The avian reoviruses (ARVs) are members of the family Reoviridae, the only group of dsRNA viruses (out of seven dsRNA virus families) that infect mammals [1,2]. The ARVs are the prototypic members of syncytia-inducing, non-enveloped viruses within the Orthoreovirus genus. This genus is divided into 3 subgroups: non-syncytia-inducing mammalian reovirus (MRV; subgroup 1; the prototype of the whole genus), avian reovirus and Nelson Bay virus (subgroup 2), and baboon reovirus (subgroup 3) [3]. In contrast to the MRV, which are rarely associated with human pathology [2,4-6], the ARV are significant pathogens of poultry, and cause a variety of diseases, including infectious enteritis in turkeys [7], viral arthritis/tenosynovitis [8], "pale bird" and runting-stunting syndromes [9], and gastroenteritis, hepatitis, myocarditis, and respiratory illness in chickens [2,8,10].

Like MRV, ARV is a non-enveloped virus with 10 linear double-stranded RNA gene segments surrounded by a double concentric icosahedral capsid shell (inner shell [also called core] and outer shell) of 70–80 nm diameter [11,12]. The ARV genomic segments can be resolved into three size classes based on their electrophoretic mobilities, designated L (large), M (medium), and S (small) [11,12]. In total, the genomic composition includes 3 large segments (Ll, L2, L3), 3 medium sized segments (Ml, M2, M3), and 4 small segments (S1, S2, S3, S4). Nine of the segments are monocistronic and encode a single different protein [11-13] while S1 is tricistronic with partially overlapping open reading frames (ORFs) that encode for three proteins [14,15]. Although ARVs share many features with the prototypic MRVs, several notable differences exist including host range, pathogenicity, hemagglutination properties, and syncytium formation [11,12,16-21].

Genomic coding differences also exist between MRV and ARV. For example, although the ARV and MRV S1 genome segments encode homologous receptor-binding proteins [19,22,23], the ARV S1 genome segment encodes two additional ARV-specific gene products, one of which is responsible for ARV's unusual cell-cell fusion ability [14,15,24], whereas the MRV S1 segment encodes only one additional protein [25]. In addition, available data [12,26] suggest each of the homologous orthoreovirus λ-class proteins are encoded by different ARV and MRV L-class genome segments. Differences in the functional properties of homologous ARV and MRV proteins have also been reported. For example, two non-homologous dsRNA-binding proteins (the ARV σA core protein and the MRV σ3 major outer capsid protein) are predicted to regulate PKR activation [27,28] while the ARV σA core protein displays nucleoside triphosphate phosphohydrolase (NTPase) activity [29], ascribed to the non-homologous MRV μ2 [30] and λ1 [31] core proteins. Based on these early comparative studies, it seems likely that additional analysis of ARV will continue to broaden our understanding of the Reoviridae family, possibly leading to the identification of novel features that impact on the distinct biological and pathogenic properties of ARV.

Recent advances have allowed sequence determinations of a growing number of virus isolates. Many ARV and MRV genome segment sequences have been reported. In addition, the complete genomic sequences of three prototype strains of MRV have been completed [32-34]. In contrast, sequence information from ARV isolates is more limited. While the entire complement of S-class genome segments (for example, [14,15,35-39]) and M-class genome segments (for example, [40,41]) have been determined for some ARV clones, and sequence information is available for some ARV L1 and L3 genome segments [42,43], there is, at present, no sequence information for the ARV L2 genome segment. This segment is presumed to encode for the viral RNA-dependent RNA polymerase (RdRp) protein, an essential enzyme for RNA virus replication. Thus, we determined the genomic sequences of the ARV L2 genome segments from two different strains of ARV (ARV138 and ARV176) in order to expand the available ARV sequence database, determine sequences of the ARV RdRp protein, and to delineate conserved structure/function features of this key viral-encoded enzyme.

DNA sequencing was performed in both directions by use of an ABI Prism BigDye Terminator v3.1 Cycle Sequencing Ready Reaction Kit (Applied Biosystems) and an Applied Biosystems Genetic Analyzer DNA Model 3100. The first two sequencing reactions were performed with the primers used for PCR amplification. Primers for subsequent reactions were designed from newly obtained sequences to completely sequence each full-length PCR product in both directions. Sequences nearer the ends of each segment were determined from PCR products that were amplified with a primer complementary to the anchor primer and an internal gene-specific primer. Sequences obtained from both directions were assembled and checked for accuracy with SeqMan^® (Lasergene^®, Version 7.1.0; DNASTAR, Inc.).

The sequences of genes that encode the RdRp protein have been determined for a number of members of the Reoviridae family of viruses (Table 1). However, this information was lacking for members of the avian orthoreovirus subgroup. We determined the sequences of two different strains' ARV L2 genome segments. The L2 genome segments of ARV138 and ARV176 were determined to be 3829 (GeneBank accession no. EU707935) and 3830 (GeneBank accession no. EU707936) nucleotides long, respectively (Table 2). The one-nucleotide length difference is attributed to the 5'-end of the non-translated region of the plus-strand, where ARV138 L2 contains a one-base deletion relative to ARV176 L2. No additional deletions or insertions were found elsewhere in the alignment. The nucleotide identity between ARV138 and ARV176 L2 genome segments is 85% (Table 3). BLAST searches indicated the ARV L2 genome segments were most similar to the mammalian reovirus (MRV) and aquareovirus (AqRV) L1 genome segments, which encode the RNA-dependent RNA polymerase [54,55]. Pairwise sequence comparisons between both of these newly-determined ARV genome segments and all currently available homologous MRV and AqRV L1 genome segments (see Table 1) showed a range of nucleotide and protein identity values. Preliminary comparative studies of all currently available AqRV L-class genome segments indicated that the grass carp reovirus (GCRV) and chum salmon reovirus (CSRV) L genes were the most distantly related amongst the AqRV (data not shown). Thus, although all currently available ARV, MRV, and AqRV L-class genome segments were aligned and compared in subsequent analyses, we limited presentation in subsequent tables and figures to these few most-distant clones for clarity. In addition, preliminary attempts to align the ARV138 and ARV176 L2 genome segments with homologous genes in other Reoviridae genera (ie. the Fijivirus Nilaparvata lugens, the Dinovernavirus Aedes pseudoscutellaris, the Coltivirus Eyach virus, the Orbivirus St. Croix River virus, the Seadornavirus Kadipiro virus, the Mimoreovirus Micromonas pusilla reovirus, and the currently unclassified virus Operophtera brumata reovirus) resulted in much lower identity values and significant gaps (data not shown); thus, these other more-distant genera were not included in subsequent analyses. Pairwise nucleotide sequence comparisons between ARV L2 and homologous MRV genome segments showed identities of ~55%, and pairwise nucleotide sequence comparisons of ARV L2 with AqRV homologues revealed ~48% identity (Table 3).

Table 1 Nucleotide sequences used in this study

Table 2 Genome-segment lengths, non-translated regions, and encoded proteins of ARV138 and ARV176

Table 3 Percent identities of the ARV L2 genome segments and homologous encoded proteins of MRV and Aquareovirusesa

The predicted open reading frames for both ARV L2 segments were determined to be nucleotides 14–3790 for ARV138 L2 and 15–3791 for ARV176 L2, resulting in deduced λB proteins of 1259 residues (Table 2). The calculated molecular weights for ARV138 λB and ARV176 λB are ~140 kDa each (Table 2). The amino acid identity between the two ARV λB proteins is 97.5%, with no insertions or deletions relative to one another. ARV protein λB is the only ARV protein whose sequence has not been reported previously. Thus, completion of the L2 sequence in this study has allowed us to assign its function at the sequence level. Amino acid alignments of ARV λB, MRV λ3, and AqRV VP2 proteins revealed several regions of high amino acid identity (Fig. 1), many of which correspond to previously identified polymerase domains [56]. A large number of amino acids were completely conserved across all 14 currently known ARV, MRV, and AqRV RdRp protein sequences (Fig. 1, closed circles). Amino acid identities between ARV λB and homologous MRV λ3 or AqRV VP2 are ~55% and ~42%, respectively (Table 3), suggesting the ARV and MRV are more closely related to each other than either are to AqRV (also seen in phylogenetic analysis – Fig. 2), reflecting that ARV and MRV belong to different species in the Orthoreovirus genus [36] whereas AqRV are members of the different Aquareovirus genus in the Reoviridae family. Window-averaged analysis of ARV λB and MRV λ3 protein identities (Fig. 3, dashed lines) revealed several regions of high amino acid identity. The highest identity scores, with window-averaged identity values > 90%, were located within canonical polymerase regions, including "fingers" domains (MRV residues 452 – 467 and 514 – 530) "fingers"/"palm" interface domains (MRV residues 542 – 571 and 673 – 699), "palm" domains (MRV residues 725 – 738, which includes the GDD motif, which is common to all viral RNA-dependent RNA polymerases [57-59]), "thumb" domains (MRV residues 864 – 878), and an "undefined" domain (MRV residues 881 – 896). Addition of the AqRV VP2 protein to the above analyses provided additional information about potentially important conserved domains. Clustal-W (Fig. 1) and T-Coffee (data not shown) alignments identified 359 amino acid residues that were identical in the 6 aligned sequences (overall average identity = 28.3% Fig. 3, horizontal solid line]). There were numerous window-averaged regions of very low conservation, with most attributed to AqRV regions that were poorly conserved compared to corresponding ARV/MRV regions, a feature also noted in MRV:AqRV comparisons [60]. Three regions showed higher-than-average conservation in the ARV:MRV:AqRV alignments, with window-averaged identity values > 75%, suggesting these polymerase regions (ARV residues G₅₁₆LRNQVQRRPRTIMP₅₃₀, H₅₄₂TLS/CADYINYHMNLSTTSGSAV₅₆₃, and T₆₇₇TTFPSGSTATSTEHTANNSTM₆₉₈, that correspond to MRV residues G₅₁₆LRNQVQRRPRSIMP₅₃₀, H₅₄₂TLTADYINYHMNLSTTSGSAV₅₆₃, and T₆₇₇TTFPSGSTATSTEHTANNSTM₆₉₈, respectively) contain important structural/functional domains. The GDD motif was located within a region of slightly lower window-averaged scores (~60%), but in a sequence (in ARV) I₇₂₄QxxYVCQGDDG₇₃₅ that, apart from the residues at positions 726 and 727, were completely conserved in all 14 currently-available ARV, MRV, and AqRV RdRp sequences. In addition to the 359 identical residues found in all 6 sequences discussed above, blossum50 weighting alignments indicated that an additional 206 positions contained either identical amino acid residues or conservative substitutions in at least 4 of the 6 aligned sequences.

Figure 1 Alignment of the deduced ARV138 and ARV176 λB amino acid sequences. All 14 currently available homologous ARV λB, MRV λ3, and AqRV VP2 proteins (determined for each clone shown in Table 1) were aligned, both by T-Coffee [50] (data (more ...)

Figure 2 Phylogenetic tree analyses of the prototype ARV L2 genome segments and homologous genes in other reoviruses. Abbreviations are as defined in the legend to Fig. 1. Lines are proportional in length to nucleotide substitution. Alignments were performed by (more ...)

Figure 3 Window-averaged scores for sequence identity among the ARV λB, AqRV VP2, and MRV λ3 RNA-dependent RNA polymerase proteins. To provide consistent weighting to the averaged scores, only the two most-distant clones from each of the three (more ...)

The atomic structure of few ARV proteins have been reported [61], and such high-resolution structures are not known for any λ-class ARV proteins. By contrast, atomic structures are known for most MRV proteins, including the RdRp [52]. Comparative sequence analyses described in this report have indicated that ARV and MRV RdRp proteins share ~55% amino acid identity, ARV and AqRV RdRp proteins share ~42% identity, and that only 359 (~28%) amino acids are completely conserved (identical) when ARV138, ARV176, MRV T1L, MRV T2J, AqRV CSRV, and AqRV GCRV are aligned (Fig. 1). Thus, to gain structure/function information about this key viral-encoded enzyme, ARV, MRV, and AqRV identical amino acids, conservative substitutions, and non-conservative substitutions were modeled in the MRV λ3 crystal structure (Fig. 4). This comparative analysis indicated that most non-conserved amino acids were located on the surfaces of the protein exposed to the core interior and in the N-terminal and C-terminal bracelet domains, whereas most identical amino acids and conservative substitutions were located within canonical fingers, palm, and thumb polymerase motifs, particularly those lining channels used by template and nascent RNA during transcription (Fig. 4). Similar observations had been reported from MRV:AqRV comparisons [60] and our results support and extend these earlier observations. As was previously reported from MRV:AqRV comparisons [60], conserved residues surround the GDD motif and additional residues shown to be important for a variety of polymerase functions are also conserved, including Arg₅₂₂, Arg₅₂₃, Arg₅₂₅, Ala₅₈₇ (which are needed to properly position the incoming NTP triphosphate), Ile₅₂₇ and Pro₅₂₉ (needed to help position template nucleosides), Thr₅₅₇, Ser₅₅₈, Gly₅₅₉, Ser₅₆₀, and Val₅₆₂ (portions of a loop that maintains priming NTP), and Asp₅₈₉, Ser₆₈₁, and Gln₇₃₁ (specifies ribonucleotides). Each of these residues is located one amino acid more N-terminal in the MRV sequence (ie. ARV Arg₅₂₂ = MRV Arg₅₂₃ and all (as well as numerous others) are completely conserved in all 14 currently available ARV, MRV, and AqRV RdRp sequences (Fig. 1, indicated by closed circles). In addition, our comparative analyses indicated many identical amino acids and conservative substitutions were located on the surface of the protein that is predicted to interact with the core shell [62]. This might imply that conserved domains are needed to help tether the RdRp to the underside of the core shell. This hypothesis could be tested by extending such ARV:MRV:AqRV sequence comparisons to the other core proteins.

Figure 4 Localization of conserved, non-conserved, and identical amino acids in ARV, MRV, and AqRV RdRp proteins. The MRV λ3 crystal structure (PDB # 1MUK [52]) was manipulated with Chimera [53]. A, Low-resolution, cutaway model of the reovirus core structure (more ...)

In conclusion, we report the first sequence analysis of the avian reovirus RdRp gene and protein. The ARV λB and MRV λ3 proteins showed the highest ARV:MRV identity values (~55%) amongst currently known ARV and MRV proteins, suggesting significant evolutionary constraints are placed on dsRNA RdRp molecules, particularly in regions comprising the canonical polymerase motifs and residues thought to interact directly with template and nascent mRNA.

The authors declare that they have no competing interests.

WX performed the experiments and analyses and WX and KC wrote the manuscript.

We thank members of our laboratory for critical reviews of this manuscript and Kolawole Opanubi for expert technical assistance. This research was supported by grant FRN-11630 from the Canadian Institutes of Health Research.