Introduction

Members of the species Rotavirus A, or group A rotavirus (hereafter referred to rotavirus), within the genus Rotavirus of the family Reoviridae, has been recognized as a very important etiological agent of severe gastroenteritis in infants and young children worldwide, causing more than 110 million diarrheal episodes, 25 million physician visits, 2 million hospitalizations, and more than half a million deaths among children less than 5 years of age, mainly in developing countries [34, 35]. The rotavirus genome consists of 11 segments of double-stranded RNA encoding six structural viral proteins (VPs) and six nonstructural proteins (NSPs); each genome segment codes for a single viral protein, with the exception of genome segment 11 which encodes two proteins (NSP5 and NSP6) [9]. VP7 and VP4 are structural proteins constituting the outer layer of the virion and the viral spikes, respectively. These two proteins define the serotype of the virus and are therefore considered to be critical for vaccine development because they are the target of neutralizing antibodies that may provide both serotype-specific and, in some instances, cross-protective immunity [18]. However, the correlates of protection were recognized as not being restricted to high levels of neutralizing antibodies [6]. Based on the antigenicity of these two outer capsid proteins, a dual classification system by G serotypes (G for glycoprotein, coded for by genome segment 7, 8 or 9, depending on the strain) and P serotypes (P for protease-sensitive protein, coded for by genome segment 4) has been adopted [9]. However, serological assays are now being replaced by molecular typing; hence, the term “serotype” is being replaced by “genotype”, and these two terms are used almost as synonyms. There is an exact correlation between G serotype and G genotype, thereby allowing the use of the same numbering system. Unfortunately, different numbering systems are adopted to designate P serotype and P genotype, with P genotype being designated within a squared bracket. While more than a dozen of G types and P types have been reported for strains detected in humans and animals and the number of potential G/P combinations is many hundred, only five major G and P combinations commonly occur in human rotaviruses; they are G1P[8], G2P[4], G3P[8], G4P[8] and G9P[8] [7, 12, 44].

Recognition of the large global burden of rotavirus disease has propelled vaccine development over the last few decades. As a result, there are currently two live-attenuated oral rotavirus vaccines licensed on the global scale. They are the monovalent, human-rotavirus-based vaccine, Rotarix, (GlaxoSmithKline Biologicals, Belgium) and the pentavalent, bovine-human reassortant vaccine, RotaTeq, (Merck & Co, USA) [42, 49]. Both vaccines have made a remarkable impact on the reduction of the number of rotavirus hospitalizations in high-income and middle-income countries [3, 4, 16, 41]. Recently, a marked increase and dominance of G2P[4] rotavirus strains was observed concurrently with the implementation of a universal mass vaccination program with the monovalent vaccine in Brazil [17], although there was also a simultaneous marked decrease in the detection rate of rotavirus [31]. It was hypothesized that vaccination with the monovalent G1P[8] vaccine created conditions in which G2P[4] rotavirus strains acquired a selective advantage over P[8] strains [31]. However, a similarly high prevalence of G2P[4] rotavirus strains was observed in Latin American countries where no rotavirus immunization program was implemented. For example, Esteban et al. [10] observed that G2 strains were the most prevalent during 2004 (44%) and 2007 (58%) in Argentina. Patel et al. [36] reported that surveillance in El Salvador, Guatemala, and Honduras showed that G2P[4] was the predominant circulating strain in 2006 (68%–81%). It was also reported from Paraguay that G2P[4] was predominant in 2006 (64%) and 2007 (46%) after a 6-year absence of this genotype [29]. Recent studies in countries such as Sierra Leone, Egypt, Jordan, Oman and Yemen have shown that G2P[4] is the most prevalent [22, 24]. Furthermore, during the post-licensure effectiveness study of Rotarix in El Salvador, de Palma et al. did not observe any increase in G2 strains [5]. Thus, these authors suggested that the predominance of G2P[4] was due to natural genotype variation and was unrelated to the introduction of the monovalent rotavirus vaccine.

However, there are several epidemiological studies in which re-emergence or sharp increase in relative frequency of G2 strains in the population was associated with amino acid substitutions in the antigenic regions of the VP7 protein that might allow such variants to escape from recognition by neutralizing antibodies [28]. In Taiwan, for example, G2 strains re-appeared in 1992 after a five-year interval and caused an epidemic in 1993 [50]. The re-emergence and epidemic of G2 strains in Taiwan was accompanied by an amino acid substitution at position 96 from aspartic acid to asparagine (hereafter abbreviated as D96N) [50]. The amino acid at position 96 was previously described as an epitope recognized by a G2-specific neutralizing monoclonal antibody (RV5:3) [28]. Furthermore, the G2 strains that re-emerged in the UK in 1995-1996 as the second-most common genotype (16-18%) possessed this amino acid substitution and were unable to be serotyped by the enzyme-linked immunosorbent assay using monoclonal antibody RV5:3 [19]. More recently, in Thailand, there was a sharp increase in G2 strains from almost negligible percentages in 2001 and 2002 to 83% in 2003 [23], and these emergent G2 strains possessed the D96N substitution [47].

During surveillance studies in Nepal since 2003 [38, 48], where no rotavirus vaccine has been introduced, we noticed a sudden increase in the detection rate of G2 strains to 39% in 2004-2005 as compared to 1% in the preceding season (2003-2004). This observation prompted us to sequence the VP7 gene of Nepalese G2 strains, for which there had been no sequence information available, in order to examine the presence of the D96N substitution. More recently, a further amino acid substitution (serine to asparagine at 242: S242N) was described based on Brazilian G2 strains to constitute a new sublineage within the lineage hallmarked by the D96N substitution [30]. Thus, we extensively searched the DNA database and extracted relevant G2 VP7 sequences deposited over the last 34 years from across the world. We then conducted a phylogenetic analysis in order to better understand these and other major substitutions in the VP7 antigenic regions from the global and evolutionary perspective and to provide a basis on which to assess the changes in the G2 VP7 genes as rotavirus vaccines penetrate into broader areas of the world.

Material and methods

Specimens

Among the stool specimens that were collected from children with acute diarrhea attending Kanti Children’s Hospital, Kathmandu, Nepal, from August 2004 through July 2005, we previously reported that there were 67 G2 rotavirus-positive specimens identified by the reverse transcription (RT)-PCR genotyping method [38]. For this study, there remained 57 genomic RNAs available, which were extracted using the QIAamp Viral RNA Mini Kit (QIAGEN Sciences, Germantown, MD, USA).

Nucleotide sequencing

Genomic RNAs that were extracted previously using the QIAamp Viral RNA Mini Kit (QIAGEN Sciences, Germantown, MD, USA) were reverse transcribed and amplified with a pair of VP7 consensus primers, Beg9 and End9 [15], using the AcessQuick RT-PCR system (Promega Corporation, Madison, WI, USA). The amplified products were then purified using a QIAquick PCR purification kit (QIAGEN) according to the manufacturer’s instructions. Nucleotide sequencing reactions were performed by fluorescent dideoxy chain termination chemistry using the BigDye Terminator Cycle Sequencing Ready Reaction Kit, version 3.1 (Applied Biosystems, Foster City, CA, USA), and nucleotide sequences were determined using an ABI Prism 3730 Genetic Analyzer (Applied Biosystems). The nucleotide sequences thus obtained were aligned using the Megalign program in the Lasergene 8 software package (DNAstar, Inc. Madison, WI, USA).

Compilation of the G2 VP7 sequences in the DNA databases

In order to place the VP7 sequence results from Nepalese strains into a broader evolutionary context, we made an exhaustive DNA database search using the BLAST program (http://blast.ncibi.nlm.nih.gov/) with the DS-1 VP7 sequence as a query sequence and retrieved 500 sequences so that all potential G2 sequences might be included. We then used for our analysis only those G2 VP7 sequences that included at least 615 nucleotide sequences from position 307 to position 921 (corresponding to amino acid residues 87-291) and for which information about the year and the place of isolation were available either in the DNA database entry or in the relevant literature (a list of all sequences used, including names of RV strains, year of isolation and database accession number, is provided in Supplementary Data 2). Accordingly, the number of G2 VP7 sequences we included from the DNA database was 339.

Phylogenetic analysis of the G2 VP7 genes

Phylogenetic analysis was performed using the MEGA4 package [46]. The nucleotide sequences of Nepalese G2 strains as well as those compiled from the DNA database were aligned using the CLUSTALW program, and the genetic distances between sequences were calculated by the Kimura two-parameter method [25]. A phylogenetic tree was then constructed by using the neighbor-joining method [43]. The statistical significance at the branching point was calculated with 1,000 pseudo-replicate datasets. A lineage was defined as a cluster of sequences having a bootstrap probability of 70% or higher at the branching point. Similarly, within the lineage, a sublineage was defined as a cluster of sequences that had a bootstrap probability of 70% or higher. In addition, amino acid substitutions in the antigenic regions of VP7 were analyzed, and their pattern was defined for the substitutions of the amino acids at residues 87, 96, 213, and 242.

Chronological and geographical variations of G2 VP7 genes

All G2 VP7 sequences were categorized according to the place of detection into five groups: Asia, Europe, Africa, Americas and Oceania, and also according to the years of detection into seven quinquennial periods: 1976-1980, 1981-1985, 1986-1990, 1991-1995, 1996-2000, 2001-2005 and 2006-2009. The number and the proportion of strains belonging to each lineage or sublineage were counted for each of the abovementioned continents or periods.

Results

Of 67 rotavirus specimens that were typed as G2 in the study performed in Nepal in 2004-2005 [38], we were able to amplify and sequence the entire open reading frame of the VP7 gene from 45 specimens. The nucleotide sequence identity between any pair of those 45 VP7 sequences ranged from 97.6 to 100%, and their deduced amino acid identity ranged from 97.6 to 100%. As for the amino acid residues that were previously identified as important epitopes in the antigenic regions of VP7 [2, 13, 14, 19, 28, 30, 33, 47, 50], all of 45 Nepalese sequences had the same amino acids: threonine at position 87 (antigenic region A), asparagine at 96 (region A), aspartic acid at 213 (region C), and asparagine at 242 (region F).

A phylogenetic tree was constructed using the 45 Nepalese G2 nucleotide sequences determined in this study and 339 G2 sequences that were detected between 1976 and 2009 in 31 countries on five continents and were available in the DNA database. In this global collection, four strains were detected between 1976 and 1980, 21 were detected between 1981 and 1985, 11 were detected between 1986 and 1990, 54 were detected between 1991 and 1995, 43 were detected between 1996 and 2000, 86 were detected between 2001 and 2005, and 120 strains detected between 2006 and 2009.

Four lineages were identified in the phylogenetic tree (Fig. 1). The first lineage (I) contained eight G2 strains from the United States of America, Australia, Taiwan and South Africa that were detected in the 1970s and 1980s. The second lineage (II) contained 16 G2 strains detected between 1992 and 2004 from South Africa, Australia, Italy, Russia and Brazil. The third lineage (III) contained only two strains from Japan detected in a single year, 1980. The fourth lineage (IV) contained 358 G2 strains detected from 1981 to 2009. In this large lineage IV, one sublineage (IVa) that contained 306 strains was identified with a bootstrap probability of 92%. The rest of the lineage IV sequences, which included 52 sequences detected in Asia, Africa, Australia and Europe, were not considered to form a sublineage due to a low bootstrap probability (19%), and their detection spanned over the period from 1981 in Taiwan to as recently as 2008 in Germany (Fig. 1).

Fig. 1
figure 1

A simplified phylogenetic tree constructed by the neighbor-joining method based on the VP7 nucleotide sequences of human rotavirus genotype G2 strains (The nucleotide sequences compared spanned a minimum of 615 nucleotides from position 307 to position 921, which corresponded to amino acid residues 87-291). This tree was rooted with porcine G2 rotavirus strains. Since the number of sequences belonging to each lineage or sublineage is too large to be presented in a figure, sequences are condensed into a triangle to the right of which information about the geographic origin and the year of detection of relevant sequences is given. At the right-hand side of the figure is a designation of lineages and sublineages identified in this study. The genetic distance (substitutions per site) is indicated at the bottom. Percent bootstrap support is indicated by the value at each node when the values were 70% or larger. Notice one exception, where there is a triangle with a bootstrap support of 19%, which conglomerates sequences that belonged to lineage IV but did not form a single sublineage. The complete version of this phylogenetic tree is in Supplementary Data 1

Full size image

Within sublineage IVa, further sublineages were identified with sufficiently high bootstrap probabilities of 70-100%: sublineage IVa-1, which contained 123 sequences from five continents over the period between 1992 and 2009, the small sublineage IVa-2, which contained seven sequences only from Africa in 1997 and 1999, and sublineage IVa-3, which contained 171 sequences from Asia, Africa, Europe and Americas, but not from Oceania, over the period between 1993 and 2009. All 45 Nepalese G2 sequences determined in this study were clustered in sublineage IVa-3 (Fig. 1). The absence of sequences with sublineage IVa-3 from Oceania may reflect the lack of G2 sequence deposition in the DNA databases from Australia since the middle of the 1990s. The vast majority (97%) of G2 strains detected in the last decade belonged to sublineage IVa. Exceptions were two strains detected in Brazil in 2002 and one strain detected in Russia in 2004 that belonged to lineage II, three strains detected in Thailand in 2001 and 2005, and one strain from Germany in 2008 that were found among the rest of lineage IV (Fig. 1).

Pairwise sequence identities between VP7 sequences were calculated. There were two distinct peaks, one corresponding to identities between lineages and the other corresponding to identities within the same lineages, lending support for the division of global G2 sequences into four lineages (Fig. 2). The peak for the former was 93% and that for the latter was 96%.

Fig. 2
figure 2

Pairwise sequence identities of VP7 genes: The number of the occurrence of different percentages of identities between lineages is represented by open square columns and that within lineages is represented by filled square columns

Full size image

The pattern of amino acid substitutions occurring in each lineage (and sublineage) of the G2 VP7 genes was analyzed relative to the DS-1 strain (lineage I) with respect to the amino acids present at four positions in the VP7 antigenic regions previously recognized as potential epitopes for G2 strains: at 87 and 96 in antigenic region A (residues 87-101), 213 in region C (residues 208-221) and 242 in region F (residues 233-242). The lineage I sequences had the same amino acids at all four positions (Fig. 3). The lineage II sequences had three amino acid substitutions from the lineage I sequences: A87T (alanine to threonine), N213S (asparagine to serine) and N242S (asparagine to serine). While the lineage II sequences, with two exceptions, shared the N213S and N242S substitutions, the A87T substitution occurred only in half of the lineage II sequences, and the amino acid remained unchanged at this position in the other half of the sequences (Fig. 3). The lineage III sequences had two substitutions (A87T and N242S) from the lineage I sequences (Fig. 3). The sequences in lineage IV that did not belong to sublineage IVa (indicated as lineage IV except IVa in Fig. 3) had three amino acid substitutions from the lineage I sequences: A87T, N213D and N242S. While they all shared the N213D and N242S substitutions, the A87T substitution occurred in only one third of the sequences, and the amino acid remained unchanged at this position in the remaining two thirds of the sequences. The salient feature of sublineage IVa sequences is the D96N substitution. What was notable within this sublineage IVa was the amino acid substitution at 242; there was serine in the IVa-1 and IVa-2 sublineages, whereas there was asparagine in the IVa-3 sublineage (Fig. 3).

Fig. 3
figure 3

The amino acids at residues 87, 96, 213 and 242 in the antigenic regions A (residues 87-101), C (residues 208-221) and F (residues 233-242) of the VP7 sequences that belong to each of the lineages and sublineages identified in this study. While the amino acids present at these four residues were examined for all 384 G2 VP7 sequences, including 45 Nepalese sequences (and provided in Supplementary Data 2), sequences were selected to present in the figure such that there was no duplication of the sequences from the same geographical location and in the same year except in cases where amino acids at any of these four positions were different

Full size image

Figure 4 depicts the temporal distribution of different lineages and sublineages of the G2 VP7 sequences over the last 34 years since 1976. Although the number of sequences available was limited for lineages I-III, there was a successive lineage shift from lineage I to lineage II, then to co-circulation of sublineages IVa-1 and IVa-3. Of note was the fact that the persistence over seven quinquennial periods of the sequences that clustered into lineage IV but that did not belong to sublineage IVa (indicated as sublineage IV except IVa in Fig. 4), although the number of sequences identified in the last quinquennial period was only one (the one detected in Germany in 2008). Lineage I was present only for a limited time period in the last quinquennial period in the 1970s and were never detected thereafter (Fig. 4). The strains belonging to lineage II circulated at low frequencies between 1991 and 2005, yet their circulation was global on the five continents.

Fig. 4
figure 4

The number of G2 VP7 sequences belonging to each lineage and sublineage during seven quinquennial periods: 1976-1980, 1981-1985, 1986-1990, 1991-1995, 1996-2000, 2001-2005 and 2006-2009. Note, however, that the bars indicated as lineage IV except IVa and sublineage IVa-except IVa1-3 did not actually form a single lineage but a collection of sequences that did not belong to IV or IVa

Full size image

Discussion

This study demonstrates that a sharp increase in the relative frequency of G2 strains in Nepal in 2004-2005 was accompanied by the D96N substitution in the VP7 gene, as was observed in the studies conducted in Taiwan [50], the United Kingdom [19], South Africa [32, 33], Italy [2], and Thailand [47]. Moreover, although the presence of the D96N substitution was not explicitly described in other reports in association with the re-emergence or an increase of G2 strains, this association was confirmed by our analysis. For example, in Paraguay, 39% and 64% of circulating strains in 2005 and 2006, respectively, were G2, whereas no G2 strains were reported from 2000 to 2004 [29]. The VP7 genes of all seven strains reported in the paper by Martinez et al. [29] contained the D96N substitution, and they belong to sublineage IVa according to our classification. In Bangladesh, there was a constant circulation of G2 strains at low frequencies during the early years of the 2000s, and then there was a sharp increase to 43% in 2005-2006 [40]. In the same or nearby regions in Bangladesh, other investigators also reported a high detection rate of G2 strains (43-54%) during the same period [1, 8, 37]. The nucleotide sequence data from these studies, as far as they were available in the databases, were analyzed, and it was shown that there was the same D96N substitution and that all strains analyzed belonged to sublineage IVa.

However, it is important to understand such substitutions of G2 sequences in the global and evolutionary context. For this purpose, we did an extensive database search that included 339 G2 sequences deposited over the last three decades, analyzed the amino acid substitutions having occurred at the four potential neutralization epitopes in the VP7 antigenic regions, and examined whether there was any distinct relationship between the phylogenetic lineages of G2 VP7 genes and the observed amino acid substitutions. We found that each lineage or sublineage had essentially a set (or two sets) of amino acid combinations at four previously recognized amino acid positions in the G2 VP7 antigenic regions (Fig. 3). Most notably, this study showed that a single D96N substitution was associated with the emergence of sublineage IVa. The first detection of strains carrying this sublineage sequence was in India in 1987. It then increased in the following years and rapidly dominated globally in the 2000s (Fig. 4). It is therefore tempting to speculate that current, globally predominant G2 rotavirus strains originated in Asia.

Similarly notable was the observation that there were two amino acid substitution patterns within sublineage IVa, where the difference was whether serine or asparagine was present at position 242; one was serine at 242 and the other asparagine. Serine at 242 was the hallmark substitution of sublineages IVa-1 and IVa-2 (in addition to the A87T, D96N and N213D substitutions), whereas asparagine at 242 was characteristic of sublineage IVa-3. However, it was not clarified in this study whether the asparagine at 242 in the sublineage IVa-3 sequence represented a back mutation from serine that was present in lineage II and later lineages or the original amino acid, which had remained unchanged from the lineage I sequence. While both sublineage IVa-1 (which had serine at 242) and sublineage IVa-3 (which had asparagine at 242) were abundantly detected at similar frequencies in different geographic locations (mostly Asia, the Americas, and Africa), there was a significant difference in their detection rates in the respective quinquennial periods over the last two decades. The rate of detection of sublineage IVa-1 remained rather constant, meaning its proportion among sublineage IVa sequences was diminishing. In sharp contrast, there was an increasing number as well as an increasing proportion of sublineage IVa-3 sequences detected, from a few sequences (5-10%) in the 1990s to 80 sequences (70%) in the 2000s. In this context, the recent observation made about 15 Brazilian G2 sequences by Mascarenhas et al. [30] is noteworthy: local Brazilian G2 sequences with the D96N substitutions were clustered into two sublineages IIa and IIc, which correspond to sublineage IVa-1 and IVa-3 in this study, respectively. They described the presence of the former sequences (serine at 242) in 2006 and that of the latter (asparagine at 242) in 2006 (only one sequence) and 2008. Also in Brazil, Gomez et al. [13] reported that, while all 54 G2 sequences detected between 2005 and 2009 had the D96N substitution (belonging to sublineage IVa), serine at 242 (sublineage IVa-1) was observed in the first three years (2005-2007) and asparagine at 242 (sublineage IVa-3) was observed in the last four years (2006-2009). Thus, it appears that there was a shift of sublineage from IVa-1 to IVa-3 in Brazil. Whether or not this sublineage shift is global is yet to be determined.

Also worthy of mention is the persistence of the sequences in lineage IV that did not belong to sublineage IVa (indicated as lineage IV except IVa in Figs. 3 and 4). All of these sequences had two amino acid substitutions from lineage I (N213D and N242S) and they might have contributed to the predominance of these “lineage IV except IVa” sequences in the three quinquennial periods between 1981 and 1995 (Fig. 4). Thereafter, these sequences decreased substantially, but they persisted for another three quinquennial periods. Such persistence may possibly be related to the abundance of genetic diversity, as indicated by the fact that they did not form a single lineage, and the small minor sublineages may have represented surviving sequences that found their niches.

Lineage II sequences were possessed only by G2 strains from five countries that were detected during the 1990s, with the exception of two G2 strains from Brazil in 2002 and one strain in Russia in 2004. It is of particular interest that only one G2 strain in lineage II was detected in Asia despite the fact that two-thirds of strains analyzed in this study were from Asian countries.

Reviewing the evolutionary history of G2 VP7 sequences over the last three decades with reference to lineage diversification in the phylogenetic tree and amino acid substitutions occurring in the VP7 antigenic regions, it is remarkable that virtually 100% of the VP7 sequences of contemporary G2 strains belong to sublineage IVa and that this shift was accompanied by the D96N substitution. In this regard, it should be mentioned that the amino acid at residue 96 was previously identified as a neutralizing epitope in an experiment using neutralization monoclonal antibody escape mutants [28]. It is therefore very likely that the D96N substitution, from a negatively charged polar amino acid (aspartic acid) to an uncharged polar amino acid (asparagine), contributed to a major antigenic change that may have helped the emergence and substantial increase of G2 strains observed in various geographic locations, including Nepal. It should be noted, however, that the high prevalence of G2 strains of sublineage IVa may not only be due to the D96N substitution in the VP7 gene but also to ongoing reassortment events within G2 strains involving various genes, as observed by Gomez et al. [13]; in this way, emerging variants may gain a competitive advantage (fitness) in particular population environments to become dominant strains. It would be of interest to investigate whether similar reassortment events were involved in the emergent Nepalese G2 strains by examining the other genome segments.

There are some limitations in this study. First, the unavailability of G2 rotaviruses before the abrupt increase in 2004-2005 in Nepal made less conclusive the association between the G2 increase and the D96N substitution. Second, uneven availability of sequence information may have produced a distorted picture of the evolutionary history of the G2 VP7 genes; most notably, the VP7 sequences of G2 strains detected in Europe and Australia in the most recent quinquennial period are lacking. Overcoming the latter limitation is considered feasible, since rotavirus strain surveillance has been actively performed in such countries as Australia, UK, and Italy for long periods of time [11, 20, 21, 26, 45].

In conclusion, a sharp increase in the relative frequency of G2 strains in Nepal in 2004-2005 occurred concurrently with the D96N substitution that was the hallmark of strains belonging to sublineage IVa, by which virtually all globally circulating G2 strains were replaced. However, within sublineage IVa, further variants emerged belonging to sublineage IVa-3 that had an additional S242N substitution and then became predominant. All Nepalese strains analyzed in this study belonged to this sublineage. Since there is continuous emergence of variants that have amino acid substitutions in the four potential epitopes in the antigenic regions of VP7, it is likely that rotavirus evolves by escaping the immunity formed in the population. Given the limited spectrum of amino acid substitutions observed in the G2 VP7 antigenic regions, directional selection rather than diversifying selection is likely to have operated on these amino acid residues. Since an increasing number of molecular evolutionary methods for detecting natural selection are available [27, 39], it would be interesting to determine the kind of natural selection forces that drive the evolution of the G2VP7 genes. In addition, it would also be interesting to investigate whether reassortment events were involved in the emergent Nepalese G2 strains, as was documented for the G2 strains from Brazil [13].