A sequence of 300 bases showing sequence similarity to bocaviruses (BLASTx E
) was initially identified by 454 pyrosequencing. The sequence was extended by degenerate PCR targeting conserved bocavirus regions, yielding a ~2.5 kb partial genome sequence. The tissue nucleic acids were further analyzed using the MiSeq Illumina platform generating 16 contigs composed of 133 reads with similarities to bocaviruses (BLASTx E
), which allowed the amplification of all three ORFs. The virus was highly divergent from the other two known canine bocaviruses, MVC and CBoV, and was provisionally named Canine bocavirus 3 (CnBoV3)
The nonstructural (NS1) protein encoded by ORF1 was 778 aa long, and contained motifs associated with rolling circle replication, helicase and ATPase. The NP1 protein encoded by the middle ORF3 was 194 aa long. The ORF2 encoded capsid proteins VP1 (689 aa) and proteolytically processed VP2 (560 aa). CnBoV3 shared 51%, 57%, 56% aa similarity with the NS1, NP1 and VP1 region of CBoV, respectively, and 49%, 52%, 57% aa similarity with the NS1, NP1 and VP1 regions of MVC. Phylogenetic analysis of the entire VP1 was performed to determine the relationship between CnBoV3 and other bocaviruses. CnBoV3 was phylogenetically distinct from the known dog bocaviruses and only loosely related to California sea lion bocaviruses, CBoV, feline bocavirus and MVC (Figure
). Phylogenetic trees constructed by NS1 and NP1 region yielded similar topology (data not shown).
Phylogenetic tree based on aligned amino acid sequences of full-length VP proteins of representative bocavirus species.
Bocaviruses are believed to replicate through the parvovirus rolling hairpin model, which generate replication intermediates of concatemers with head-to-head or tail-to-tail structure
]. Recent experimental evidence showed the presence of head-to-tail concatemers or circularized genomes of human bocavirus (HBoV) 1&3
] and porcine bocaviruses
], indicating that some bocaviruses may use a rolling - circle replication model.
Using inverse PCR with primers directed outward from the 5’ and 3’ extremities of the partial genome sequence we were able to amplify head-to-tail sequence of the CnBoV3 non-translated regions (NTR) between the VP termination and the NS1 initiation codons (Figure
A). The generation of specific PCR products indicated the presence of concatemerized or circular forms of the genome in the liver
Figure 2 A) Inverse PCR was performed to obtain the NTR sequence between the VP termination and the NS initiation codon of episomal forms of CnBoV3. Primers (P1 and P2) were situated at VP and NS region respectively. Secondary structures (symbolic in the box) (more ...)
To confirm these results, multiple repeats of the inverse PCR were generated and directly sequenced as well as subcloned into a plasmid vector. PCR amplicons and plasmid inserts were Sanger sequenced using protocol for GC-rich/hairpin sequences. The resulting sequences were aligned and the length of the region between the VP stop and the NS start codon calculated. This region varied in length from 392 to 506 nt (Figure
C). Mfold analyses of the longest NTR sequence (506 nt) showed that the long palindromic hairpin terminal repeats (e.g. approximately 150 nt at both ends in MVC or bovine parvovirus, BPV) were missing, but a typical “rabbit ear” structure was detected that was nearly identical to a structure in the MVC 5'NTR sequence (Figure
B). These results suggested that inverse PCR may have been impeded by strong secondary DNA structures (Figure
A &C, between two red arrows). None of the currently reported NTR regions of bocavirus obtained by inverse PCR contained both complete inverted terminal repeats seen in MVC or BPV