGTCYGGDGTGAA -3') fragment of the 3D pol gene of the potential novel picornaviruses. The components of the PCR mixtures and the cycling conditions were the same as those described above. Purification of the PCR products and DNA sequencing were performed as described above, using the corresponding PCR primers. The sequences of the PCR products were compared with known sequences of the 3D pol genes of picornaviruses in the GenBank database.

Genome sequencing and analysis
Thirteen genomes of the two novel picornavirus species were amplified and sequenced, with RNA directly extracted from respiratory or alimentary samples as templates [6,24,29]. RNA was converted to cDNA by a combined sequence-specific-priming,random-priming and oligo (dT) priming strategy. As initial results showed that the two novel picornaviruses are distantly related to known picornaviruses, the cDNAs of three initial strains were amplified by 5'-rapid amplification of cDNA ends (RACE) using the SMARTer RACE cDNA Amplification Kit (Clontech, USA). The first strand cDNA for the 5' sequence of the genome was constructed with specific primers designed according to results of the first and subsequent rounds of sequencing and SMARTer II A Oligonucleotide by SMARTScribe Reverse Transcriptase. The 3' sequence of the genome is completed by specific primers designed for the 3' end from the results of the first and subsequent rounds of sequencing and oligo (dT) primer. Sequences were assembled to produce final sequences of the viral genomes. The genomes of the remaining strains were amplified and sequenced by the specific primers designed from the initial three genomes and the 5' ends of the viral genomes were confirmed by RACE using the SMARTer RACE cDNA Amplification Kit (Clontech, USA).
The nucleotide (nt) sequences of the genomes and the deduced amino acid (aa) sequences of the open reading frames (ORFs) were compared to those of known picornaviruses. Phylogenetic tree construction was performed using maximum-likelihood methods from PhyML 3.0 program. Secondary structure prediction in the 5'UTR was performed using RNAfold [67] and the IRES elements were determined based on sequence alignment with EMCV as described previously [28,41].

Estimation of synonymous and non-synonymous substitution rates
To estimate the selective pressure in driving viral evolution among different regions of the genomes, the number of synonymous substitutions per synonymous site, Ks, and the number of non-synonymous substitutions per non-synonymous site, Ka, for each coding region between different strains of rosavirus B and C were calculated using the Nei-Gojobori method (Jukes-Cantor) in MEGA 5.0 as described previously [68].

Western blot analysis
Since the VP1 is the largest and most surface-exposed protein which contains most of the motifs important for interaction with neutralizing antibodies in picornaviruses, (His) 6 -tagged recombinant VP1 proteins of rosavirus B strain RNCW1002091R from a Norway rat and rosavirus C strains, RASK8F from an Indochinese forest rat and NCGX12IN from a Coxing's white-bellied rat, were cloned as described previously [28,30]. Briefly, the VP1 gene was amplified and cloned into the NheI site of expression vector pET-28b(+) (Novagen, Madison, WI, USA) in frame and downstream of the series of six histidine residues. The (His) 6 -tagged recombinant VP1 polypeptide was expressed and purified using the Ni 2+ -loaded HiTrap Chelating System (GE Healthcare, Buckinghamshire, UK) according to manufacturer's instructions. Western blot analysis was carried out using available rodent sera using the purified recombinant VP1 protein as described previously [30]. Briefly, the purified (His) 6 -tagged recombinant VP1 protein was loaded into each well of a sodium dodecyl sulfate (SDS)-10% polyacrylamide gel and subsequently electroblotted onto a nitrocellulose membrane (Bio-Rad, Hercules, CA, USA). The blot was cut into strips and the strips were incubated separately with serial dilutions of sera collected from different rodent species with for IgG detection. Antigen-antibody interaction was detected with horse radish peroxidase-conjugated secondary antibodies and ECL fluorescence system (GE Healthcare, Buckinghamshire, UK).

Viral cultures
Eleven samples tested positive for the novel picornaviruses were subject to virus isolation in various cell lines including Vero E6 (African green monkey kidney; ATCC CRL-1586), CrFK (Crandell feline kidney; ATCC CCL-94), in-house HFL (human embryonic lung fibroblast), 3T3 (mouse embryonic fibroblast, ATCC CCL-92) cells, RD (human embryo rhabdomyosarcoma; ATCC CCL-136), RK3E (rat kidney; ATCC CRL-1895) and TCMK1 (mouse kidney; ATCC CCL-139) cells as described previously [28,69]. Briefly, after centrifugation, samples were diluted five folds with viral transport medium and filtered. Filtrates were inoculated to Minimum Essential Media (MEM) and the mixtures were added to 24-well tissue culture plates by adsorption inoculation. After 1 h of adsorption, excess inoculum was discarded, and the wells were washed twice with phosphate buffered saline and replaced by serum-free MEM. Cultures were inspected daily by inverted microscopy for CPE. Subculturing to fresh cell line was performed from time to time even if there was no CPE and culture lysates were collected for RT-PCR for monitoring viral replication. Immunostaining and electron microscopy were performed on samples that were RT-PCR positive.
Electron microscopy 3T3 cells successfully infected by rosavirus C RASM14A were subject to negative contrast electron microscopy as described previously [69,70]. Tissue culture cell extracts infected with rosavirus C RASM14A were centrifuged at 19 000 g at 4°C, after which the pellet was resuspended in phosphate-buffered saline and stained with 2% phosphotungstic acid. Samples were examined with a Philips EM208s electron microscope.

Neutralization assays
Neutralization assays for rosavirus C RASM14A were carried out as described previously with modifications [11]. Briefly, rodent sera serially diluted from 1:10 to 1:80 were mixed with 100 TCID 50 of rosavirus C RASM14A. After incubation for 2 h at 37°C, the mixture was inoculated in duplicates onto 96-well plates of 3T3 cell cultures. Results were recorded after 3 days of incubation at 37°C.

Mice challenge experiments
Virus stock used to inoculate mice was obtained from at least the 18 th passage of rosavirus C RASM14A in 3T3 cells. Groups of 20 suckling balb/c mice were infected orally (4-day-old) and intracerebrally (1-day-old) as described previously [71]. Approximately 200μl (500 TCID 50 ) of virus suspensions was applied orally and 30μl (100 TCID 50 ) intracerebrally. Two mice challenged with culture media from uninfected cells were included as negative controls in both groups. Mice were monitored daily for signs of disease. Nine/ten, five, three and three mice were sacrificed at day 3, 7, 14 and 21 respectively. After euthanasia, necropsies of mice were performed to obtain the following tissues: intestine, spleen, kidney, liver, lung and brain. Blood was collected for antibody tests by western blot analysis as described above.

Histopathological and immunohistochemical studies
To perform immunhistochemical staining on infected cell lines and rodent tissues, guinea pig antiserum against the VP1 protein of rosavirus C was produced by subcutaneously injecting 100 μg purified recombinant rosavirus C VP1 protein to three guinea pigs, using an equal volume of complete Freund's adjuvant (Sigma) as described previously [28]. Incomplete Freund's adjuvant (Sigma) was used in subsequent immunizations. Three inoculations at once every two weeks per guinea pig were administered. Two weeks after the last immunization, 1 ml of blood was taken via the lateral saphenous vein of the guinea pigs to obtain the sera.
To examine the histopathology and viral replication of rosavirus C RASM14A in tissues of challenged mice, necropsy organs of the mice were subject to both viral RNA detection by RT-PCR and immunohistological studies as described previously [28]. Tissues for histological examination were fixed in 10% neutral-buffered formalin, embedded in paraffin, and stained with hematoxylin and eosin (H&E). Histopathological changes were observed using Nikon 80i microscope and imaging system. Infected cell lines and tissues from challenged mice that were tested positive for rosavirus C RASM14A by RT-PCR were subject to viral load studies and immunohistochemical staining for viral VP1 protein as described previously [28,69]. Tissue sections were deparaffinized and rehydrated, followed by blocking endogenous peroxidase with 0.3% H 2 O 2 for 25 min, and then with 1% BSA/PBS at room temperature for 25 min to minimize non-specific staining. The tissue sections were then pre-treated with streptavidin solution and biotin solution at room temperature for 30 min respectively to avoid high background signals due to the endogenous biotin or biotin-binding proteins in the tissues. The sections were incubated at 4°C overnight with 1:2000 dilution of guinea pig anti-VP1 anti-serum, followed by incubation of 30 min at room temperature with 1:2000 dilution of biotin-conjugated rabbit anti-guinea pig IgG, H & L chain (Abcam). Streptavidin/peroxidase complex reagent (Vector Laboratories) was then added and incubated at room temperature for 30 min. Sections were counterstained with hematoxylin. Cells infected or uninfected by rosavirus C RASM14A were included as positive and negative controls respectively in each staining. Cells were fixed in chilled acetone at -20°C for 10 min before incubation with antibodies for staining. Color development was performed using 3,3'-diaminobenzidine and images captured with Nikon 80i imaging system and Spot-advance computer software.

Ethics statement
The . The mice study was carried out in strict compliance with animal welfare regulations. The mice were anesthetized by fetanyl/fluanisone/diazepam during the whole experiment. Standard guidelines prescribed in Pain and distress in laboratory rodents and lagomorphs, Laboratory Animals 28, 97-112 (1994) were strictly followed and the well-being of animals were monitored daily with a scoring sheet to ensure minimal pain and distress experienced by the mice.