Plasma virome of cattle from forest region revealed diverse small circular ssDNA viral genomes

During 2016, 80 bovine blood samples were collected from a forested region in Northeast China. All blood samples were taken from the caudal vein of the animals using disposable syringes and immediately transported to the laboratory on dry ice. These samples were collected as part of a survey for arthropod-borne viral infection in cattle. Library preparation and computational analysis were performed as previously described [3, 19]. Briefly, 80 plasma samples were pooled into five pools which included 16 plasma samples each. Supernatants were then collected after centrifugation (10 min, 13,000×g). Plasma pools were filtered through a 0.45-μm filter (Millipore) to remove eukaryotic- and bacterial cell-sized particles, and 200 μL of each pool was then subjected to a mixture of nuclease enzymes to reduce the concentration of free (non-viral encapsidated) nucleic acids [3]. Remaining total nucleic acid was then isolated using QiaAmp Mini Viral RNA kit (Qiagen) according to manufacturer’s protocol. Viral cDNA synthesis was performed using SuperScript III reverse transcriptase (Invitrogen) according to the manufacture’s instructions. The second strand of cDNA synthesis was performed by incubation of reverse transcribed (RT) products with Klenow Fragment DNA polymerase (New England Biolabs). Five libraries were then constructed using Nextera XT DNA Sample Preparation Kit (Illumina) and sequenced using the MiSeqIllumina platform with 250 bases paired ends with dual barcoding for each pool. Bioinformatics analysis was performed according to a previous study [20, 21]. In bioinformatics analysis, an in-house analysis pipeline running on a 32-nodes Linux cluster was developed to process the data. Cattle host reads and bacterial reads were subtracted by mapping the reads to the reference genome Bos taurus (assembly Bos_taurus_UMD_3.1.1) and bacterial RefSeq genomes release 66 using bowtie2 [22]. The reads whose 5 bp to 55 bp from 5 prime end is identical were considered duplicates and only one random copy of duplicates was kept. Low sequencing quality tails were trimmed using Phred quality score ten as the threshold. Adaptors were trimmed using the default parameters of VecScreen which is NCBI BLASTn with specialized parameters designed for adapter removal. The cleaned reads were de-novo assembled within each barcode by SOAPdenovo2 version r240 using Kmer size 63 with default settings. The assembled contigs, along with singlets were aligned to an in-house viral proteome database using BLASTx with an E-value cutoff of <10⁻⁵. Candidate viral hits are then compared to an in-house non-virus non-redundant (NVNR) protein database, which was compiled using non-viral protein sequences extracted from NCBI nr fasta file, to remove false positive viral hits. Contigs without significant BLASTx similarity to viral proteome database are searched against viral protein families in vFam database [23] using HMMER3 [24‐26] to detect remote viral protein similarities. A web-based graphical user interface was developed to present users with the virus hits, along with taxonomy information and processing meta-information.

To characterize the complete genome of the CRESS-DNA genomes, inverse PCR and Sanger sequencing was used with outward pointing PCR primers designed based on the initial Rep encoding contigs. Using the ORF finder of the Geneious software version 9, putative open reading frames (ORFs) and the stem-loop in the circular genomes were located. PCR screenings were also performed to investigate the prevalence of the five CRESS-DNA genomes in the 80 bovine plasma samples by nested PCR with specific primers designed based on the five genomes, respectively. Sequences and characteristics of the primers used in inverse and screening PCRs are shown in an additional file (see Additional file 1). PCR products were then subjected to Sanger sequencing.

For generating a phylogenetic tree, sequence alignment was performed using CLUSTAL X with the default setting [27], phylogenetic analyses were constructed through full-length Rep protein of novel CRESS-DNA viruses and other genetically-close relatives. Using the Maximum-Likelihood method, phylogenetic trees with 1000 bootstrap resamples of the alignment data sets were generated, visualized using the program MEGA [28]. Bootstrap values (based on 1000 replicates) for each node are shown if >50%.

The 80 cattle plasma samples in the five libraries generated a total of 2,493,791 unique sequence reads using the Illumina MiSeq sequencing run with 250 bases paired ends. Sequence reads were de novo assembled using the Ensemble program [29] and compared to the GenBank non-redundant protein database using BLASTx. Results indicated that CRESS-DNA sequences accounted for a major part of total putative mammalian virus reads, with 1848 reads showing sequence similarity to the viruses from Circoviridae family, and 2212 reads related to viruses from Genomoviridae family [30]. Table 1 list the other mammalian virus sequences detected namely bovine parvovirus (BPV) (2695 reads), papillomavirus (4 reads), and picobirnavirus (3 reads), where the sequence reads of papillomavirus and picobirnavirus were shown in an additional file (see Additional file 2). Besides these mammalian viruses, a small number of insect viral sequences were detected showing low-level sequence similarity to Iflaviridae and Dicistroviridae. The CRESS-DNA viruses showed lower levels similarities to known genomes available in GenBank, whose divergent genomes were then further characterized.

Two complete CRESS-DNA genomes showing the highest sequence similarity to viruses from Genomoviridae family were obtained through inverse PCR based on two large contigs from these libraries and Sanger sequencing. Genomes were 2192 nt (BGmv001, from library 5) and 2926 nt (BGmv002, from library 1) in length, respectively. As shown in Fig. 1a and b, the two genomes contained two bidirectional ORFs, encoding the putative Rep and Cap proteins. The two gemycircularviruses (the only genus currently in the Genomoviridae family) in this study did not contain putative intron in the Rep gene. BLASTp search in GenBank based on the amino acid sequence of Rep showed BGmv001 shared the highest sequence similarity of 56% to a gemycircularvirus genome sequenced from dragonfly abdomens (NC_023872) [8]. BGmv002 shared the highest protein sequence similarity of 36% to a gemycircularvirus from bird feces (KF371630) [11]. Figure 1 C showed the CRESS-DNA genome of Bvch001 with 2717 nt in length, whose Rep and Cap are arranged in inverse direction. The best matches in GenBank of Bvch001 based on amino acid sequence of Rep include two smacoviruses from Gorilla (KP233191 and KP233192) [31], two unidentified circular ssDNA virus from Rhesus macaque (KU043429 and KU043430) [32] and four porcine stool-associated circular viruses (KJ577812-KJ577815) [33], sharing 52% sequence similarity with them. Two complete CRESS-DNA genomes showing the highest similarity to circoviruses were determined through inverse PCR based on two large contigs from these libraries and Sanger sequencing, which were 1881 nt (Bvch002, from library 5) and 2926 nt (Bvch005, from library 4) in length, respectively. Figure 1 D and E indicated the genomic organization of Bvch002 and Bvch005, where the predicted Rep and Cap of Bvch002 are arranged in the same direction within the genome and those of Bvch005 are inversely arranged. BLASTp search in GenBank based on the amino acid sequence of Rep showed Bvch002 shared the highest sequence similarity of 38% to a circovirus-like virus from dragonfly (KM598397) and BGmv005 shared the highest sequence similarity of 48% to a bovine faeces-associated circular DNA virus (NC_030137) [34].

Based on the alignment of the Rep amino acid sequences herein detected with the best matches of BLASTp search in GenBank and those of representative CRESS-DNA genomes including circovirus, cyclovirus and gemycircularvirus, a phylogenetic tree was constructed. As shown in the Fig. 2 BGmv001 and BGmv002 belong to the cluster of gemycircularvirus, Bvch002 and Bvch005 fall into the clade showing close relationship with circovirus, and Bvch001 clustered into the group neighbouring to the gemycircularvirus clade.

PCR screening results indicated Bvch001 was detected in three samples, and sequencing results indicted the three 310 bp sequence fragments were different from each other, sharing >99.3% similarity (the two new sequences were deposited in GenBank with nos. MG601761 and MG601762). BGmv001 was detected in two samples and one of them (with GenBank no. of MG601763) had one base pair difference from the original genome based on the 290 bp PCR fragment. Two samples were positive for BGmv002 and the two sequences had two base pairs difference based on the 280 bp PCR product (the GenBank no. of new sequence is MG601764). The rest two CRESS-DNA genomes, Bvch002 and Bvch005, were respectively detected in a single sample, where their sequencing results were identical to the original genomic sequences, respectively.

Bovine parvovirus 3 (BPV3) was detected in all the five cattle plasma pools and BPV2 was detected in two pools. Three complete coding regions of BPV genomes could be assembled, where one (with stain name ujs2665) belongs to BPV2 (from CattleB03) and two (with stain name ujs1794 and ujs497) belong to BPV3 (from CattleB03 and CattleB04, respectively). Figure 3 showed the genomic organization of the three BPVs and phylogenetic trees based on the nonstructural (NS) and VP1 proteins, respectively. The BPV2’s NS showed 96–98% identities to the 3 BPV2 genomes currently available in GenBank database, while the VP protein showed 84–86% identities to them, suggesting that ujs2665 is a putative new strain in BPV2 group. The NS proteins of the two BPV3 shared 98% sequence identity with each other and both showed 96% sequence identity with the only BPV3 strains in GenBank, and the VP proteins of them shared 99% sequence identity with each other and 97% identity to the BPV strain.

The viral genomes described in detail here and the PCR screening sequence fragments were deposited in GenBank under the following accession numbers MG026727- MG026729, MF669476- MF669480, and MG601761- MG601764. The raw sequence reads from the viral metagenomic libraries were deposited in the Short Read Archive of GenBank database under the accession number SRX3235902.