Stool samples were thawed on ice from storage at –80°C and a 20% (vol/vol) stool suspension of total volume 1 mL made in water (pH 7.0). The sample was centrifuged for 1 min at 13,000 × g; the supernatant was then removed and centrifuged for a further 7 min at 13,000 × g. Viral RNA was extracted by using the QIAmp Viral RNA kit (Qiagen, Hilden, Germany) according to manufacturer's instructions. Amplification of the capsid region and a portion of the polymerase region was carried out as described previously (17). Amplification of a 507-bp region of the putative recombinant Sydney 2212/98/AU (corresponding to nucleotides 4610–5117 in Lordsdale virus, GenBank accession no. X86557) encompassing the 3´ end of the polymerase region and the 5´ end of ORF2 was achieved by using a nested reverse transcription–polymerase chain reaction approach. In brief, outer primers CB1 (17) and NoV2oR (5´-GTR AAC GCR TTY CCM GC-3´) (R = A or G, Y = C or T, M = A or C) and inner primer pairs 2212F (5´-GTG AGC ACA GAT ATM AAM TTA-3´) and 2212R (5´-AGA TGG AGY GGC GTC ATT CG-3´) were used in reaction conditions, as described previously (17). The ORF1/ORF2 overlap and flanking polymerase and capsid regions of another 2 suspected recombinants, NoV/Sydney C14/02/AU and NoV/Picton/03/AU, were amplified with hep170 (5´-TCH TTY TAT GGT GAT GA-3´) and GV29 (5´-CAA GAM ACW GTR AAM ACA TCA TCM CCA G-3´) (W = A or T) to produce a 1,070-bp product. Products were sequenced directly on an ABI 3730 DNA Analyzer (Applied Biosystems, Foster City, CA, USA).

Recombinant NoV were identified by constructing 2 phylogenetic trees, 1 using 420 bp of the 3´ end of the RNA polymerase region and the other using 550 bp of the 5´ end of ORF2. Strains that did not cluster with the same group of viruses in both trees were considered putative recombinant strains. Evolutionary distances between sequences were determined by using the GCG program, DNAdist (Kimura 2-parameter method) (18). The computed distances were used to construct phylogenetic trees with Fitch (18). To gain an internal estimate of how well the data supported the phylogenetic trees, bootstrap resampling (100 datasets) of the multisequence alignments was carried out with the program Seqboot (18). The consensus tree was calculated with Consense (18). Tree branch lengths were determined by analyzing the consensus tree with Puzzle, and trees were plotted by using the program TREEVIEW (version 1.6.6) (19).

The recombination breakpoint of putative recombinant strains was determined by using 2 methods: the maximum chi-squared method (20) and Simplot (version 2.5) (21). The maximum chi-squared method is recognized as being among the most accurate when compared independently with 13 other methods (22).

We identified 3 recombinant NoV GII isolates responsible for outbreaks of acute gastroenteritis in New South Wales, Australia. Phylogenetic analysis of polymerase and capsid sequences of these and other recombinant NoV GII isolates showed 9 recombinant NoV GII sequences. All other recombinant NoV GII were close relatives of these (Table). The 3 NoV GII recombinant sequences identified in this study are constructed from only 2 polymerase sequences and 2 capsid sequences. They share either capsid or polymerase sequences, which suggests that the regions were derived from a pool of circulating viruses. The close geographic relationship of recombinants that share sequences in only 1 part of their genome may indicate the source location of the recombination event. In addition to the above example, this phenomenon is seen with 2 isolates from Vietnam, Vietnam 026 and Vietnam 0703, that share polymerase sequence with another Vietnamese isolate 9912-02F (Table) (13). However, the global distribution of recombinants such as Sydney C14, found in Australia, the United States, Germany, and Japan, is evidence against this hypothesis and indicates a much higher prevalence of recombinant strains in relation to other strains than was previously considered.

The putative crossover point was identified on either side of the overlap in 9 recombinant NoV GII (Table). Recombination within the overlap cannot be identified because it is 100% conserved across all NoV GII sequences and masks the breakpoint. This fact and the identification of the breakpoint at position 5359 in the recombinant NoV GI strongly suggest that recombination takes place within the reading frame overlap in NoV. The reading frame overlap and 6–7 bp of downstream sequence are closely related to sequence found at the start of the genome. In NoV GII are 28 bp that are highly conserved at both the 5´ end of ORF1 and around the ORF1/ORF2 overlap, with a consensus sequence of 5´-GTG AAT GAA GAT GGC GTC KAR YGA CGC Y-3´ (bases involved in the formation of stem loop structures are underlined). In NoV GI, 27 bp are highly conserved at the 5´ end of the genome and the ORF1/ORF2 overlap region with a consensus sequence of 5´-GYR AAT GAT GAT GGC GTC KAA RGA CGY-3´. The 2 highly conserved regions for NoV GI and GII contain 2 in-frame and 3 in-frame start codons, respectively. The duplication of a conserved sequence at the start of ORF1 and ORF2 is characteristic of caliciviruses and is seen in all 4 genera, NoV, sapovirus, lagovirus, and vesivirus (5,26–28). This repetition at the 5´ end of the 2 major ORFs led us to consider the role of ORF1/ORF2 as a negative-strand subgenomic RNA promoter site. Indeed, a subgenomic RNA promoter is required for subgenomic RNA synthesis and is often found in close proximity to the 5´ end of subgenomic RNA species (29). The presence of a subgenomic RNA has not been proven in NoV, but it is highly likely based on transcription in related viruses (26,27,30). For example, subgenomic RNA species have been identified, with 5´ ORF2 sequences, in 2 caliciviruses, namely, feline calicivirus (26,27) and rabbit hemorrhagic disease virus (RHDV) (30). The recent and first report of a calicivirus subgenomic RNA promoter in RHDV at the 5´ end of ORF2 provides evidence to support this hypothesis (30). Additionally, RNA promoter regions often have stem loop structures (reviewed in [31]); such structures have been identified within the repeated sequences found at the start of ORF1 and ORF2 of NoV (see sequences above) (28). Taken together, strong evidence exists that the conserved 27/28-bp sequence found at the 5´ end of the NoV genome and ORF2 is part of an RNA promoter sequence.

The primary mechanism involved in recombination in RNA viruses is the copy-choice model (32). In this model homologous recombination is driven by the viral encoded RdRp when pausing occurs during the transcription of a region of secondary structure. The polymerase then loses processivity and switches between RNA templates (reviewed in [7,8]). A number of models of subgenomic synthesis have been proposed, but the most widely recognized is the internal initiation mechanism (33). Here the replicase initiates positive-strand subgenomic transcription internally on a negative-strand copy of genomic RNA (29). Using these 2 well-supported models, we propose a simple mechanism for recombination in NoV (Figure 3). Replication and internal subgenomic RNA synthesis generate 2 positive RNA species. These templates direct RNA synthesis that leads to the generation of both a full-length negative genome and a negative subgenomic RNA species, in the second round of replication. The negative subgenomic RNA is the key player in our proposed model, and such species have been identified in viruses that produce subgenomic RNA (34). We propose that recombination occurs when the enzyme initiates positive-strand synthesis at the 3´ end of the full-length negative strand, loses processivity at the stem loop of the ORF1/ORF2 overlap, then hops across (template switching) to an available negative subgenomic RNA species generated by a co-infecting virus (Figure 3). Alternatively, the RdRp could also template switch directly from 1 genomic RNA to another genomic RNA in the highly conserved ORF1/ORF2 overlap. The net result of both possibilities is a recombinant virus that has acquired new ORF2 and ORF3 sequences.

The decline in the prevalence of previously dominant strains, such as US-95/96 in the United States and Australia (3,17), suggests immunity in the community might be an important factor in reducing further spread of NoV. Recombination offers NoV an attractive mechanism for immune evasion. Subgenomic RNA promoters have been proposed to be recombination hotspots (35,36). In this study we have presented data to support this hypothesis, and we have described a simple mechanism of how recombination might occur in NoV.