Introduction

Despite the introduction of rotavirus vaccination into the national immunisation programs (NIPs) of almost 100 countries, group A rotaviruses (RVA) still remain a leading cause of deaths and hospitalisations due to gastroenteritis (GE) among children aged <5 years all over the world. Recently, the Global Burden of Diseases Study, 2015 [1] confirmed RVA to be the most common cause of diarrhoea in children younger than age 5 years, causing an estimated 259,700 (211,200-323,500) deaths per year in 2015. Prior to the widespread use of rotavirus vaccines in Brazil, RVA infection was responsible for approximately 650,000 outpatient visits, 92,000 hospitalisations, and 850 deaths annually in children under five years of age [2]. Worldwide, Brazil was among the first and largest countries to introduce a rotavirus vaccine into the public sector, covering a birth cohort of approximately 3 million infants [3]. In March 2006, the live-attenuated monovalent human-derived G1P[8] rotavirus vaccine Rotarix® (GlaxoSmithKline Vaccines, Rixensart, Belgium) was incorporated into the country’s national immunisation program with the aim of reaching 90% coverage (2 doses) among infants aged two to four months [4]. Twelve years have elapsed since the implementation of rotavirus vaccination into the public sector in Brazil, and a considerable amount of evidence has accumulated showing a substantial reduction in GE-related hospitalisations and deaths, particularly among children aged less than one year [4,5,6,7,8,9]. Of public-health importance was the fact that most of these studies also showed a decline in hospitalisations for GE among children too old to receive the rotavirus vaccine, suggesting a herd protection effect.

Rotaviruses constitute a genus within the family Reoviridae. Their genome consists of 11 segments of double-stranded RNA, which are surrounded by a non-enveloped, icosahedral triple-layered capsid [10, 11]. RVA possesses two outer layer proteins, named VP7 and VP4, which define the G (glycoprotein) and P (protease-sensitive) genotype, respectively. A total of 36 G and 51 P genotypes have been differentiated to date, of which at least 12 G types and 15 P types are commonly associated with infection in humans [11, 12]. Although over 60 G-P dual combinations have been reported in humans, five (G1P[8], G2P[4], G3P[8], G4P[8] and G9P[8]) are known to account for greater than 80% of the circulating genotypes causing childhood rotavirus gastroenteritis globally [13, 14]. Rotavirus strains with G12 genotype specificity were first detected in the Philippines in 1987 [15], but in the last decade, their prevalence has increased exponentially on a global scale, namely in combination with genotype P[6] or P[8]. Therefore, G12 is now recognised as the sixth epidemiologically important genotype associated with infections in humans [13, 16, 17]. Interestingly, genotype G3 has been reported to infect not only humans but also a broad range of other host species, and some uncommon G/P combinations, such as G3P[3] and G3P[9], are believed to be of canine, feline or human origin [18,19,20].

A comprehensive classification system has been proposed for RVA that is based on the whole viral genome and assigns a genotype to each of the 11 gene segments, as differentiated on the basis of specific cutoff points of nucleotide sequences [10, 21]. In this system, the abbreviation Gx-P[x]-Ix-Rx-Cx-Mx-Ax-Nx-Tx-Ex-Hx is used to define the genotypes of the VP7, VP4, VP6, VP1, VP2, VP3, NSP1, NSP2, NSP3, NSP4, and NSP5/6 genes, of which at least 36 G, 51 P, 26 I, 22 R, 20 C, 20 M, 31 A, 22 N, 22 T, 27 E, and 22 H genotypes has been identified in humans and/or animals [12, 22].

Although Brazil is likely to be approaching the established target of 90% rotavirus vaccination coverage, at least for some regions, rotaviruses still contribute significantly to the burden of diarrhoea in children aged <5 years, particularly in the impoverished Northern and Northeast regions [4]. Moreover, the nationwide implementation of rotavirus vaccination in Brazil in 2006 raised some concern that vaccine-induced selective pressure might have occurred, since a considerable increase in the prevalence of G2P[4] rotavirus strains was observed throughout the country, particularly during 2007-2009 [4, 23, 24]. However, more-recent surveillance studies conducted in Brazil have shown a sharp decline in the prevalence of G2P[4] genotypes followed by an increase in the occurrence of G1P[8] and others such as G12P[8], supporting the hypothesis of natural fluctuation of strains temporally and regionally, rather than a replacement mechanism [25,26,27,28].

In a previous, 3-year hospital-based monitoring study of rotavirus strains conducted in Belém, Northern Brazil, we focussed our VP7 characterisation only on the genotypes that are most prevalent worldwide which have G1, G2, G3, G4, or G9 type specificity [26, 29]. In the present study, we sought mainly to provide additional information about the circulating rotavirus strains during the early post-vaccine-introduction period in our region, including those of G and P genotypes that are currently recognised as emerging or unusual and that may eventually pose a challenge to current vaccination strategies.

Materials and methods

From May 2008 to May 2011, a hospital-based study was carried out in Belem, Northern Brazil, with two main objectives. First, to determine the effectiveness of the monovalent G1P[8] human rotavirus vaccine (Rotarix®) between May 2008 and May 2009, and second, to monitor the occurrence of circulating rotavirus strains throughout the whole study period (May 2008 to May 2011). A full description of the methods and results from these original studies were provided elsewhere, including the genotype characteristics of the viruses from 1076 rotavirus-positive samples, using G or P oligonucleotide primers targeting the G (G1, G2, G3, G4, and G9) and P (P[4], P[6], P[8], and P[9]) RVA types [26, 29]. Of these, 122 (11.3%) strains could not be fully G- and/or P-typed, and a further attempt to genotype them by including a primer for G12 was the main objective of the current study.

The study protocol was approved by the IRB Committee of the Brazilian Ministry of Health’s National Rotavirus Reference Laboratory at Instituto Evandro Chagas (IEC), under reference number 579.295, and the study was conducted in accordance with the Declaration of Helsinki and Good Clinical Practice. Written informed consent was obtained from the parents/legal guardians before enrolment.

Stool samples were prepared as 10% v/v suspension in 0.01 M Tris-Ca++, pH 7.2, and viral double-stranded (ds) RNA was extracted from the faecal supernatant using guanidinium isothiocyanate-silica as described by Boom et al. [30]. To determine electropherotypes, the extracted dsRNAs were further electrophoresed in a 5% polyacrylamide gel, followed by silver staining as described by Pereira et al. [31].

Reverse transcription polymerase chain reaction (RT-PCR) of the VP4 and VP7 genes was done for all strains, using the consensus primers 4Con3-4Con2 and Beg9-End9, respectively, as described previously [32, 33]. The G and P genotyping was performed in a second-round RT-PCR (hemi-nested PCR) using different specific primers targeting G5, G6, G8, G10 e G12 and P[1], P[5], P[7], P[10] and P[11] [32, 34].

Nucleotide sequencing was carried out with isolates that could not be genotyped previously by RT-PCR. In brief, gel-purified first-round RT-PCR amplicons from the VP7 and VP4 genes were sequenced using a BigDye Terminator Cycle Sequencing Reaction Kit v3.1 (Applied Biosystems, Foster City, CA) on an ABI Prism 3130xl Genetic Analyser (Applied Biosystems).

Sequences were analysed, and consensus sequences were prepared using the CAP3 sequence assembly program. Multiple consensus alignments were made using the software MAFFT v.7.221 [35], and sequence editing was done using the Geneious Bioinformatics software platform v.8.1.7 [36]. The data were compared with the corresponding sequences from the National Center for Biotechnology Information GenBank database using the BLAST alignment tool [37].

Molecular phylogenetic analysis was done using the IQ-TREE v.1.3.2 package for inferring maximum-likelihood trees [38]. Trees were drawn to scale with branch lengths measured in the number of substitutions per site, and the statistical significance was assessed by bootstrap resampling analysis (1000 pseudoreplicates). The resulting phylogenetic trees were visualized using the program FigTree v.1.4.2 (http://tree.bio.ed.ac.uk/software/figtree/). The sequences obtained in this study have been deposited in the GenBank database under the accession numbers MF695027-MF695065.

We further selected two G12P[6] strains with identifiable RNA profiles (2A1194 and 2A3406 with long and short electropherotypes, respectively) to perform a whole-genome analysis using specific primers for amplification of the VP1 (686 bp), VP2 (686 bp), VP3 (702 bp), VP6 (1356 bp), NSP1 (1,565 bp), NSP2 (1,038 bp), NSP3 (1,062 bp), NSP4 (738 bp), and NSP5 (664 bp) genes as reported before [39,40,41,42]. Amplicons were subjected to nucleotide sequencing, essentially as described above, and a partial nucleotide sequences for each gene segment was determined and deposited in the GenBank database under the following accession numbers: KX279732.1 (VP6), KX279702.1 (VP3), KX274667.1 (VP2), KX274608.1 (NSP5), KX274578.1 (NSP4), KX274548.1 (NSP3), KX274518.1 (NSP2), and KX274488.1 (NSP1) for the Wa-like strain 2A1194; and KX279740.1 (VP6), KX279710.1 (VP3), KX274675.1 (VP2), KX274645.1 (VP1), KX274616.1 (NSP5), KX274586.1 (NSP4), KX274556.1 (NSP3), KX274526.1 (NSP2), and KX274496.1 (NSP1) for the DS1-like strain 2A3406.

Statistical analysis was performed using SAS version 9.1 (SAS Institute Inc., Cary, NC). The statistical significance of differences between prevalence rates of genotypes over time was evaluated by the χ2-test or Fisher’s exact test (whenever applied). All p-values were two-tailed, and p < 0.05 was considered to be statistically significant.

Results

From May 2008 to May 2011, there were a total of 122 (11.3%) out of the original 1076 RVA strains that could not be partially or fully typed in a previous assessment of circulating genotypes in Belém, Brazil, because the set of primers used did not target G12 strains [26]. These strains were thus reexamined using either a broader array of primers or nucleotide sequencing. Overall (combining the previous and current genotyping assessments), G2P[4] isolates were predominant, accounting for 58.0% (625/1076) of the genotyped samples (Table 1). The second and third most prevalent strains were G1P[8] and G12P[6], representing 16.9 (182/1076) and 7.5% (81/1076) of the genotyped strains, respectively.

Table 1 Distribution of G and P genotypes identified among 1076 children hospitalised for acute rotavirus gastroenteritis in Belém, Brazil, between May 2008 and May 2011
Full size table

Among the remaining dual G and P combinations, mixed and partially or fully untypeable RVA strains represented 10.9% (117/1076) and 1.5% (16/1076) of the total samples, respectively. In addition, the globally emerging G12P[8] and the rare G12P[9] and G3P[9] genotypes were found among typed strains at rates of 0.6% (6/1076), 0.4% (4/1076) and 0.1% (1/1076), respectively.

Fig. 1 shows the electropherotypes of six representative G12P[6] rotavirus strains, of which four displayed identical long patterns with a 4-2-3-2 distribution of the RNA segments, whereas two others had a typical short electrophoretic profile also exhibiting clustering of RNA fragments into a 4-2-3-2 configuration.

Fig. 1
figure 1

Representative electrophoretic profiles of G12P[6] rotavirus strains (lanes 1-4, long RNA profiles; lanes 5-6, short RNA profiles) detected in Belém, Brazil, and controls (C1, G1P[8], long RNA profile; C2, G2P[4], short RNA profile)

Full size image

The distribution of RVA genotypes throughout the 37-month study is shown in Fig. 2. G12P[6] genotypes were identified at similar rates during the first (2.5%; 13/525) and second (3.9%; 10/256) years, and this rate jumped to 15.6% (46/295) in the third year. G2P[4] strains were found to be predominant from May 2008 to February 2010, occurring at rates that ranged from 61% to 100%. From March 2010 onwards, the prevalence rate of the G2P[4] genotype declined sharply, reaching proportions as low as 8% in December 2010. Whilst the G1P[8] genotype was not identified from May 2008 to January 2009, an increase in G1P[8] prevalence was observed from April 2010 to May 2011, yielding rates in the range of 36% to 79% of circulating strains.

Fig. 2
figure 2

Annual distribution of rotavirus genotype G and P combinations in Belém, Brazil, from May 2008 to May 2011. Others, untyped strains; mixed, mixed G and/or P types

Full size image

The rest of the isolates included a variety of uncommon rotaviruses, as well as mixed and partially or fully untypeable strains, which were in general circulating at rates <10%.

The results of analysis of the VP7 sequences of G12P[6] (n = 17), G12P[8] (n = 4) and G12P[9] (n = 4) strains are depicted in Fig. 3a. Brazilian G12P[6] and G12P[8] strains clustered within lineage III, together with recent isolates from Korea, Slovenia, the USA, Bangladesh, Italy, and Botswana with nucleotide sequence identity among these strains ranging from 95.2% to 100%. When compared with other G12 strains used as references (the first G12 strains detected in our setting), the median nucleotide sequence identity was 98.8% and 96.4% for G12P[6] and G12P[8], respectively. G12P[6] strains displaying short (n = 8) and long (n = 4) electropherotypes were found to co-circulate during the study period. A high degree of genetic relatedness (98.9%-100% sequence identity) was observed among the four Brazilian G12 strains associated with the P[9] genotype, all of which clustered in lineage II, whereas less sequence similarity (<90% identity) was seen between these strains and other G12P[9] reference strains from Brazil, Argentina and Paraguay. Nucleotide sequence analysis of P[6] genotype associated with G12 revealed that all ten strains clustered in lineage I, and their sequences were 99.1% to 100% identical (Fig. 3b). Compared with other intra-lineage reference strains, nucleotide sequence identity rates ranged from 98.2% to 100%. Results from the analysis of VP4 sequences of the P[9] genotypes, including four G12P[9] strains and one G3P[9] strain, are summarized in Fig. 3c. All Brazilian G12P[9] strains clustered in lineage II, whereas the G3P[9] strain (1A3739/2011) clustered with strains of lineage III. It was shown in our study that P[9] strains with G12-type specificity exhibited high nucleotide sequence similarity to each other (99.6%-100% identity), whereas the P[9] strain bearing G3-type specificity showed low genetic relatedness with both lineage I (86% to 88% identity) and lineage II (86.4% to 87.8% identity).

Fig. 3
figure 3figure 3

(a) Phylogenetic dendrogram based on VP7 gene sequences of local G12P[6], G12P[8] and G12P[9] strains (highlighted in shaded areas) detected during May 2008-May 2011 in Belém, Brazil. Closed triangles and circles represent G12P[6] strains displaying short and long RNA profiles, respectively. Bootstrap values based on 1000 replicates are shown. Intra-genotypic lineages are indicated in square brackets at the right. (b) Phylogenetic dendrogram based on VP4 gene sequences of local G12P[6] strains (highlighted in shaded areas) detected during May 2008-May 2011 in Belém, Brazil. Bootstrap values based on 1000 replicates are shown. Intra-genotypic lineages are indicated in square brackets at the right. (c) Phylogenetic dendrogram based on VP4 gene sequences of local G12P[9] strains (highlighted in shaded areas) detected during May 2008-May 2011 in Belém, Brazil. Bootstrap values based on 1000 replicates are shown. Intra-genotypic lineages are indicated in square brackets at the right

Full size image

A full genomic characterisation of two G12P[6] rotaviruses demonstrated that while strain 2A1194 (with a long RNA profile) possessed a Wa-like (I1-RX-C1-M1-A1-N1-T1-E1-H1) genotype constellation, strain 2A3406 (short RNA profile) possessed a DS-1-like (I2-R2-C2-M2-A2-N2-T2-E2-H2) genotype constellation.

Discussion

The present analysis represents an extension of a previous three-year surveillance follow-up of RVA strains in Belém, Brazil, circulating within five years after introduction of RVA vaccination into the Brazilian national immunisation program [26]. It was observed during the previously published study that a high proportion of strains remained untypeable for the G and/or P type, most probably due to the fact that we used a set of primers that did not target G12, which was then known to be a globally emerging genotype, and there had been a previous report on the detection of G12P[6] strains in the Northern region of Brazil [43]. In addition, nucleotide sequencing could not be performed with the strains that were previously untypeable by RT-PCR, suggesting that we might have missed G12 strains bearing either P[6] or P[8] types. In order to improve our knowledge of circulating rotaviruses in the post-vaccination period in our region, we reassessed VP7 and VP4 genotypes in the subset of strains that had not typed previously, using a set of primers targeting a broader range of genotypes.

Of particular interest in our study was the observation that G12 strains accounted for 10% of the isolates; of these, >80% were found to possess P[6] genotype specificity. G12P[6] strains were identified at rates ranging from 2.5% to 4% during the first and second years, and this rate jumped to 15.6% in the third year. This trend toward a higher frequency during the third year of follow-up is consistent with recent findings showing that G12 genotype possessing either P[6] or P[8] specificity became predominant in several countries throughout the world [13, 44]. Moreover, the proportion of circulating G12P[6] strains detected in our study appeared to be in line with the rates that were first reported in the Brazilian Amazon region [42, 45] and elsewhere [27, 28, 46,47,48,49,50], which were within the range of 1% to 17%. In this context, specific data from Brazil suggested that a G12P[6] “season” might have occurred in our region in 2010/2011.

A finding of particular interest in our study was the detection of co-circulating G12P[6] strains with long and short RNA patterns, raising the question as to whether they might belong to distinct genotype constellations. We were able to carry out a whole-genome characterization of two representative G12P[6] isolates found that that strains with long and short RNA migration patterns belonged to the Wa-like and DS1-like genotype constellation, respectively. This might eventually demonstrate that, in our surveillance study, G12P[6] rotaviruses evolved in Belém, Brazil, through the occurrence of reassortment events between locally co-circulating strains. Likewise, Nakagomi et al. [49] have reported the co-circulation of G12P[6] rotaviruses in Blantyre, Malawi, between 2007 and 2008, where strains possessing a long RNA profile displayed a complete Wa-like genotype constellation, whereas those with a short RNA pattern were found to be single VP3 gene substitution reassortants with a DS-1-like genetic backbone. Additional research at both the nucleotide and amino acid level should be conducted with these distinct G12P[6] strains in Belém for a better understanding of their evolutionary dynamics in our region.

Interestingly, phylogenetic analysis of VP7 gene sequences of all of our G12P[6] strains showed that they clustered in lineage III, which is known to include the majority of the G12 strains bearing either P[6] or P[8] genotype specificities in Brazil and elsewhere [17, 43, 44]. Of note in this context, data from the current study confirm previous findings in northern Brazil by Soares et al. [43], who demonstrated that the VP7 gene of G12P[6] rotavirus strains also belonged to lineage III. Altogether, these results show that a broader phylogenetic analysis of G12 strains that have emerged and spread across the Amazon during the past ten years is warranted and would provide a better understanding of their evolutionary genetics.

Since several pre- and post-rotavirus-vaccine-introduction studies have shown that the two licensed vaccines (Rotarix® and RotaTeq®) provide protection against a broad variety of RVA strains, including those not incorporated in these formulations, it seems improbable that the emergence of G12P[6] in our region will pose a challenge to the current vaccination strategies [29, 51, 52]. Furthermore, it should be pointed out that studies in Africa have shown that Rotarix® provides protection against diverse circulating rotavirus strains, including genotypes bearing P[6] and G12 specificities [53].

The occurrence of uncommon genotypes in our study also included the detection at low rates of five rotavirus strains bearing P[9] type specificity, combined with G12 (four strains) or G3 (one strain). Our results showing that all VP7 sequences from G12P[9] rotaviruses clustered together with lineage II of the VP4 gene are consistent with those reported previously in Korea, Japan, Argentina, Brazil and Paraguay [54,55,56,57]. In addition, the fact that P[9] is a highly prevalent genotype in the feline population raises the hypothesis of interspecies transmission, which could have generated the emerging G12P[9] strain in the Northern region [17, 58]. In order to determine the true origin of these strains, however, further sequencing studies are required, including the whole-genome characterisation of G12P[9] strains. In our study, there was a single isolate of the G3P[9] genotype of human rotavirus, which accounted for only 0.8% of all rotavirus-related gastroenteritis cases. Worldwide, the occurrence of P[9] strains has been associated with sporadic cases of gastroenteritis in humans, in general representing <2.5% of cases, even though significantly higher prevalence rates have been reported in Brazil (10.2%) and Poland (67.2%) [59,60,61]. In line with our findings, the majority of the P[9] genotypes that have been isolated elsewhere have had G1, G2, or G3 VP7-type specificity, and specifically the unusual G3P[9] combinations have been proposed to have been generated through reassortment events involving human and animal strains [62].

Interestingly, phylogenetic analysis revealed that G3P[9] strain 1A3739/2011 has a distant relationship to the prototype AU-1-like G3P[9] strain discovered in Japan in 1982, which is known to be closely related to a feline rotavirus strain, suggesting a possible zoonotic transmission from a cat to a human [63, 64]. Notably, our G3P[9] strain 1A3739/2011 clustered with another local G3P[9] strain (QUI-35-F5/2010), which was detected in an African-descendant community in Pará state, Brazil (J D P Mascarenhas, personal communication). Altogether, these findings suggest the emergence of a novel genetic variant of the P[9] genotype, possibly representing a new lineage. Surprisingly, previous molecular characterisation of the 1A3739/2011 strain in our setting revealed that it does not have an AU-like genotype constellation but rather possesses a multi-reassortant nature, with genes derived from chiropters, alpacas, equines, and simians [65].

In addition to expanding our observations on the occurrence of circulating G12P[6] strains in our setting, we identified common G1P[8] and G2P[4] genotypes at prevalence rates of 14.7% and 9.2%, respectively, among samples that had been characterised as untypeable when analysed previously [26]. While we currently do not have a plausible explanation for this, it seems likely that technical factors might have accounted for these discrepant results. It is worth mentioning in this regard that a number of studies have shown a high failure rate of primer-based genotyping, which is often attributed to single-nucleotide polymorphisms at primer-binding sites, resulting in inconsistent results [66, 67]. Moreover, Iturriza-Gomara et al. [68] have shown that amino acid substitutions at positions 87 (Ala to Thr) and 96 (Asp to Asn) within the VP7 gene of G2 rotavirus strains may be associated with failure to obtain a genotype determination. Further analysis will be performed with our local strains to assess whether possible failure of genotyping primers may have occurred due to changes in the primer-binding regions.

We found one genotype, G3P[8], that is commonly detected worldwide in humans, but further genomic characterisation was not performed to assess whether it would eventually represent a newly emergent reassortant equine-like strain, as reported in several recent studies in Brazil and elsewhere [69,70,71,72,73,74,75].

A major limitation in this study was the lack of a full analysis of the complete RVA genome constellations, which would allow a better understanding of the evolutionary dynamics of G12P[6] and G3P[9] strains spreading in our region. Furthermore, analysis of complete genome sequences would expand our knowledge on the genetic relationships between rotavirus strains of both human and animal origins.

In the context of the broad diversity of rotavirus strains circulating worldwide, this study strengthens the notion that long-term, continuous rotavirus surveillance will be essential to further clarify the overall impact of rotavirus vaccination. Our findings also provide additional evidence that simultaneous monitoring of rotavirus strains in humans and animals will lead to a better understanding of rotavirus ecology.