Isolation and Analysis of Rare Norovirus Recombinants from Coinfected Mice Using Drop-Based Microfluidics

Abstract

Human noroviruses (HuNoVs) are positive-sense RNA viruses that can cause severe, highly infectious gastroenteritis. HuNoV outbreaks are frequently associated with recombination between circulating strains. Strain genotyping and phylogenetic analyses show that noroviruses often recombine in a highly conserved region near the junction of the viral polyprotein (open reading frame 1 [ORF1]) and capsid (ORF2) genes and occasionally within the RNA-dependent RNA polymerase (RdRP) gene. Although genotyping methods are useful for tracking changes in circulating viral populations, they report only the dominant recombinant strains and do not elucidate the frequency or range of recombination events. Furthermore, the relatively low frequency of recombination in RNA viruses has limited studies to cell culture or in vitro systems, which do not reflect the complexities and selective pressures present in an infected organism. Using two murine norovirus (MNV) strains to model coinfection, we developed a microfluidic platform to amplify, detect, and recover individual recombinants following in vitro and in vivo coinfection. One-step reverse transcriptase PCR (RT-PCR) was performed in picoliter drops with primers that identified the wild-type and recombinant progenies and scanned for recombination breakpoints at ∼1-kb intervals. We detected recombination between MNV strains at multiple loci spanning the viral protease, RdRP, and capsid ORFs and isolated individual recombinant RNA genomes that were present at a frequency of 1/300,000 or higher. This study is the first to examine norovirus recombination following coinfection of an animal and suggests that the exchange of RNA among viral genomes in an infected host occurs in multiple locations and is an important driver of genetic diversity.

Importance: RNA viruses increase diversity and escape host immune barriers by genomic recombination. Studies using a number of viral systems indicate that recombination occurs via template switching by the virus-encoded RNA-dependent RNA polymerase (RdRP). However, factors that govern the frequency and positions of recombination in an infected organism remain largely unknown. This work leverages advances in the applied physics of drop-based microfluidics to isolate and sequence rare recombinants arising from the coinfection of mice with two distinct strains of murine norovirus. This study is the first to detect and analyze norovirus recombination in an animal model.

FIG 1

Differential PCR detects recombination between norovirus strains. (A) Similarity plot of MNV-1 and WU20 (gray trace) with a diagram of the expected PCR products (amplicons A to I [black bars]) (see also Table 1). The x axis indicates the position on the viral genome (in kilobases). The y axis indicates the similarity score at each nucleotide position. (B) Primer pair design. For each PCR product, four primer sets were used: p1 (MNV-1-forward and MNV-1-reverse), p2 (WU20-forward and WU20-reverse), p3 (MNV-1-forward and WU20-reverse), and p4 (WU20-forward and MNV-1-reverse). Primer sets 1 and 2 anneal to parental strains only (MNV-1 or WU20), while primer sets 3 and 4 detect recombinants (Rec MNV).

FIG 2

Validation of differential PCR. (A) Efficiency and specificity of p1 to p4. p1 and p2 specifically amplify MNV-1 and WU20, respectively, while p3 and p4 did not generate PCR products from parental strain cDNA. Mock indicates no viral template input. (B) The full panel of RT-PCR products (amplicons A to I) representing the parental strains MNV-1 (p1) and WU20 (p2) as well as recombinants (p3 and p4) in viral RNA purified from RAW264.7 cells coinfected with MNV-1 and WU20. (C) Differential PCR products analyzed by drop-based digital PCR. Mixed MNV-1 and WU20 RNA (lane 1) or RNA from RAW264.7 cells coinfected with MNV-1 and WU20 (lane 2) was encapsulated in drops with p4 and subjected to RT-PCR. The 1,200-bp recombinant product was amplified only from RNA purified from coinfected cells. M, molecular size markers (top, 1,500 bp; bottom, 1,000 bp).

FIG 3

Detection of recombination events in cells coinfected with MNV-1 and WU20. (A) Differential PCR of viral cDNA from in vitro coinfections. By using the primer pair combinations outlined in Fig. 1B, differential PCR for amplicons D to F was performed on samples of cells singly infected with MNV-1 or WU20, the mock lysate (top three panels), mixed RNA and cDNA from singly infected cells (middle two panels), and total or RNase-treated samples following coinfection (bottom two panels). (B, top) Recombination events detected in the region encompassing nt 4700 to 5904, a region that spans the ORF1/2 junction. The PCR products generated from total RNA after coinfection and from p3 and p4 (PCR products F3 and F4, respectively) (Fig. 2A, bottom right) were cloned into TOPO vectors. (Bottom) Five individual plasmid clones from each set were sequenced and aligned with the MNV-1 and WU20 reference sequences. Gaps between bars indicate 100% identity between WU20 and MNV-1 sequences.

FIG 4

Drop-based microfluidic platform for detecting recombinant MNV. (A) Schematic representation of RT-PCR in drops. (Top left) A one-step RT-PCR mixture with purified viral RNA was encapsulated in monodispersed picoliter drops by using a microfluidic drop maker. Each drop contained no more than one template (blue lines) and specific primers (red lines) for the detection of recombination. (Top right) Drops were collected in a PCR tube and subjected to thermocycling conditions for RT-PCR. (Bottom right) Drops with successfully amplified recombinant sequences generated a fluorescence signal from the DNA-intercalating dye EvaGreen. After thermocycling, drops were examined by fluorescence microscopy. (Bottom left) The thermocycled drops were reinjected into a microfluidic sorter, and drops containing amplicons (green) were detected via laser-induced fluorescence (488 nm) and sorted by dielectrophoresis. (B) Total RNA from in vitro coinfections was encapsulated in drops with primers specific for nt 4700 to 5904 (F amplicon; p4 [WU20-forward and MNV-1-reverse]). The contents of 9 drops containing 6 unique recombinants were sequenced and aligned with the MNV-1 and WU20 reference sequences. Gaps between bars indicate 100% identity betweenWU20 and MNV-1.

FIG 5

Detection of MNV recombinants from in vivo coinfections. Individually housed STAT1^−/− mice were infected with MNV-1 and WU20 orally (PO) or intraperitoneally (IP). Feces (FE) were removed every 24 h. The indicated tissues were harvested at 72 hpi. SP, spleen; LI, liver; ME, mesentery. (A) Viral loads in tissues and feces (48 to 72 hpi) were determined by a plaque assay. Each symbol represents one mouse. The mean viral titers for the combined group are indicated by black lines. (B) Genome titers in tissues and feces (48 to 72 hpi) were measured by drop-based RT-PCR. Each symbol represents one mouse. The means are indicated by black lines. *, P < 0.05; **, P < 0.01; ***, P < 0.001 (by Student t test). (C, top) Partial MNV genome showing coding regions within amplicons D and F (nt 2297 to 3803 and 4700 to 5904, respectively). (Bottom) Total recombinants in different tissues were counted by drop-based RT-PCR using specific primers. Recombinants were sorted and sequenced to determine the locus of recombination. Identical sites of recombination in different mice are highlighted with gray boxes.

FIG 6

Analysis of recombination site RNA sequences, RNA structure, and protein domains. (A) In vivo recombination sites in amplicon D (nt 2297 to 3803) that cooccur in the samples more frequently than by chance (see Result for details) were biased toward sequence similarity (two-sample Kolmogorov-Smirnov P value of 1.1 × 10⁻³) and toward protein domain similarity (two-sample Kolmogorov-Smirnov P value of 0.04). Recombination sites that occurred sporadically in amplicon D were not biased toward sequence or protein domain similarity. Here sequence mismatch is the number of nucleotides mismatching in a 10-bp window and domain mismatch is the distance to the nearest domain boundary in mismatching amino acids (see Materials and Methods). A P value of <0.05 rejects the null hypothesis that the two samples were drawn from the same continuous distribution. (B) In vivo recombination in amplicon F (nt 4700 to 5904) was significantly biased toward both sequence similarity (two-sample Kolmogorov-Smirnov P value of 3 × 10⁻⁵) and known protein domain similarity (two-sample Kolmogorov-Smirnov P value of 10⁻²⁰). (C) Localization of detected recombinants in amplicon D in the secondary structure. Detected recombination regions are plotted over the secondary RNA structure of amplicon D and the genome. Recombination sites tended to localize in 3 stem/loop regions of the molecule. This observation was robust across the first 25 configurations proposed by Mfold for the complete WU20 genome (the inset shows the first configuration), suggesting that some features of the RNA secondary or tertiary structure correlate with a higher probability of recombination. (D) Proposed model for recombination. During transcription, the RdRP dwells longer at locations where single-stranded RNA (loop) hybridizes to another strand to become a stem, presumably increasing the potential for template switching at those sites. (E) To test for recombination bias toward loop-to-stem junctions, the distance to the nearest loop-to-stem junction was measured for each recombination site. Detected recombination sites were on average 1.3 ± 0.03 nt away from a loop-to-stem junction, significantly closer than potential recombination sites that were on average 1.8 ± 0.02 nt away from a junction. The data include recombination sites that are 6 nt or shorter; junctions were calculated for 25 different secondary RNA structures.