Emergence of Human G2P[4] Rotaviruses in the Post-vaccination Era in South Korea: Footprints of Multiple Interspecies Re-assortment Events

After the introduction of two global rotavirus vaccines, RotaTeq in 2007 and Rotarix in 2008 in South Korea, G1[P8] rotavirus was the major rotavirus genotype in the country until 2012. However, in this study, an emergence of G2P[4] as the dominant genotype during the 2013 to 2015 season has been reported. Genetic analysis revealed that these viruses had typical DS-1-like genotype constellation and showed evidence of re-assortment in one or more genome segments, including the incorporation of NSP4 genes from strains B-47/2008 from a cow and R4/Haryana/2007 from a buffalo in India, and the VP1 and VP3 genes from strain GO34/1999 from a goat in Bangladesh. Compared to the G2 RotaTeq vaccine strain, 17–24 amino acid changes, specifically A87T, D96N, S213D, and S242N substitutions in G2 epitopes, were observed. These results suggest that multiple interspecies re-assortment events might have contributed to the emergence of G2P[4] rotaviruses in the post-vaccination era in South Korea.

The VP4 genes showed 96.2-97.5% sequence identities with the cognate genes of Malawian human strain TB-Chen (G2P [4]). In phylogenetic analysis, the VP4 gene clustered into two sub-lineages of IV (IV-a and -b) (Fig. 2b). Sub-lineage IV-b included seven of the twelve G2P [4] strains analysed, together with G2P[4] strains previously detected in the USA and Italy. Five strains in sub-lineage IV-a were closely related to strains SSKT-133, CU497-BK, and VU10-11-2, which were detected in Thailand and the USA during 2009-2013. Among the Korean P [4] strains, VP4 nucleotide identities were 96.5-99.8%.
Furthermore, several intra-genogroup re-assortment events were identified in G2P [4] strains co-circulating during the 2010-2015 season. Strains CAU14-09 and CAU14-10 clustered in the same sub-lineage, based on the analysis of eight genes, VP2, VP4, VP6, VP7, NSP1-NSP3, and NSP5. The remaining three genes, VP1, VP3, and NSP4, of these strains clustered in different sub-lineage groups. Similarly, with the exception of NSP1, which was closely related to VU12-13-26 from the USA, the other 10 genome segments of CAU15-11 grouped closely with those of the Thai SSKT-133 strain (Fig. 5a). Overall, the great genomic diversity of DS-1-like G2P [4] strains seems to have been generated through re-assortment events between human and animal strains. Comparison of VP7 amino acid sequences with those of vaccine strains. To identify substitutions in neutralisation epitopes and variable regions, the deduced amino acid sequences of the 12 G2P [4] strains, analysed in this study, were aligned with those of representative strains, including the G2 vaccine strain SC2-9 (lineage II) and prototype G2 strain DS-1 (lineage I; residue numbering based on the DS-1 sequence). The amino acid sequences of VP7 of the Korean G2P [4] strains exhibited 17-24 and 13-21 mismatches with the corresponding sequences in the G2 component of RotaTeq and the prototype G2P [4] strain DS-1, respectively. Among the amino acids in the VP7 epitopes, all Korean G2P [4] strains, (except CAU15-11) showed 3-4 substitutions (two in 7-1a and one or two in 7-1b) relative to that of the RotaTeq G2P [5] (Fig. 8). These differences were observed at positions 87 (AT; alanine to threonine) and 96 (DN; aspartic acid to asparagine) in the 7-1a epitope and at positions 213 (SD; alanine to threonine) and 242 (SN; serine to asparagine) in 7-1b epitope of strains CAU11-03 and CAU12-06, whereas the amino acid residues in the 7-2 epitope were all conserved. In contrast, three amino acid residues (87, 96, and 242) in the 7-1a and 7-1b epitopes were conserved in CAU15-11 and SC2-9. The analysis also showed that the Korean G2 strains contained three amino acid substitutions (V40L, V42A, and I44M) in one or two linear cytotoxic T lymphocyte epitopes at amino acid positions 40-52) when compared to the SC2-9 sequence.

Discussion
The prevalence of rotavirus genotypes varies depending on the socio-economic status of the population under study, climates of different countries, and histo-blood group antigen types in different individuals and populations around the globe 20,21 . According to long-term surveillance of rotavirus genotypes in France, the United Kingdom, other western European countries 22 , and Thailand 23 , G1P [8] was the most prevalent rotavirus genotype during 2007-2014, whereas G3P [8] was dominant in China and Japan 11 . However, in Korea, a different trend of rotavirus genotypes has been observed. According to the current study, G2P [4] was the dominant genotype (35.9%) for three years in one region, followed by G1P [8], G9P [8], and G3P [8]. Although the reason for the prolonged dominance of G2P [4] in Korea remains uncertain, G2P [4] human rotaviruses should be considered a primary target for rotavirus vaccines in countries like South Korea.  Over the last 12 years, surveillance in Seoul, South Korea has shown that rotavirus frequencies vary across seasons. Before the introduction of a rotavirus vaccine, the frequency of genotypes varied over time, and G2P [4] RVAs were infrequently detected. G1P [8], G3P [8], and G4P [6] were the strains most frequently circulating in Seoul between 2004 and 2007 24,25 . In the initial years of rotavirus vaccination era (2007-2013), G9P [8] was predominant during 2007-2009, G1P [8] resurged as the predominant strain during 2008-2010 and 2011-2013, and G3P [8] was predominant during 2010-2011 ( Fig. 1) [26][27][28][29] . Previous studies on different provinces in Korea revealed that G2P [4] was the fourth most common type, following G1P [8], G3P [8], and G9P [8], between 2006 and 2012, accounting for approximately 2.71% of circulating strains 11 . In contrast, the current study clearly indicates the emergence of G2P [4] strains, representing more than 35.9% of rotavirus infections in Seoul from March 2013 to August 2015. G2P [4] genotype was predominant in many other countries, including Belgium, Austria, Brazil, and Australia, where the Rotarix vaccine was introduced 11-15,30,31 . However, the samples in this study were collected from Seoul, where vaccination in children is optional (by parental choice) and administered according to opportunity and availability, with RotaTeq being more popular than Rotarix 32 . G2P [4] strains were found to be predominant even after the introduction of RotaTeq in Australia, Nicaragua, and other South and Central American countries, even though there was no established rotavirus vaccination program at this time 9,11,12,33,34 . These data suggest that the genetic changes observed in this study might have resulted from natural variation or re-assortment events between human and animal strains.
The twelve Korean G2P [4] RVA strains, analysed in the current study, had a complete DS-1-like background; however, other genetic variants were also observed to circulate in Korea. Several of the dendrograms, such as those for NSP1-5, VP1-4, and VP6, suggested re-assortment among the co-circulating strains. All genes, except for VP1, VP3, and NSP4, in the Korean G2P [4] strains, were closely related to those of other human RVA strains, but showed difference from the prototype DS-1 strain, with nucleotide identities between 86.2 and 96%. VP1, VP3, and NSP4 were the most divergent RVA genes, and appeared to share a distant common ancestor with strains obtained from a goat, buffalo, and cow, respectively. For example, strains CAU15-12, CAU13-08, and CAU12-05 (VP3), and CAU12-05, CAU13-07, CAU13-08, CAU14-10, and CAU15-12 (VP1) appear to share a distant common ancestor with strain RVA/Goat-tc/BGD/GO34/1999/G6P[1] from a goat (97% identity). Strains . Similar constellations, with one (or sometimes two) genes from other species, have recently been described for G2P [4] RVA strains detected in the USA, Italy, and Brazil [35][36][37] . Re-assortment between animal and human strains is predicted to rapidly alter RV diversity by introducing novel gene constellations 10,38,39 . However, available data regarding typical goat, buffalo, and cow RVA strains are limited, making it difficult to determine the species of origin for genome segments from non-human animal strains incorporated into human RVA strains.
The efficacy of vaccines against diarrhoea caused by various rotavirus genotypes has been well studied and ranges between 61% and 88% for Rotarix and 88% and 95% for RotaTeq, depending on the virus genotype 40 . In recent vaccine trials, both Rotarix and RotaTeq were shown to be equally effective against G1, G3, G4, and G9 strains with P[8] specificity 41 . The efficacy of the Rotarix vaccine against G2 strains is somewhat lower than that of RotaTeq, since the latter might contain a G2 antigen that is highly similar to the circulating G2 strains 42 . However, the Korean strains, analysed in this study, did not show a close phylogenetic relationship with the G2 VP7 gene in RotaTeq strain SC2-9. Furthermore, due to the number of amino acid changes and the absence of a P [4] component, the RotaTeq vaccine might also be less effective in inducing the production of antibodies capable of protecting against the circulating G2P[4] strains 42 .
The present study compared the amino acid motifs in the neutralising epitopes of VP7 proteins across the circulating Korean G2P [4] rotaviruses, their prototype DS-1 strains, and available vaccine strains. The VP7 protein contains three antigenic epitopes, 7-1a, 7-1b, and 7-2, which comprise of 29 amino acids from positions 87 to 291 43 . Several previous studies have indicated that substitutions at these positions, with or without changes in glycosylation, can change the antigenicity of human rotaviruses and allow them to escape host immunity [44][45][46] . Four notable changes were observed within the antigenic epitopes 7-1a and 7-1b: A87T, D96N, S213D, and S242N. These positions are commonly known as immunodominant sites of the VP7 protein, and are critical for reactions with neutralising monoclonal antibodies. Lazdins et al. 47 reported that rotaviruses with a mutation in the antigenic epitope 7-1b showed a 10-fold increase in resistance to neutralisation by antiviral antiserum. Amino acid substitutions at these positions were found to be associated with an inability to serotype a strain and are responsible for antigenic changes 48 . Recently, amino acid substitutions in the three main antigenic regions, especially at positions 96 and 213 were postulated to be associated with the emergence of rotavirus G2 in Korea during 2010-2015. This study represents the first large-scale whole-genome study to understand the evolution of rotaviruses after the introduction of rotavirus vaccines into the Korean immunisation program. Continuous molecular epidemiological monitoring of RVAs will be necessary to prevent and control them and to determine the need for the inclusion of an RVA vaccine in the national immunisation program.

Materials and Methods
Ethics statement. The  Seoul, Korea. Acute gastroenteritis was defined in this study as increased stool frequency (i.e., at least three loose and watery stools within a 24-h period), with or without vomiting and fatigue, occurring within the preceding 48 h. Approximately 10% suspensions of stool samples were prepared by vortexing stool sample (0.1 g) with phosphate-buffered saline (1 mL) (PBS; pH 7.4). The stool sample suspensions were centrifuged at 12,000x g for 15 min, and the supernatants were used as the faecal suspensions. RNA extraction. Viral RNA was extracted from the faecal suspensions using a QIAamp Viral RNA Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer's instructions. Extracted viral RNA was stored at −70 °C until it was used in reverse transcription-polymerase chain reactions (RT-PCRs).
Whole genome sequencing. G2P[4] strains, representative of the surveillance period, were selected as described previously 26,27 , and the G2P[4]-positive samples having sufficient amount of the original faecal sample available were subjected to whole genome analysis. Nearly full-length nucleotide sequences, including the 5′ and 3′ termini of gene segments encoding VP7, VP4, VP6, VP1, VP2, VP3, NSP1, NSP2, NSP3, NSP4, and NSP5/6 of the selected RVA strains, were obtained as described elsewhere 53 . RT-PCRs were performed using a One Step RT-PCR Kit (Qiagen). Each PCR product was confirmed, purified, and sequenced as described above. The genotypes of the G2P[4] RVA strains were determined according to the recommendations of the RCWG, using the RotaC online classification tool (http://rotac.regatools.be) 54 . The nucleotide sequences were deposited in GenBank under accession numbers KX171450-KX171581.
Phylogenetic analysis. The nucleotide sequences of the G2P [4] strains were compared with representative rotavirus sequences available in the GenBank database. Phylogenetic trees were constructed using the neighbour joining algorithm 55 in PHYLIP 56 and the Kimura two-parameter model in MEGA6 57 . Evolutionary distances calculated in neighbour joining analysis were based on a model described by Jukes et al. 58 . Tree topologies based on the neighbour joining analysis were evaluated by bootstrap resampling with 1,000 replicates, using the SEQBOOT and CONSENSE programs in the PHYLIP suite.
Ethics approval. Participation was voluntary, and written informed consent was obtained from all participants. The Institutional Review Board of the Chung-Ang University College of Medicine approved the protocol of this study (IRB number #2010-10-02).