Development of a reverse-genetics system for murine norovirus 3: long-term persistence occurs in the caecum and colon

Human noroviruses (HuNoV) are a major cause of viral gastroenteritis worldwide, yet, due to the inability to propagate HuNoV in cell culture, murine norovirus (MNV) is typically used as a surrogate to study norovirus biology. MNV-3 represents an attractive strain to study norovirus infections in vivo because it establishes persistence in wild-type mice, yet causes symptoms resembling gastroenteritis in immune-compromised STAT1−/− mice. The lack of reverse-genetics approaches to recover genetically defined MNV-3 has limited further studies on the identification of viral sequences that contribute to persistence. Here we report the establishment of a combined DNA-based reverse-genetics and mouse-model system to study persistent MNV-3 infections in wild-type (C57BL/6) mice. Viral RNA and infectious virus were detected in faeces for at least 56 days after inoculation. Strikingly, the highest concentrations of viral RNA during persistence were detected in the caecum and colon, suggesting that viral persistence is maintained in these tissues. Possible adaptive changes arising during persistence in vivo appeared to accumulate in the minor capsid protein (VP2) and the viral polymerase (NS7), in contrast with adaptive mutations selected during cell-culture passages in RAW264.7 cells that appeared in the major capsid protein (VP1) and non-structural protein NS4. This system provides an attractive model that can be readily used to identify viral sequences that contribute to persistence in an immunocompetent host and to more acute infection in an immunocompromised host, providing new insights into the biology of norovirus infections.


INTRODUCTION
Gastroenteritis remains one of the top five causes of death worldwide and the second most common cause in lowincome countries (Monroe, 2011). The development of a rotavirus vaccine places human noroviruses (HuNoV) as the leading cause of food-borne disease and non-bacterial gastroenteritis worldwide (Kahan et al., 2011;Koo et al., 2010). It has been estimated that HuNoV infections in developing countries result in .200 000 deaths and .900 000 hospitalizations of children younger than 5 years every year (Patel et al., 2008). In high-income countries, however, low mortality rates are reported for HuNoV, with the major impact being observed on elderly and immunocompromised patients. It is estimated that 23 million symptomatic infections occur each year in the USA alone, resulting in at least 300 deaths (Harris et al., 2008;Mead et al., 1999;van Asten et al., 2011). Frequent large outbreaks caused by HuNoV are reported during the winter season, affecting the operational capacity of many closed environments (e.g. schools, hospitals, military camps). Large economic losses (estimated to be .£100 million in the UK alone) are therefore associated with hospital disruptions caused by HuNoV outbreaks (Lopman et al., 2004). In addition to acute gastroenteritis, noroviruses have recently been linked to a number of significant clinical diseases, such as the exacerbation of inflammatory bowel disease (IBD) (e.g. Crohn's disease, ulcerative colitis), seizures in infants and other neurological disorders (Byrnes & Griffin, 2000;CDC, 2002;Chen et al., 2009;Ito et al., 2006;Khan et al., 2009). Recent studies carried out with animal models have supported the connection between norovirus infection and the exacerbation of IBD (Cadwell et al., 2010;Lencioni et al., 2008).
The lack of efficient systems to cultivate HuNoV in cell culture has limited studies on norovirus replication and pathogenesis. Discovery of the closely related murine norovirus (MNV), capable of replication in several cell lines, has provided an ideal alternative with which to study the molecular biology of noroviruses and the identification of antiviral strategies to control them (Karst et al., 2003;Wobus et al., 2004). MNV-1 was the first strain to be isolated and was found to lead to a rapid systemic and lethal infection in immunocompromised STAT1 2/2 mice (Karst et al., 2003). MNV-1 replicates in both immunocompromised and wild-type mice, although in the presence of a competent immune system, virus replication is rapidly controlled and cleared, becoming undetectable in faeces and organs from 7 days post-infection (Hsu et al., 2006;Kahan et al., 2011;Karst et al., 2003;Mumphrey et al., 2007). Following the identification of MNV-1, three new MNV strains were isolated from asymptomatic mice from different geographical location colonies, referred to as MNV-2, -3 and -4. These three new strains were able to establish persistent infections in wild-type ICR (imprinting control region) mice that lasted for .10 weeks and resulted in large amounts of viral RNA being shed in the faeces (Hsu et al., 2006). Subsequent studies confirmed that MNV is highly prevalent in research colonies and that the different strains isolated are genetically closely related and form a single serotype (Barron et al., 2011;Hsu et al., 2007;Thackray et al., 2007).
The MNV genome is a positive-stranded RNA molecule around 7.5 kb in length, containing four ORFs (Fig. 1a). Viral RNA replication, catalysed by the viral RNAdependent RNA polymerase NS7, results in the generation of new viral genomes, as well as a subgenomic RNA encompassing the ORF2-ORF4 coding region. Replication is primed by VPg (NS5), which also plays a key role in recruiting host factors to initiate VPg-dependent viral protein synthesis (Chaudhry et al., 2006;Daughenbaugh et al., 2003Daughenbaugh et al., , 2006Goodfellow et al., 2005). Translation of ORF-1 leads to the synthesis of a large polyprotein that is cleaved into the different non-structural mature proteins (NS1-7) (Sosnovtsev et al., 2006). Translation of the subgenomic RNA leads to production of the capsid proteins VP1 and VP2 (ORF2 and 3, respectively), as well as the innate immune antagonist VF1 (ORF4) (McFadden et al., 2011;Wobus et al., 2006).
The development of several reverse-genetics approaches for MNV has allowed molecular characterization of the viral and host factors that regulate norovirus replication (Chaudhry et al., 2007;Ward et al., 2007;Yunus et al., 2010). We have shown previously that genetically defined wild-type MNV-1 and mutant viruses recovered in cell culture can establish an acute infection in mice, which has been instrumental for the molecular characterization of MNV-1 in cell culture and in vivo (Bailey et al., 2008(Bailey et al., , 2010McFadden et al., 2011). The recent identification of other MNV strains able to establish stable persistent infections in mice has opened new avenues to understanding viral sequences that may contribute to viral persistence. However, the absence of reverse-genetics approaches for these new MNV variants has limited the  MNV-3 contains four ORFs. Translation of ORF1 results in the synthesis of a viral polyprotein that is then processed proteolytically into different mature non-structural viral proteins as indicated. Translation of the subgenomic RNA results in the synthesis of major capsid protein VP1 (ORF2), minor capsid protein VP2 (ORF3) and virulence factor VF1 (ORF4). The sequence of pT7 : MNV-3 differs from the previously reported sequence in 53 nt positions along the genome, represented as short arrows with non-filled arrowheads (synonymous) and as long arrows with filled arrowheads (non-synonymous). establishment of an animal model to study persistent infections in mice with a genetically defined MNV isolate.
Here, we describe for the first time the combination of reverse-genetics approaches and animal models to study the establishment of a persistent norovirus infection in wild-type C57BL/6 mice. Viral RNA was detected in faeces for at least 56 days. In addition, all infected animals displayed high MNV-3 RNA loads in the caecum and colon, suggesting that these are the major sites for persistent replication.

Reverse-genetics recovery of MNV-3
To enable the generation of a full-length cDNA clone of MNV-3, a sample of MNV-3 provided by Robert Livingston (Hsu et al., 2007) was used to infect RAW264.7 cells and total RNA was extracted 24 h post-infection as explained in Methods. This collected sample represented the viral population after three rounds of infection at low m.o.i. in RAW264.7 cells. The full viral genome was then amplified by RT-PCR using primers that introduced a truncated T7 RNA polymerase promoter at the 59 end and a tail of 26 adenines at the 39 end. Cloning of the amplified fragment into the backbone plasmid used previously to generate an MNV-1 reverse-genetics system (Chaudhry et al., 2007) resulted in the insertion of a self-cleaving d-ribozyme sequence after the A 26 sequence to ensure the generation of a defined 39 end (Walker et al., 2003). A number of constructs were sequenced in their entirety and screened for their ability to produce infectious virus (data not shown). Surprisingly, the consensus sequence and the sequences of three individual constructs displayed .50 common sequence alterations with respect to the initially reported MNV-3 sequence (Hsu et al., 2007).
MNV-3 derived entirely from cDNA can be propagated successively in RAW264.7 cells MNV-3 recovered above, referred to as passage 0 (MNV-3 p0), was used to infect RAW264.7 cells at an approximate m.o.i. of 0.001 TCID 50 per cell. Ninety-six hours postinfection, cytopathic effect was not apparent. Infectious MNV-3 passage 1 (MNV-3 p1) was released by two successive freeze-thaw cycles and the filtered lysate was used to inoculate a new monolayer of RAW264.7 cells (m.o.i. 0.1). MNV-3 p2 was recovered at 48 h postinfection, associated with large degrees of cytopathic effect. Consecutive passages of MNV-3 at m.o.i.s ranging from 0.1 to 0.3 resulted in an increase in the virus titre and in the cytopathology in RAW264.7 cells during infection. Virus yields (titres) by passage 6 (MNV-3 p6) were typically 15fold higher than at passage 2 ( Fig. 1c). This was also followed by an increase in the appearance of cytopathic effect (not shown).
To determine whether increased virus yield was associated with changes in the viral genome, viral RNA was extracted from MNV-3 p2 and p6 stocks and their full genomes were amplified by RT-PCR. Sequence analysis revealed only two sequence changes in MNV-3 p6 relative to the cDNA clone and in both cases a degree of heterogeneity was observed; typical chromatogram signal intensities were in the range 60-80 %/40-20 % for the mutant/wild-type sequence. The mutations identified were A2269G and C5957U, resulting in the changes D50G and T301I in the viral NS4 and VP1 proteins, respectively. No mutations were found in the MNV-3 p2 genome, suggesting that changes in MNV-3 p6 may reflect tissue-culture adaptation. We and others have previously observed that, during repeated passage of MNV-1 in RAW264.7 cells, similar adaptation occurs, with changes also being observed in NS4 and VP1 (V11I and K296E, respectively; Bailey et al., 2008;Wobus et al., 2004).

MNV-3 recovered by reverse genetics establishes persistent infections in C57BL/6 mice
To investigate whether MNV-3 recovered in cell culture from cDNA was able to establish a persistent infection in immunocompetent mice, as shown previously for field isolates of MNV-3 in ICR mice (Hsu et al., 2006), we performed oral inoculations of 4-5-week-old C57BL/6 male mice. The maximum possible dose of virus was used in this initial trial, namely 100 ml of either MNV-3 p2 or MNV-3 p6 (1.0610 5 and 1.5610 6 TCID 50 , respectively). C57BL/6 mice were chosen due to the ready availability of a wide number of knockout and knock-in lines in this genetic background. Inoculated animals showed no clinical symptoms or differences in weight increase relative to mock-infected or uninfected animals for the whole period of surveillance (56 days) (Fig. 2).
To determine whether the cDNA-derived virus had established a persistent infection, stool samples were collected from the animals at days 28 and 56 post-infection, and analysed by both quantitative and semiquantitative RT-PCR. Two different methodologies were used for further validation of the levels of viral RNA and to overcome limitations observed with the detection limits of each assay. We detected genomic MNV RNA in the faeces of all the inoculated animals at both days 28 and 56 post-infection by semiquantitative RT-PCR, but not in faeces of mockinfected animals (Fig. 2). Semiquantitative and quantitative analyses were in agreement with each other, and showed that most animals excreted viral RNA at day 28 in the range of 10 5 molecules (mg stool) 21 [corresponding to approx. 10 4 molecules of viral RNA (ml sample analysed) 21 ] (Fig. 2). At day 56, a decrease in viral RNA levels was observed when compared with the same animals at day 28 [approx. 10 3 -10 4 molecules (mg stool) 21 ; P,0.05, two-way ANOVA], suggesting that virus replication had been partially restricted. Although not statistically significant, mean values for animals infected with MNV-3 p2 tended to be higher than in animals infected with MNV-3 p6. Sequence analysis of MNV-3 VP1 in faeces of three different animals infected with MNV-3 p6 (day 56) indicated that the dominant sequence present at amino acid position 301 in VP1 was T. This result might suggest that adaptive mutation T301I, selected during passage in RAW264.7 cells and present in the passage 6 stock but absent after two passages, resulted in decreased fitness in vivo and was outcompeted by wild-type variants, detected as a minority in MNV-3 p6.

MNV-3 infection occurs at a low inoculating dose and shows fast replication kinetics in mice
To investigate the infectivity of MNV-3 in an immunocompetent background in vivo, we inoculated C57BL/6 mice with 10-fold-increasing doses of MNV-3 p2, ranging from 10 to 10 4 TCID 50 (Fig. 3). At 28 days post-infection, the four groups of animals inoculated with virus displayed detectable viral RNA by RT-PCR, whereas mock-infected animals or animals inoculated with 10 5 TCID 50 UVinactivated MNV-3 were negative for MNV RNA (Fig. 3 and data not shown). Hence, the ID 50 (amount required to infect 50 % of the animals) of MNV-3 is likely to be ,10 TCID 50 . We carried out the quantification of viral RNA in animals previously infected with 10 4 and 10 TCID 50 by quantitative PCR, and compared them with the values obtained previously for animals inoculated with 10 5 TCID 50 (Figs 2 and 3). The three groups of animals displayed similar mean values, although animals infected with 10 5 TCID 50 presented higher concentrations on average (although not significant by statistical analysis, one-way ANOVA test).
To determine whether the viral RNA detected in the faeces correlated with the presence of infectious virus in the stool pellets, we carried out TCID 50 assays for the samples analysed by quantitative PCR (Fig. 3c). We detected infectivity in the stool pellets of 17 out of 18 animals. In the positive samples, virus titres ranged from 1.2610 3 to 1.9610 5 TCID 50 (g stool sample) 21 . The calculated virus titres per stool pellet (20-50 mg on average) ranged from 5.2610 1 to 6.1610 3 TCID 50 shed in every faecal sample after 28 days post-infection. Interestingly, all animals initially inoculated with 10 TCID 50 MNV-3 p2 (with the exception of animal 118, which was negative for infectivity) released .200 TCID 50 in a single pellet.
Given the low inoculating dose required to establish an MNV-3 infection in vivo, we decided to investigate the replication kinetics of MNV-3 after low-dose inoculation. Mice were inoculated with either 10 or 10 2 TCID 50 and the kinetics of viral RNA shedding in the faeces of each animal infected were determined by quantitative PCR. The data indicated that MNV-3 infection occurs rapidly, with all animals actively secreting viral RNA on day 3 (Fig. 4).

MNV-3 persists in the caecum and large intestine
To identify the major tissue(s) for MNV-3 replication and persistence in C57BL/6 mice in vivo, we extracted various organs previously identified to support MNV replication. RNA was then isolated from the spleen, liver, mesenteric lymph node (MLN), different sections of the small intestine (duodenum/jejunum, proximal and distal ileum), caecum and colon from infected animals at 28 days postinoculation. Strikingly, the highest levels of viral RNA were detected in the colon followed by the caecum, with mean values .10 5 and .10 4 MNV molecules (mg total RNA) 21 , respectively (Fig. 5). Semiquantitative RT-PCR was also used to confirm these observations (not shown). MNV  102 103 104 105 106 113 114 115 116 117 118 149 150 151 152 153 154 107 108 109 110 111 112  RNA was also detected in the MLN of some animals, but at considerably lower levels than in tissues from the large intestine.

Persistent replication of MNV-3 in C57BL/6 mice results in the selection of repeated mutations in NS7 and VP2
To determine whether persistent replication of MNV-3 in mice resulted in the selection of adaptive mutations in the viral genome, we sequenced genomic RNA extracted from the faeces of five different animals at day 28 post-infection and from one animal at day 56 post-infection (Fig. 6). We identified three mutations repeated in three or more different animals: U3574C, C4179U and A6690G, which resulted in amino acid changes V13A and L215F in the NS7 polymerase and T4A in the VP2 minor capsid protein.
In addition, only three other non-synonymous mutations were identified in the analysis: A381T and T441I in VP1, found in two clones, and T91A in NS3, found in one clone (Fig. 6).

DISCUSSION
Here, we describe the establishment of a persistent infection in vivo with MNV-3 recovered from cDNA by reverse genetics. To our knowledge, this is the first time that a persistent infection of C57BL/6 mice with MNV recovered from cDNA has been achieved. However, during the course of this study, a reverse-genetics system for the CR6 strain of MNV, shown previously to establish a persistent infection , has been described. Recombinant virus derived from this system has to date only been used to identify virulence determinants in the VP1 protein within a STAT1 2/2 model (Strong et al., 2012).
The combination of an efficient reverse-genetics system for MNV-3 and an animal model to study stable persistent infections constitutes a significant step forward in the development of systems for understanding norovirus molecular biology and pathogenesis. Persistent infections in humans by HuNoV have also been reported, although they mainly affect immunocompromised individuals (Capizzi et al., 2011;Ludwig et al., 2008). Nevertheless, recent data indicate that asymptomatic norovirus prevalence is approximately 12 % in the UK, although whether this represents subclinical acute infections or long-term asymptomatic secretion of virus is not known (Phillips et al., 2010). A recent report has shown that MNV-3 infection in STAT1 2/2 mice results in an acute infection with typical signs of gastroenteritis (i.e. delayed gastric emptying, changes to diarrhoeal symptoms) (Kahan et al., 2011). This, combined with the fact that MNV-3 can establish a persistent infection in C57BL/6 mice, the most commonly used strain for the generation of knockout mice, provides additional support for the use of MNV-3 for the study of norovirus biology.
The observation that high viral RNA levels were detected in caecum and colon at day 28 post-infection, but not in the small intestine or other organs (MLN, spleen, liver), represents one of the most significant findings of this study. These data suggest that MNV-3 establishes a persistent  infection in the large intestine, allowing efficient virus shedding in faeces over a prolonged time frame. The strong association between ulcerative colitis, a colon disease, and the presence of HuNoV during exacerbated disease (Khan et al., 2009) supports the idea that the colon may also play an important role for HuNoV persistence. Two additional studies have provided further evidence for the connection between various colon disorders and active secretion of HuNoV in faeces, specifically in children with persistent diarrhoea and in patients with chronic diarrhoea associated with leukaemia (Capizzi et al., 2011;Vernacchio et al., 2006). A recent study comparing MNV-1 and MNV-3 in vivo has shown that the viruses replicate in different organs, including the caecum and the colon, but they are rapidly cleared, being undetectable by day 7 post-infection (Kahan et al., 2011). Possible explanations for these apparently conflicting data may be the sensitivity of the assays used and differences in the experimental set-up, i.e. age and strain of mice. It is also worth noting that, whilst the MNV-3 used here was derived from cDNA, that used in the previous study was a virus passed in tissue culture several times.
Interestingly, MNV persistent infections have also been established in 7-to 15-week-old mice using the CR6 strain, with viral RNA also being detected in colon 14 days postinfection at levels lower than those observed in the ileum and MLN (Cadwell et al., 2010). The results presented in this study differ somewhat from these results, which may be a reflection of the differing virus strain under study (CR6 versus MNV-3) or the age of the animals used, i.e. 7-15 weeks versus 4-5 weeks in our study. Alternatively, MNV replication for longer periods (.14 days) may have resulted in clearance from the ileum and MLN, but not from the caecum and the colon.
Analysis of full MNV-3 genome sequences derived from serial passage of MNV-3 in RAW264.7 cells and persistent replication in mice resulted in marked differences in genome evolution. Serial passage of MNV-3 in cell culture selected for non-synonymous mutations in NS4 and VP1. We have previously reported the selection of mutations in NS4 and VP1 during MNV-1 cell-culture passage: K296E in VP1 and V11I in NS4. K296E in MNV-1 VP1 was shown to cause attenuation in STAT1 2/2 mice (Bailey et al., 2008;Strong et al., 2012). Interestingly, MNV-3 and MNV CR6, both causative of persistent infections in wild-type mice, encode E296 in VP1 instead of K. The substitution E296K in MNV CR6 resulted in non-recoverable virus, which may indicate that this mutation is not tolerated in the context of the CR6 VP1 (Strong et al., 2012). Position 296 lies in close proximity to position 301 in the structure of VP1, both at the tip of the protruding P2 domain (Taube et al., 2010), which we interestingly observed to be mutated (T301I) during repeated passage of MNV-3 in RAW264.7 cells. Our observations would suggest that, upon inoculation of animals with a mixed population of viruses containing both T and I at position 301, namely MNV-3 p6, only T at position 301 in MNV-3 VP1 persists in vivo. Clearly, further studies are warranted to examine the role of this position in tissue-culture adaptation and the possible associated fitness cost in vivo. In addition, adaptive mutations in the picornavirus homologue of NS4 (3A) have been linked to increased virulence in cell culture and/ or in vivo (Arias et al., 2010;Harris & Racaniello, 2005;Nú ñez et al., 2001), warranting further studies on the precise function of the norovirus NS4 protein.
Persistent replication of MNV-3 in vivo did not select for any of the substitutions found during cell-culture passage, Fig. 6. Sequence analysis of viral RNA from stools at days 28 (animals 113, 122, 149, 152, 154) and 56 (animal 16, which presented the highest viral RNA concentration at that time point). Empty and filled symbols represent synonymous and nonsynonymous substitutions, respectively. e/ X represent mutations not totally imposed in the consensus sequence (30-80 % relative to wild-type sequence based on chromatogram quantification); $ represents a totally imposed change.
but instead selected for changes in VP2 and NS7. The mutations T4A in VP2, and V13A and L215F in NS7 were found in six, four and three animals, respectively. Additional mutations identified include A381T and T441I in VP1, and T91A in NS3. The T4A substitution in VP2 was found in the faeces of all animals analysed, but not during the serial passage of MNV-3 in RAW264.7, suggesting that this change may be important in vivo but not relevant for cell-culture replication. For the related feline calicivirus, the VP2 protein is essential for virus replication and for the assembly of infectious particles (Sosnovtsev et al., 2005). Recent studies with HuNoV VP1 and VP2 have revealed co-evolution of both proteins in a time-dependent manner, highlighting the integral role of the VP2 protein in the norovirus life cycle (Chan et al., 2012). Strikingly, we have found that alanine at position 4 in VP2 is absolutely conserved in .30 natural MNV isolates analysed. Nevertheless, the presence of threonine at position 4 in VP2 had no effect on virus recovery by reverse genetics or on virus replication in cell culture, being stably maintained after multiple passages in RAW264.7 cells. This highlights that the VP2 sequence may contribute to virus fitness in vivo due to selective pressures not observed in cell culture.
Mutations V13A and L215F in the viral RNA polymerase NS7 (Fig. 6) appear to affect polymerase surface residues, distant from the catalytic site, in the palm and fingers domains, respectively (Lee et al., 2011). Given their accessibility, these changes may be affecting the interaction of NS7 with potential viral or cellular factors, although this hypothesis requires further investigation. Interestingly, substitutions found in VP1 at lower frequency (A381T and T441I) are predicted to lie within exposed residues of flexible loops situated in the apical region of P2 and P1 domains, respectively, which could be indicative of selection by neutralizing antibody response. In particular, A381 is situated in loop E9-F9, identified previously as a major immunodominant epitope with antibody-escape mutants being mapped in position 386 (Lochridge & Hardy, 2007;Taube et al., 2010).
In summary, we have described for the first time the establishment of a persistent infection of C57BL/6 mice with MNV-3 recovered by reverse genetics. MNV-3 derived from cDNA persistently replicated in C57BL/6 mice for at least 56 days and was associated with high viral loads in the caecum and the colon. Preliminary data indicate that viral RNA is shed by infected mice for .6 months, emphasizing the ability of recovered MNV-3 to establish long-term persistent infections in vivo (N. McFadden, A. Arias, I. Goodfellow & P. Simmonds, unpublished results To enable the generation of a full-length cDNA clone of MNV-3, a sample of MNV-3 provided by Robert Livingston (University of Missouri, Columbia, MO, USA) was used to infect RAW264.7 cells at an m.o.i. of 0.1 TCID 50 per cell (Hsu et al., 2007). RNA was extracted from the infected cells 24 h post-infection. This sample collected represented the viral sequence after three rounds of low-m.o.i. infection in RAW264.7 cells. RNA was copied into cDNA using SuperScript II (Invitrogen) and a reverse primer containing a 39 SpeI site, a poly(A) of 27 nt and the last 36 nt of the MNV-3 genome. Genomic MNV cDNA was fully amplified using KOD polymerase (Merck), the primer used for reverse transcription and a 59 primer containing a SpeI site, a truncated T7 RNA polymerase promoter and the first 35 nt of the MNV genome. Gel-purified PCR product was digested with SpeI and ligated into the pT7 : MNV-1 backbone vector digested with SpeI and NheI. The resulting pT7 : MNV-3 contains 53 sequence changes (GenBank accession no. JQ658375) with respect to the originally published MNV-3 sequence ( Fig. 1) (Hsu et al., 2007).
For the recovery of infectious MNV-3, BSR-T7 cells were infected with a helper fowlpox virus expressing recombinant T7 RNA polymerase (FPV-T7) and transfected with pT7 : MNV-3 as described previously for MNV-1 (Chaudhry et al., 2007). Generated MNV-3 infectious particles were released from BSR-T7 cells by two consecutive freeze-thaw cycles and supernatants were filtered. For the amplification and generation of MNV-3 stocks, virus recovered from BSR-T7 cells was inoculated on a monolayer of RAW264.7 cells (m.o.i. 0.001) to obtain MNV-3 p1 (passage 1). Serial passage of MNV-3 (passages 2-6) was carried out in RAW264.7 cells (m.o.i. 0.1-0.3). Infected cells were then incubated at 37 uC until cytopathic effect was apparent (typically 48 h) and then subjected to two consecutive cycles of freezing and thawing and subsequent filtering to obtain infectious particles. Virus titres were determined by TCID 50 assays in RAW264.7 cells for 5 days followed by visual inspection.
Establishment of persistent MNV infections in vivo. Groups of six C57BL/6 male mice of 4-5 weeks of age (Harlan or Charles River) were inoculated by oral gavage with 100 ml sample containing varying amounts of MNV-3 virus stocks. Mock-infected animals were administered either filtered cell-culture lysates or 10 5 TCID 50 of MNV-3 previously inactivated by UV exposure. For UV inactivation, 1.6 ml containing 2610 7 TCID 50 of MNV-3 was cross-linked under high-intensity UV light at 4 uC using a Spectrolinker XL-1500 (Spectronics Corporation).
TCID 50 titration of viral samples obtained from animal faeces.
Stool pellets collected from infected animals were placed on ice and dispersed into PBS to reach a final concentration of 50 mg ml 21 . Resuspended faeces were then centrifuged at maximum speed (15 000 g) for 5 min and 100 ml supernatant was subjected to a second centrifugation step to remove any traces of faecal debris. The double-purified samples were titrated by TCID 50 assay in RAW264.7 cells.
RNA extraction and RT-PCR-based detection of MNV-3 in faeces and tissues. Viral RNA was extracted from 100 ml supernatant of faeces dispersed into PBS (50 mg ml 21 ) and from an approximately 20 mg portion of each tissue following the indications provided with the GenElute Mammalian Total RNA Miniprep kit (Sigma-Aldrich). Different tissues/organs collected were stored in RNAlater solution (Ambion) at 280 uC until RNA extraction was performed. For semiquantitative detection of MNV-3 RNA, one-or two-step RT-PCRs were performed using primers spanning genomic residues 3395-3430 (forward) and 3770-3734 (reverse) (reference GenBank accession no. JQ658375), and Moloney murine leukemia virus (M-MLV) reverse transcriptase and GoTaq DNA polymerase (Promega). The resulting 376 bp product was then resolved on a 2 % agarose gel. To obtain an accurate determination of the number of MNV-3 RNA molecules in faecal or tissue samples, we carried out reverse transcription-quantitative PCR following protocols described previously for the broad detection of MNV strains (Kitajima et al., 2010). Reverse transcription of MNV cDNA was performed using M-MLV reverse transcriptase (Promega) and a primer complementary to genomic positions 5345-5380. MNV cDNA was then quantified by quantitative PCR with primers spanning residues 5028-5047 (sense) and 5177-5138 (antisense), and a TaqMan FAM-TAMRA-labelled probe complementary to residues 5077-5062. Quantitative PCR determinations were carried out with Precision 26 qPCR MasterMix (Primerdesign) in a ViiA7 Real-Time PCR system apparatus (Applied Biosystems). In all the experiments, a standard curve for MNV RNA with a known number of molecules was carried out in parallel.