The proposed new species, cacao red vein virus, and three previously recognized badnavirus species are associated with cacao swollen shoot disease

Sixty-five leaf or shoot samples were collected from symptomatic cacao trees in commercial plantations, or at experimental plots located at the Centre National de Recherche Agronomique in Cote d’Ivoire during 2012. An additional thirty-five samples were collected in 2015 from commercial plantations or the historical collection maintained at the Cocoa Research Institute of Ghana. Each sample was placed in a 50-ml screwcap tube containing 100% glycerol, transported to the University of Arizona, and held at 4 °C. Total DNA was isolated from 100 mg of plant tissue from all samples using the cetyltrimethyl ammonium bromide (CTAB) DNA extraction method of Doyle and Doyle [32]. Purified DNA (2 μl) was enriched for circular DNA using the Templiphi rolling circle amplification (RCA) kit (GE Healthcare Bio-Sciences), according to manufacturer’s instructions, with modifications [33, 34].

Illumina paired-end libraries were constructed from purified DNA or from RCA products using the TruSeq PE cluster kit, to obtain a 350 bp mean insert size. The Illumina HiSeq 2500 sequencing run was carried out at the University of Arizona Genetics Core facility (Tucson, AZ). The reads were de-multiplexed, and quality was assessed using FASTQC (https://​www.​bioinformatics.​babraham.​ac.​uk/​projects/​fastqc/​). The adapters were removed, and reads were trimmed using TRIMMOMATIC v0.32 [35] and assembled de novo using SeqMan NGen software v12 (DNASTAR), with and without plant host sequence filtering. For filtering, the T. cacao nuclear genome (host plant) contigs were downloaded from GenBank Accessions CM001879.1, CM001888.1, FR7222157.1, and KE132922.1 [36, 37]. Also used for filtering were the cacao chloroplast genome sequence, NC014676.2 and cotton Gossypium hirsutum (L.) (cacao relative) mitochondria genome, JX065074.1.

The assembled apparently full-length genome sequences were first annotated using BLAST2GO [38] to confirm their identity as badna-like viruses, and then subjected to BLASTn [39] analysis to determine the top hits (similarity scores) based on all sequences available in the Genbank database. Based on established convention for the genus Badnavirus, the first nt of the predicted tRNA^met priming site was assigned as the first nt coordinate [40, 41] for each full-length CSSD-badnaviral genome sequence. The predicted ORFs were identified for CSSD genome sequences using the NCBI ORF Finder tool, and the conserved protein domains (CPD) were predicted using the NCBI Conserved Domain Database (CDD) [42]. The predicted CPDs were aligned for comparisons using MUSCLE [43], implemented in CLC Sequence Viewer 7.5.

Genome sequences for 122 badnavirus isolates representing 37 species were downloaded from GenBank and the genome was considered intact or based on the RT-RNase H region used for taxonomic assignment to species. A haplotype search implemented in FaBox v1.41 was used to identify and remove sequences sharing 100% nt identity [44], leaving 87 full-length genome and RT-RNase H region sequences for the analyses, respectively. Pairwise distances were determined using Standard Demarcation Tool (SDTv1.2) software [43], for 87 full-length genomes or RT-RNase H regions (1230 bp). The sequence alignments were carried out in MUSCLE [43], implemented in CLC Sequence Viewer 7.5.

Phylogenetic analysis of genomic and RT-RNase H sequences was carried out with the Maximum Likelihood (ML) algorithm available in MEGA6 [45]. Tree reconstructions used the General Time Reversible substitution model, determined by the highest Bayesian Information Criterion score obtained, with gamma distribution for invariable sites, and 1000 bootstrap iterations. Confidence values were placed at major nodes having a ≥ 70% bootstrap.

To validate the de novo-assembled Illumina full-length genome sequences, six representative isolates, three each from Cote d’Ivoire and Ghana, representing Cacao vein virus (CRVV) and the established CSSV species group (four and two sequences, respectively) (Table 1), were selected for PCR amplification, cloning, and bi-directional capillary (Sanger) DNA sequencing. Purified total DNA was enriched for circular DNA by RCA. The partially enriched badnaviral DNA was used as template for PCR amplification of full-length viral genomes with sequence-specific abutting primers (Additional file 1: Table S1). Amplicons were cloned using the CloneAmp™ HiFi PCR Premix (Clonetech), according to the manufacturer’s instructions. The reaction contained 1X CloneAmp™ HiFi PCR Premix, 0.2 μM each of reverse and forward primers, 2 μL of RCA product, and nuclease-free water to a final volume of 50 μL. The PCR conditions were: denaturation at 98 °C for 2 min, followed by 40 cycles of denaturation at 98 °C for 20 s, annealing at 55 °C for 15 s, extension at 72 °C for 8 min, and final extension at 72 °C for 10 min. Amplicons were separated by agarose gel (0.8%) electrophoresis in TAE buffer, pH 8.0. Bands ~7 kbp in size were excised and gel-purified using the illustra GFX PCR and DNA Gel Band Purification kit (GE Healthcare Bio-Sciences), according to manufacturer’s instructions. The concentration of DNA was determined using the Nanodrop 2000 UV–vis spectrophotometer (Thermo Scientific). Amplicons were cloned into the pGEM5 plasmid vector (Promega), previously linearized by digestion with Not I (New England Biolabs). Transformation was carried out using the In-Fusion HD Cloning kit (Clontech), according to the manufacturer’s instructions. Insert size was confirmed following purification of the plasmid vector with the GeneJET Plasmid Miniprep Kit (ThermoFisher Scientific), and by restriction with Not I. The DNA sequence was determined for plasmids bearing a cloned insert of ≥7 kbp in size by bi-directional, capillary Sanger DNA sequencing, and primer walking (Eton Bioscience, San Diego, CA). Sequences were assembled using SeqMan Pro v.12 (DNASTAR, Madison, WI). Viral ORFs, non-coding regions, and predicted, CPDs were identified for comparison to the assembled Illumina sequences using ORF Finder, CDD, and pairwise distance analysis (SDT), as described above.

The Illumina reads were assembled into 14 full-length badnaviral genomes ranging from 6920 to 7172 bp in size. Also, BLASTn analysis against the GenBank database identified a large number of partial contigs with significant matches to CSSD-badnaviral sequences. However, in this report, only complete badnaviral genome sequences were considered, and they are derived from 410,662–31,417,994 reads with a depth of coverage of 36–6600.

PCR amplification was carried out using sequence-specific primers (Additional file 1: Table S1) and Sanger sequencing of six full-length genome sequences, representing two CSSV and three CRVV full-length genomes, and one partial genome assembled from CRVV isolate CI286. Illumina and Sanger sequence comparisons indicated that the cloned viral amplicons were nearly identical in length and sequence, at 99.0–99.9% nt identity, between the respective sequences. BLASTn analysis for the 14 apparently full-length CSSD-badnaviral genome sequences indicated the top hits in the GenBank database were consistent to previously reported CSSD-badnaviral sequences based on similarity score, sequence coverage, and e-value of zero. A comparison of the resultant Next-gen Illumina and Sanger DNA sequences using the above criteria, as well as: location of the first nt coordinate, number and location of predicted ORFs ≥10 kDa, and presence of the hallmark-badnaviral non-coding sequence region, indicated that both approaches yielded the same genome sequence. The Illumina sequences have been deposited in GenBank as the Accession numbers, KX592571-KX592584.

Identification of predicted coding regions (>50 aa) on the CSSD-badnaviral sense strand for the 14 CSSD-genomes revealed three types of genome arrangements, arbitrarily herein referred to as I, II, and III (Fig. 1). Samples CI275, CI286, GH64, and GH67, representing isolates from Cote d’Ivoire and from Ghana, had a type I arrangement, which consisted of four predicted ORFs, 1–3 and Y (Fig. 1a). Seven isolates, CI44, CI134, CI135, CI215, CIS2, CIS3, and CIT5, all from Cote d’Ivoire, had a type II genome arrangement, comprising five predicted ORFs, 1–3, X, and Y (Fig. 1b). And, CI301, CI311, and GH75, which represent isolates from Cote d’Ivoire and Ghana, had a type III arrangement, or six ORFs 1–4, X, and Y (Fig. 1c). The CSSD-associated genome sequences each contained a predicted tRNA^met binding site, located at nt coordinates 1–18, at which reverse transcription is initiated [24, 46].

Predicted CPD searches for the 21 genomes indicated a number of similarities, but also discernable differences, among CSSD-species and well-studied non-cacao-associated badnaviruses [10, 25, 47]. With respect to all of the CSSD-badnavirus species, CPDs were identified on ORFs 1 and 3. The ORF1 consistently harbored a domain of unknown function (DUF) 1319 [42], classified as a member of a badnavirus-specific, uncharacterized protein superfamily (Fig. 2a, Table 1). The ORF3 CPDs were identified as domains of zinc knuckle finger (Zn), pepsin-like aspartate protease (Pepsin), RT, and RNase H proteins (Table 1) present in all genomes except for CSSCDV [28], which lacked the Zn domain. Domains that were shared among all cacao-infecting badnaviruses were DUF1319 in ORF1, and Zn, Pepsin, RT and RNase H domains on ORF3. No CPDs were identified for CSSD-badnaviral species in the ORFs 2, 4, X, or Y, except for CSSCDV ORFY, which harbored a DUF3187 domain.

A survey of CPDs for the two New World cacao-infecting species, Cacao mild mosaic virus (CaMMV) and Cacao yellow vein-banding virus (CYVBV) [48], indicated that certain CPDs are conserved in ORFs 1 and 3 by all six species, while several domains were found only on ORF3 of CaMMV or CYVBV, and absent in CSSD-genomes (Table 1, Fig. 2a). In particular, DUF4200, which is unique to CaMMV ORF3, was located between the Zn and Pepsin domains. In CYVBV ORF3, the plant homeodomain finger (PHD) was identified adjacent to the Zn domain, and the trimeric-dUTPase (Trim) domain was located between the zinc finger and the aspartate protease domains (Table 1, Fig. 2a).

The SDT analysis of the RT-RNase H region identified four species based on ≥80% badnavirus species cutoff established by the International Committee on the Taxonomy of Viruses (ICTV): (i) CRVV, (ii) CSSV, (iii) Cacao swollen shoot CD (Cote d’Ivoire) virus (CSSCDV), and (iv) Cacao swollen shoot Togo A virus (CSSTAV) (Table 2). The best-represented species is CSSV, at 10 isolates, including those reported here and GenBank, from Ghana (AJ608931, AJ609019, and AJ609020) and Togo (AJ534983 and L14546), making it the most widely distributed species thus far. CSSCDV and CSSTAV are reported so far from Cote d’Ivoire (JN606110) and Togo (AJ781003), respectively. And, the herein proposed species, CRVV, represented by isolates from Cote d’Ivoire (CI275, CI286) and Ghana (GH64 and GH67), had the greatest extent of intraspecific divergence among the known CSSD-badnaviruses, at 28–30% (70–72% identity) (Table 3).

By pairwise distance analysis of the RT-RNase H region, the 14 newly determined and seven GenBank CSSD-associated genome sequences shared 72–100% nt identity (Table 2). In contrast, the corresponding RT-RNase H region for the 21 isolates and 80 other badnaviral RT-RNase H sequences (GenBank) was 58–70% identical. Notably, CaMMV and CYVBV are as divergent from each other (40%), at 60% shared nt identity, as they are from West African CSSD-badnaviruses. Although SDT distinguished four species (≥80%) irrespective of RT-RNase H or complete viral genome sequences, species group composition was inconsistent, in particular, for the CSSD isolates. For example, SDT analysis for complete CRVV genomes (≥80% threshold) delimited CI286, and the group containing CI275, GH64, and GH67, as two groups. In contrast, the RT-RNase H analysis identified the same four isolates as one species (Tables 2 and 3). Also, the RT-RNase H analysis placed TG_AJ781003 with CSSTAV species, while by complete genome analysis, the isolate grouped with CSSV species (Tables 2 and 3). Further, the CSSTAV complete genome shared 80% nt identity with Togo isolates L14546 and AJ534983, compared to 75% for the corresponding RT-RNase H region, resulting in a conflicted species status. Finally, only CSSCDV (CI_JN606110) was delimited as the same species by both the RT-RNase H and the full-length genome sequence analysis.

The phylogenetic relationships were determined, based on the RT-RNase H region of 87 GenBank accessions and the 14 CSSD-badnaviral sequences. Despite the accepted use of this region for badnavirus species demarcation, the ML tree was unresolved, e.g. as a polytomy, at ≥70% bootstrap support. Within the badnavirus polytomy, the West African CSSD isolates grouped separately from the other badnaviruses, with 99% bootstrap support, indicating a close phylogenetic relationship among cacao-infecting viruses in West Africa (Fig. 3a). In contrast, ML analysis of the complete genome sequences for the same isolates (CSSD and 87 badnaviral genome sequences from GenBank), resolved three clades, each well supported by ≥70% bootstrap values. For discussion purposes here, the clades have been arbitrarily labeled as I, II and III (Fig. 3b). Clade I (70% bootstrap) contained all known (21) CSSD-associated badnaviral genomes (Fig. 3b). Clade I also contained badnaviruses associated with diverse host species, including one recently reported cacao-infecting virus from Trinidad, CaMMV [48]. A second cacao-infecting badnavirus from Trinidad, CYVBV, grouped with seven non-cacao-infecting badnaviruses in Clade II (Fig. 3b). Clade III contained 18 badnaviruses from non-cacao hosts, including the banana streak virus species group (Fig. 3b).