Genetic Characterisation of South African and Mozambican Bovine Rotaviruses Reveals a Typical Bovine-like Artiodactyl Constellation Derived through Multiple Reassortment Events

This study presents whole genomes of seven bovine rotavirus strains from South Africa and Mozambique. Double-stranded RNA, extracted from stool samples without prior adaptation to cell culture, was used to synthesise cDNA using a self-annealing anchor primer ligated to dsRNA and random hexamers. The cDNA was subsequently sequenced using an Illumina MiSeq platform without prior genome amplification. All strains exhibited bovine-like artiodactyl genome constellations (G10/G6-P[11]/P[5]-I2-R2-C2-M2-A3/A11/A13-N2-T6-E2-H3). Phylogenetic analysis revealed relatively homogenous strains, which were mostly related to other South African animal strains or to each other. It appears that these study strains represent a specific bovine rotavirus population endemic to Southern Africa that was derived through multiple reassortment events. While one Mozambican strain, MPT307, was similar to the South African strains, the second strain, MPT93, was divergent from the other study strains, exhibiting evidence of interspecies transmission of the VP1 and NSP2 genes. The data presented in this study not only contribute to the knowledge of circulating African bovine rotavirus strains, but also emphasise the need for expanded surveillance of animal rotaviruses in African countries in order to improve our understanding of rotavirus strain diversity.


Introduction
Rotavirus is an enteric pathogen that affects the young of many mammalian and avian species [1]. The virus belongs to the Reoviridae family and contains an 11-segmented double-stranded RNA (dsRNA) genome. The genome encodes six structural proteins that form the viral triple-layered particle (TLP), which consists of the outer capsid (VP4 and VP7), the inner capsid (VP6) and the core (VP2 encasing VP1 and VP3) proteins that enclose the nucleic acid material. The viral genome also encodes five or six non-structural proteins, NSP1-6 [2].
Rotavirus, which is transmitted via the faecal-oral route, is a common cause of diarrhoea in calves, which can cause economic loss either by mortality or affecting the growth of animals. Severity of disease ranges from asymptomatic carrier animals to mild, self-limiting diarrhoea and, in severe cases, dehydration and death. Mortality is influenced by various factors, including virulence of strains, age of the host and environmental stresses. Although infections are mostly mild, rotavirus is associated with high morbidity [6]. However, in many countries, rotavirus infections in cattle or other animals are not reported. A global review of bovine rotavirus in 24 countries reported prevalence as 33.7% (n = 14,076) [7]. The only African countries represented in this study were Tunisia and Nigeria. The majority of bovine rotaviruses had a G6 (50.7%), followed by a G10 (20.6%) G type and the most common P types were P [5] (25.9%) and P [11] (21.5%) [7].
The north-eastern region of South Africa shares a border with the south-western region of Mozambique. Due to the lack of bovine surveillance data, rotavirus prevalence in these countries is not known but a vaccine for cattle, Rotavec ® Corona (MSD Animal Health), is available in South Africa. Only limited data for whole genomes of RVA strains originating from various animal species have been described from South Africa. These include strains detected in African buffalo [8], sable antelope [9], horses [10] and cattle species [11,12]. No bovine rotavirus strains have been reported from Mozambique, although a partial strain of possible animal origin was detected in a child, suggesting interspecies transmission [13].
South Africa and Mozambique are both sub-Saharan African countries where rotavirus infections have a major public health impact [14]. Interspecies transmission and reassortment of rotaviruses increase the diversity of the human rotavirus population which can influence vaccine effectiveness. Transboundary transmission of strains between animals will also impact the diversity of rotavirus circulating in a specific population. Characterisation of animal strains is therefore important to aid in the understanding of the genetic diversity of rotavirus strains and their host range specificity. In this study, the whole genomes of bovine rotavirus strains detected in South Africa and Mozambique were characterised.

Genome Assembly and Genotyping
Genomic data were generated for all seven samples and average coverage for these sequences ranged from 158.2 to 10,555.7 per sequence (Supplementary Materials Table S1). Full-length consensus sequences were assembled for all strains, Bov7, Bov4, MRC-DPRU457, Bov1, MPT93 and MPT307, except for 1162. All the gene sequences of strain 1162 covered at least 80% of the coding region except for the NSP5-encoding sequence, for which only 61.5% of the coding region was obtained (Supplementary Materials Table S1).

I2
In the VP6 phylogenetic tree, the study strains grouped in three different clusters independent of their G and P genotypes. Four study strains (Bov1, Bov4, Bov7 and MRC-DPRU457) formed a distinct monophyletic cluster with other South African bovine strains, two detected in the Western Cape (RVA/Cow-wt/ZAF/1603/2007/G6P [5] and RVA/Cowwt/ZAF/1605/2007/G6P [5]) and RVA/Cow-wt/ZAF/MRC-DPRU3010/2009/G6P [5], as well as a South African porcine strain, RVA/Pig-wt/ZAF/MRC-DPRU3878/2008/G5P[X]. MRC-DPRU457 was identical to an unpublished South African strain, RVA/Cow-wt/ZAF/ MRC-DPRU456/2009/G6P [11] (Supplementary Materials Table S2). These strains shared 98.49-99.16% nucleotide identity (99.2-99.7% amino acid identity) (Figures 1 and 2, Supplementary Materials Table S2). The VP6 sequences of MPT307 and 1162 clustered with two buffalo strains from South Africa to form a minor South African cluster in lineage X (Figure 2, Supplementary Materials Table S2). These buffalo strains were detected in 2002 and 2007 in Limpopo province [8] (Figure 1). Finally, the VP6 sequence of the Mozambican strain, MPT93, clustered in lineage VI with bovine strains as well as some human strains and shared a 98.74% nucleotide identity (99.5% amino acid identity) with a human strain from India ( Table S2). The phylogenetic relationships of this Indian strain in not known. Additionally, the VP6 origins of the human strains in this lineage was inconclusive in previous studies [16,17]. However, the clustering with bovine strains in this study strongly suggests a bovine origin for these human strains.

T6
The T6 NSP3 genes of the study strains formed two sub-clusters. MRC-DPRU457 clustered with the two Mozambican strains, MPT307 and MPT93, and formed a diverse monophyletic cluster with previously identified South African animal strains that was distinct from global strains.  Table S2).

E2
The E2 NSP4 genes of the study strains formed three sub-clusters; two of which fell within lineage XV. MRC-DPRU457 and MPT307 clustered with previously characterised South African animal strains and a bat strain form Zambia in the southern African region. MPT93, Bov1 and 1162 clustered with a South African buffalo strain (RVA/Buffalowt/ZAF/1442/2007/G10P [11]) and a French bovine strain (RVA/Cow-xx/FRA/DijonA03 6/2006/GXP[X]) ( Figure 2). These sub-clusters fell within a lineage predominantly comprised of strains from the African continent. Bov7 and Bov4 clustered in lineage XVIII and were closely related to two South African bovine strains previously detected in the Western Cape in 2007 [11] (Figure 1).
These study strains contained genotypes typically associated with rotavirus detected in bovine hosts: G10/G6-P [11]/P[5]-I2-R2-C2-M2-A3/A11/A13-N2-T6-E2-H3. It is note-worthy that three of the study strains contained the rarely identified A13 genotype adding to the 14 complete sequences in the NCBI GenBank. The only other published study describing an African strain with this genotype was detected in an African buffalo [8]. The study strains had the same genotypes for all segments, except those encoding for VP7, VP4 and NSP1. However, even though the same genotypes were observed for eight of the genes, none of the study strains contained similar sequences across their genomes, indicating that they have probably evolved through various reassortment events and extensive genetic drift. This degree of diversity observed suggests the circulation of multiple, distinct subtypes/alleles in endemic bovine strains in the region that have evolved over decades.
The majority of the study strains almost always clustered with each other or other African animal strains. Of note, study strain RVA/Cow-wt/ZAF/MRCDPRU457/2009/ G10P [11] clustered with the contemporary, previously characterised strain RVA/Cowwt/ZAF/MRC-DPRU456/2009/G6P [11] across all genes analysed, except for the VP7 en- This indicates prolonged and extensive circulation of rotavirus strains throughout the two countries. However, the lack of whole genome data of animal rotavirus strains from South Africa and Mozambique must be noted. Unsampled local diversity is also represented in the long branch lengths in some of the trees and the low nucleotide identity between the study strains and their closest relatives.
In some cases, the study strains clustered with rotaviruses detected in humans from African countries such as the NSP5 sequence of Bov7 which grouped with RVA/Humanwt/MOZ/0060b/2012/G12P[8]P [14] in a separate clade. The majority of the strains in this clade was, however, detected in animals and 0060b was also described as a bovine rotavirus strain detected in a child [13]. This indicates that Bov7 has a bovine-like NSP5 sequence. Similarly, the VP3 sequence of MPT93 clustered in lineage VI with human strains, mostly which are suspected to be a result of interspecies transmission events between humans and cows [21].
The VP1 sequence of MPT93 may also be derived from an interspecies reassortment event as it clustered in lineage V with mostly human strains. The closest relative was a human rotavirus strain, RVA/Human-tc/MWI/QEC287/2006/G8P [8] with a shared nucleotide identity of 98.48% (Figure 2; Supplementary Materials Table S2). This human rotavirus strain was not associated with any interspecies transmission events [16] and the rest of the strains in lineage V were also detected in humans, suggesting that the VP1 sequence of MPT93 is human-like. The NSP2 sequence of MPT93 clustered with strains detected in humans and animals (yak, goat and cow) in lineage V. There is evidence that the NSP2 sequences of the bovine strains (RVA/Cow-tc/THA/A5-10/1988/G8P [1] and RVA/Cowtc/THA/A5-13/1988/G8P [1]) as well as the goat strain (RVA/Goat-xx/BGD/G034/1999/ G6P [1]) are more closely related to human NSP2 sequences [21,22]. This points to a human NSP2 sequence in MPT93; however, the long branch lengths in this lineage suggests unsampled global diversity and increased sequencing of diverse strains from varied hosts may refine the origins of strains in this lineage. Segments VP6, VP2, VP3 and NSP1 of MPT93 clustered with strains from India, Bangladesh and Italy, although the nucleotide identity varied between 94.35 and 98.64%. This indicates that MPT93 is divergent from the rest of the study strains, which are, for the most part, endemic to Southern Africa.
The other Mozambican strain, MPT307, clustered closely to the South African strains for all the segments except that of NSP1-A13, indicating a close relationship between these strains. Additionally, these study strains group closely to strains from other African coun-tries indicating possible transboundary movement of animals and subsequent transmission events. The frequency and extent of transboundary movement in the region is, however, unknown. This observation calls for combined regional efforts between veterinary services to manage infections across international borders.
The limited data available due to local subsampling complicate full characterisation of these study strains but at the same time highlight the significant contribution this study makes to current knowledge about bovine rotaviruses in Africa. Apart from the limited knowledge, the data presented in this study represent a diverse genetic pool of bovine rotavirus strains in Africa that are shaped by extensive reassortment events. These strains circulated for more than a decade in various geographical regions, across country borders. Extensive reassortment is seen, resulting in endemic variants, as well as interspecies transmission in the VP1 and NSP2 genes. Results suggest that transboundary movement and interaction of hosts influence the diversity of rotavirus in Africa. This study calls for extensive surveillance of bovine rotavirus in African countries to understand bovine rotavirus diversity and the extent of interspecies transmission.   Figure 1) and tested positive for the presence of rotavirus using the ProSpecT ™ Rotavirus EIA kit (Oxoid, Ely, UK) and electropherotyping. The two Mozambican samples were collected as part of an exploratory study in 2016 from the Manhiça and Marracuene districts in the Maputo province ( Figure 1). As it is known that animals can be asymptomatic, samples were collected irrespective of clinical signs (both diarrhoetic and non-diarrhoetic) and sex from newborn to one-year-old animals. Samples were tested for the presence of rotavirus with the ProSpecT ™ Rotavirus EIA kit (Oxoid, Ely, UK) at the Direcção de Ciências Animais (DCA), Directorate of Animal science. The South African samples were taken from commercial herds where all the adult cows had been vaccinated against rotavirus. The Mozambican samples were taken from animals at informal non-commercial smallholdings without any vaccination.

RNA Extraction, cDNA Synthesis and Sequencing
RNA extraction was performed as previously described [23]. Briefly, total RNA was extracted with Tri-Reagent (Sigma) and single-stranded RNA was precipitated with lithium chloride. A self-annealing anchor primer (PC3-T7 loop; Integrated DNA Technologies) was ligated to the dsRNA in order to obtain full-length sequences, with the exception of RVA/Cow-wt/ZAF/1162/2012/G6P [11]. Complementary DNA was synthesised using the Maxima H Minus Double Stranded kit (ThermoFisher Scientific, Waltham, MA, USA). The manufacturer's instructions were followed with the following modifications. Firstly, the dsRNA was denatured (95 • C; 5 min) immediately before annealing random hexamers and subsequently, first strand synthesis was carried out for two hours at 50 • C.
Sequencing was performed at the University of the Free State Next Generation Sequencing (UFS-NGS) Unit using the Nextera XT DNA Library Preparation Kit (Illumina, Inc., San Diego, CA, USA.) and the MiSeq Reagent Kit V3 (600 cycles).

Maximum Likelihood Phylogenetic Analysis
Sequencing data were assembled as previously described and consensus sequences were analysed using BLASTn [23]. Genotypes were identified using the Rotavirus A Genotype Determination tool available in Virus Pathogen Database and Analysis Resource (ViPR) [24]. Each gene was compared with sequences available in GenBank and nucleotide alignments were constructed using the MUSCLE algorithm in MEGA X [25]. Phylogenetic trees were generated using MEGA X implementing the Maximum Likelihood method and the robustness of branches was assessed by bootstrap analysis using 1000 pseudo replicate runs [25]. The optimal nucleotide substitution model was determined based upon the Akaike information criterion (corrected) (AICc) ranking implemented in jModelTest [26]. For VP7 (G10), VP6 (I2) and NSP4 (E2) the Tamura 3 model + G G4 and for VP7 (G6), VP4 (P [5]), NSP1 (A3) and NSP5 (H3) Tamura 3 + GG4 + I was used. The Tamura-Nei model + GG4 + I was used for VP4 (P [11]) and VP1 (R2). For VP2 (C2), NSP2 (N2) and NSP3 (T6) the General Time Reversible + G G4 and for VP3 (M2) and NSP1 (A11 and A13) General Time Reversible + G G4 + I was used. Lineages for the DS-1-like genotypes [27] and G-and P-genotypes [28] were assigned as previously determined. Nucleotide and amino acid distance matrixes were calculated using the p-distance algorithm in MEGA X.5.