S. sclerotium strain Sunf-M, which was originally isolated from a sclerotium on a diseased sunflower plant (Helianthus annuus), was a normal wild-type strain in colony morphology and virulence. Agarose gel electrophoresis of dsRNA isolated from mycelial extracts of strain Sunf-M revealed the presence of three dsRNA bands, termed L-, M- and S-dsRNA respectively (Figure 1A). The largest L-dsRNA was generally more abundant and migrated slightly slower than the M-dsRNA.

The L-dsRNA was purified from agarose gel with a gel extraction kit, and then subjected to cDNA synthesis, PCR amplification, cloning and sequencing as described before 3 . Computer-assisted sequence assembly showed that the full-length of L-dsRNA cDNA is 9124 bp in length that lacked a poly (A) tail at its 3′-end. Northern blot hybridization confirmed that the sequence was derived from the L-dsRNA (Figure 1B).

Molecular cloning and sequencing of the smallest S-dsRNA was also carried out and the complete nucleotide sequence was determined. Sequencing and Northern hybridization analysis revealed that the S-dsRNA band was actually a doublet consisting of two co-migrating dsRNA segments and thereby the resulting S-dsRNA band was designated as S-1 and S-2 dsRNA, which are 1856 bp and 1783 bp in length, excluding the poly (A) tail, respectively (Figure 1B).

The genomic organization of L-dsRNA revealed that it contains two large open reading frames (ORFs): ORF1 (nt 1089–5006) and ORF2 (nt 5054–9071) in different frames on the genomic plus strand. The L-dsRNA has a long 5′-UTR of 1088 bp, but relatively short 3-UTR of 48 bp (Figure 1C). ORF1 potentially encodes a 1305 amino acid (aa) protein with a predicted molecular mass of 144 kDa. A sequence search with BLASTp showed that it shares significant sequence similarity (E value of ≤ 3e-20) with only the hypothetical proteins of three unclassified dsRNA viruses: Grapevine associated totivirus-2 (GrAV-2) 22 , Fusarium graminearum dsRNA mycovirus-3 (FgV-3) 23 and Phlebiopsis gigantea mycovirus dsRNA 2 (PgRV-2) 24 in the databases. Although a search of conserved domain database (CDD) revealed a significant match (E value of 8.09e-03) with a partial consensus sequence of the phosphoheptose isomerase (SIS_GmhA; cd05006) (Figure 1C), the function of ORF1 protein is unclear.

ORF2 potentially encodes a 1338-aa protein with a predicted molecular mass of 146.2 kDa. BLASTp searches showed that this protein was most closely related (E value of ≤ 9e-92) to the putative RdRps of the same three viruses mentioned above as well as another unclassified dsRNA virus Diplodia scrobiculata RNA virus 1 (DsRV-1; accession no. NC_013699). In addition, this protein also shares significant sequence similarity with RdRps of members of the families Totiviridae and Chrysoviridae. CDD searches and multiple protein alignment confirmed that ORF2-encoded protein contained a conserved viral RdRp domain (RdRp_4; pfam02123) with eight conserved motifs characteristic of the RdRps in dsRNA viruses (Figure 2).

Intriguingly, a search of CDD revealed that a region of L-dsRNA ORF2 downstream of the RdRp domain shared significant sequence similarity (E value of 5.96e-03) with a consensus sequence of the Phytoreo_S7 domain (pfam07236, aa 266–349) (Figure 1C). The Phytoreo_S7 domain is characteristic of a family consisting of several phytoreovirus S7 proteins known to be viral core proteins with nucleic acid binding activities. The phytoreovirus S7 domain has also been found in FgV-3 23 , chrysovirus Penicillium chrysogenum virus (PcV) 25 as well as other viruses (see below).

The ORF2 is likely expressed via a −1 frameshift mechanism as a fusion protein with ORF1-encoded protein. A heptanucleotide sequence (GAAAAAC_4997-5003), located in a stretch in-frame with ORF1 stop codon region (nt 4987–5054) and upstream of ORF2 start codon, was identified as a putative shifty heptamer motif that could facilitate ribosomal frameshifting (Figure 1C). In addition, a pseudoknot structure that facilitates pausing of the translating ribosome and increasing the frequency of frameshifting 26 27 was also found immediately downstream of the shifty heptamer (nt 5027–5114) (Figure 1D). Similar genomic organization and expression strategy are characteristic of some members of the family Totiviridae. Therefore, the results presented in this section suggest that L-dsRNA probably represented a novel dsRNA mycovirus infecting S. sclerotiorum strain Sunf-M. This virus was hence named as Sclerotinia sclerotiorum nonsegmented virus L (SsNsV-L).

Both S-1 (1856 bp) and S-2 dsRNA (1783 bp) contain a single ORF on their plus-stranded RNA and share conserved sequences in their 5′-UTRs (Figure 1E and Additional file 1: Figure S1). The S-1 and S-2 dsRNAs potentially encode proteins of 579 and 507 aa, respectively. BLASTp searches showed that the S-1 protein had highest sequence similarity (identity of 47%) to the putative RdRp of Flammulina velutipes isometric virus (BAH08700.1) and the S-2 protein shared highest sequence similarity (identity of 25%) with the putative CP of Rosellinia necatrix partitivirus 2 (BAK53192.1). Both of these two viruses are members of family Partitiviridae. Furthermore, CDD database searches and multiple protein alignment revealed that the S-1 protein contained the consensus motifs of partitivirus RdRps (RNA_dep_RNAP, cd01699) (Figure 2). These results suggest that S-dsRNA is a new member of family Partitiviridae infecting S. sclerotiorum Sunf-M and was named as Sclerotinia sclerotiorum partitivirus S (SsPV-S).

To elucidate the evolution of mycovirus-related dsRNA viruses including SsNsV-L and SsPV-S, we compared the genome structures and performed phylogenetic analyses for representative members of the families Chrysoviridae, Partitiviridae and Totiviridae as well as other unclassified dsRNA viruses (Figures 3 and 4, Additional file 2: Table S1, and Additional file 1: Figure S2). The analyses revealed that there are diverse evolutionary lineages in the known mycovirus-related dsRNA viruses including at least ten monopartite, three bipartite, one tripartite and three quadripartite lineages.

The monopartite dsRNA family Totiviridae is not a monophyletic group since it comprises five distant evolutionary lineages: GLV-like, IMNV-like, victorivirus-like, ScV L-like, and UmV-H1 (Figure 3 and Additional file 1: Figure S2). In addition, there were variations in genome structures for several members compared with typical members (e.g. ScV-L-A) of the family Totiviridae (Figure 4). For example, the genome of Piscine myocarditis virus AL V-708 (PMV-AL V-708) contains an extra ORF at its 3′-terminal other than the CP and RdRp ORFs 28 (Figure 4); this is not known to occur in other totiviruses. In addition, contrary to typical totiviruses, the RdRp ORF of Glomus sp. RF1 medium virus (GMRV-RF1) is located at the 5′ terminal half of the genome upstream of CP ORF (Figure 4).

In addition to the five totiviral lineages, the unclassified monopartite dsRNA viruses consist of five other evolutionary lineages: Amalgamaviridae, CiTV-1-like, PgRV-1-like, SsNsV-L-like, and NrV-L1, which are genetically distantly related to each other. The Southern tomato virus (STV) and three other related viruses isolated from plants 6 7 8 29 clustered together and were distantly related to totiviruses. Their genomes are shorter than those of totiviruses and no typical virions are associated with them. A new family, Amalgamaviridae, has been proposed to accommodate these four viruses 6 . Our phylogenetic analysis revealed that a yeast virus Zygosaccharomyces bailii virus Z (ZbV-Z) was also likely to be the member of this proposed family. The SsNsV-L-like lineage contained SsNsV-L and four other mycoviruses (Figure 3 and Additional file 1: Figure S2). They have similar genome structures with typical members of the family Totiviridae but their genomes (>9 kb) are much larger than those of known totiviruses (Figure 4). It has been reported that the PgRV-2 and FgV-3 possibly do not form true virions 23 24 . It remains to be determined if SsNSV-L has a capsid. The CiTV-1-like lineage included two insect viruses and an unclassified plant virus. Despite the similarity in genome organization with members of the family Totiviridae, their genomes (~8 kb) were slightly larger than those of typical totiviruses (4.6–7.0 kb) (Figure 4) and might not assemble in conventional virions 30 . The P. gigantean mycovirus dsRNA 1 (PgRV-1)-like lineage consists of three large monopartite dsRNA viruses and no viral particles have been reported in association with infections by PgRV-1 24 . A mycovirus Nectria radicicola virus L1 (NrV-L1) is most closely related to partitiviruses and has been assigned as an unclassified member of the family Partitiviridae in database. However, its genome is 6 kb in length 31 , clearly suggesting that it is not likely to be a partitivirus. Hence, it may represent a novel group of monopartite dsRNA viruses.

The family Partitiviridae generally consists of viruses with bipartite genomes comprising two dsRNA segments. All members of this family clustered together in the phylogenetic tree and could be further separated into four clades: clade I-IV (Figure 3 and Additional file 1: Figure S3). The recently isolated mycovirus Rosellinia necatrix megabirnavirus 1 (RnMBV1), a member of the proposed family Megabirnaviridae 9 was most closely related to the PgRV-1-like monopartite lineage (Figure 3), but it has a bipartite genome encapsidated in isometric virions. Therefore, it may represent a novel bipartite evolutionary lineage of dsRNA viruses. More recently, a novel bipartite dsRNA mycovirus Botrytis porri RNA virus 1 (BpRV1) was isolated from Botrytis porri that belongs to a separate clade distinct from those of other known RNA mycoviruses 32 . Interestingly, the BpRV1-related bipartite dsRNA virus was also isolated from S. sclerotiorum in our lab (Liu LJ et al., unpublished data). These results indicated that there are diverse evolutionary lineages of bipartite dsRNA viruses in nature.

Although the typical patitiviruses have two genome segments, some viruses in clade IV harbor three segments, two of which encoded CPs. These CPs formed two sister groups in the CP tree and the two CPs in each virus clustered within different groups (see Additional file 1: Figure S3B), suggesting that a possible duplication event occurred before the divergence of these viruses. Because the genomes of many viruses in the clade IV were not completely sequenced, it is not yet known whether all members of this clade possess two CP genes.

Currently, ICTV recognizes six viruses as members or probable members in the family Chrysoviridae 5 (Figure 3). Phylogenetic analysis revealed that other related viruses might also be members of this family (Figure 3 and Additional file 1: Figure S2). The extended family is a monophyletic group and can be further divided into two subgroups: clade I and clade II. Members of clade II have only 4 genome segments whereas members of clade I generally have more than four segments. At present, none of clade I viruses is recognized as member or probable member of the genus Chrysovirus. It is worth noting that Aspergillus mycovirus 1816 (AMV1816) has been incorrectly assigned by the GenBank as an unclassified member of the family Totiviridae in database. Four dsRNA segments have been shown to be associated with this virus 33 . Our phylogenetic analysis also suggests that it is probably a member of Chrysoviridae.

Like members of the family Chrysoviridae Alternaria alternate dsRNA mycovirus (AaRV) 34 and Aspergillus mycovirus 341 (AMV-341) have four genomic dsRNA segments that are encapsidated in isometric virions. However, each genomic segment has a 3′-poly (A) tail, which is not found in chrysoviruses. In addition, the dsRNA 4 segments of these two viruses (~1.5 kb) are much shorter than those of chrysoviruses (~2.4 kb) (Figure 4). Furthermore, phylogenetic analysis revealed that these two viruses are distantly related to chrysoviruses (Figure 3). These results suggest that the two viruses are members of a novel lineage of quadripartite dsRNA viruses.

Rosellinia necatrix quadrivirus 1 (RnQV1), a novel dsRNA virus, has recently been reported from Rosellinia necatrix 35 . Its genome encompasses four segments but all of which are much larger than those of chrysoviruses (Figure 4). Interestingly, the genome of Amasya cherry disease-associated mycovirus (ACDAV) also encompasses four segments but two of which putatively encode RdRps that are related to each other 36 and which cluster with RnQV1 in the phylogenetic tree (Figure 3). It has been suggested that the two RdRps of ACDAV might derive from two related viruses 35 . Although RnQV1 and ACDAV clustered within the totiviral clade (Figure 3), the feature of multipartite genome indicated that they represented a new lineage of quadripartite dsRNA viruses. Based on the properties of RnQV1, a new family Quadriviridae has been proposed 35 .

From both phylogenetic trees and network, we can delineate different lineages of multipartite dsRNA viruses: AaRV-like, RnQV1-like, RnMBV1 and chrysoviruses. These are generally most closely related to certain monopartite dsRNA viruses and the monopartite viruses were branching deeply at the base of the multipartite viral lineages. These results suggest that multipartite dsRNA viruses likely evolved from different monopartite dsRNA viruses. The extra segments in monopartite dsRNA viruses could be generated through gene acquisition (e.g. RnMBV1 and chrysoviruses) and/or genome separation (e.g. RnQV1 and AaRV). In addition, phylogenetic analysis places the NrV-L1 at the base of partitiviral clade III (see Additional file 1: Figure S3B), raising the possibility that the ancestor of clade III partitiviruses possibly originated from NrV-L1-like monopartite viruses. This finding raised our awareness of the possibility that the partitiviruses in the three other clades probably also evolved from different monopartite dsRNA viruses that have yet to be discovered. Therefore, generation of multipartite genomes may be an important macroevolutionary mechanism in dsRNA viruses.

In this study, genome comparisons and phylogenetic analyses revealed that at least 10 monopartite, 3 bipartite, 1 tripartite and 3 quadripartite lineages of mycovirus-related dsRNA viruses are known. Among these, some lineages have been considered as members or tentative members of the families, Totiviridae, Partitiviridae and Chrysoviridae. However, the taxonomy of viruses belonging to the other evolutionary lineages has yet to be considered. Because of differences in genome organization, particle morphology and phylogeny from members of these three families, the establishment of new virus families or new genera is warranted to accommodate the unclassified dsRNA viral lineages.

Current taxonomy of genera in the family Partitiviridae is based on viral hosts (plants or fungi). Our phylogenetic analysis, however, shown that partitiviral clade I and II consisted of a mosaic of plant partitiviruses (of genus Alphacryptovirus) and fungal partitiviruses (of genus Partitivirus). Therefore, the host taxon is not a distinguishing character and it does not reflect the true evolutionary relationships of viruses. Because classification of viruses based on phylogeny not only helps to understand the evolution of viruses but also facilitates the prediction of new virus emergence. Therefore, we propose the establishment of four new genera to reflect the four clades in family Partitiviridae.

It has been considered that dsRNA viruses are polyphyletic and have originated from different lineages of positive-strand RNA viruses 37 38 39 . The families Partitiviridae and Totiviridae have been suggested to belong to the picorna-like superfamily 40 . Our study revealed that members of the family Chrysoviridae and other diverse evolutionary lineages of mycovirus-related dsRNA viruses are closely related to totiviruses and partitiviruses. Therefore, these viral lineages may also be included in this expanded superfamily, remarkably expanding the known diversity of the picorna-like viruses.

The S7 protein domain is thought to be characteristic of members of the genus phytoreovirus in family Reoviridae. The finding that S7 domain homologs occur in SsNsV-L, FgV-3 and PcV raised our interest in exploring whether other viruses also have S7 domain homologs. For this purpose, we performed PSI-BLAST searches using the S7 domain sequences of these three viruses and plant phytoreoviruses as seed sequences against NCBI nr database. The results showed that homologs of S7 domain are widely distributed in diverse dsRNA viral lineages (Table 1).

^aThe positions of Phytoreovirus S7 domain in proteins for alignment and phylogenetic analysis were indicated.

The S7 domain was found in all known chrysoviruses. Interestingly, It occurs in both RdRp and p3 proteins of Clade II chrysoviruses (p3 homolog in several members is p4), but only found in the RdRps of Clade I chrysoviruses. In fact, the 5′-terminal regions of RdRps in Clade II chrysoviruses are homologous with those of their p3/p4 proteins 25 (Figure 5). The S7 domain is located in these 5′-terminal homologous regions. The homology between RdRp and other proteins was not found in Clade I chrysoviruses.

In addition to four members in genus phytoreovirus, we also found that the S2 protein (guanylyltransferase) of Scylla serrata reovirus SZ-2007 (SsReV-SZ) 41 , a member of a proposed new genus in the family Reoviridae, contains an S7 domain homolog at its 3′-terminal. This domain was also identified in SsNsV-L-related mycoviruses FgV-3, DsRV-1 and PgRV-2 as well as in another unclassified monopartite dsRNA virus, PgRV-1. Moreover, S7 domain homologs were also found in the CP protein of the totivirus GMRV-RF1 and the polyproteins of three endornaviruses.

Compared with those of phytoreoviruses, the S7 domain in other viruses contains only partial sequence of the S7 domain. Multiple alignment of the S7 domain revealed that the sequences were likely to be conserved in diverse dsRNA viral lineages (Figure 6). The function of this domain in these non-phytoreoviruses is not known. It will be of interest to determine whether it plays a role in viral RNA binding and packaging.

Comparison of the domain arrangement in relevant viral proteins shows that the S7 domain is located downstream of the RdRp domain in SsNsV-L, FgV-3, DsRV-1 and PgRV-2. However, the S7 domain resides upstream of the RdRp domain in endornaviruses, chrysoviruses and PgRV-1 (Figure 5). In addition, this domain is also found in other proteins, such as the CP of GMRV-RF1 and the guanylyltransferase of SsReV-SZ. These results suggest that multiple recombination events may have occurred among different viral domains or proteins.

To elucidate the possible evolution of the S7 domain homologs, we constructed phylogenetic tree and network for those from different viral lineages (Figures 7 and 8). The S7 domains from five members of phytoreovirus clustered together and formed a distinct clade in both tree and neighbor-net. All of these from chrysoviruses also clustered together. The phylogeny relationships of the p3/p4 of clade II chrysoviruses were consistent with those of their RdRps, suggesting that the ancestor of clade II chrysoviruses also possessed the S7 domain in its RdRp and p3/p4. Considering that the branches of p3/p4 proteins were located at the base of chrysoviruses clade and that the 5′-terminal regions were homologous between RdRps and p3/p4s (Figure 5), it is most likely that the S7 domains were firstly obtained by the p3/p4 proteins of the ancestral clade II chrysoviruses and then transferred to the RdRp proteins by recombination during evolution. Given the close relationships in the neighbor-net (Figure 8), the S7 domains of clade I chrysoviruses may have been acquired by their progenitor from those in RdRps of the ancestor of clade I chrysoviruses via inter-species HGTs.

Some viruses from unrelated families constituted the third clade. The positions of their S7 domains were different (Figure 5). This suggests that HGTs may have occurred among these distant viruses. The closely related four unclassified mycoviruses SsNsV-L, FgV-3, DsRV-1 and PgRV-2 cluster together and the arrangement of their S7 and RdRp domains were also similar, suggesting that the ancestor of these four viruses may have contained the S7 domain. Interestingly, PgRV-1 and PgRV-2 that coinfect one fungal strain clustered together but the positions of their S7 domains were different, suggesting that HGT between these two viruses may have occurred. Similarly, the endornavirus Helicobasidium mompa endornavirus 1 (HmEV-1) 35 possibly acquired its S7 domain from the DsRV-1-like virus via HGT.

dsRNA viruses represent an extremely divergent group. Although RdRps are the most highly conserved proteins among RNA viruses, there is little primary-sequence similarity among those encoded by dsRNA viruses from different genera, even those belonging to the same family 1 . For example, the RdRps from several distinct genera in the family Reoviridae show only 10–20% amino acid identity 42 . So far, the phylogenetic relationship of members of Reoviridae with viruses in other families or genera is not clear. Therefore, the obvious conservation of S7 domain in diverse dsRNA viral lineages even from different families implies that it is not likely to be the results of vertical inheritance. Furthermore, the S7 domain only occurred in certain viruses of a given viral group; their phylogenies were not consistent with those of RdRps. For example, only three members of the family Endornaviridae have the S7 domains, and these did not cluster together in S7 domain tree (Figure 6). In addition, the S7 domains in the non-phytoreoviruses are truncated and have different domain arrangement with RdRps (Figure 5). Given these facts, the S7 domain sequences in non-phytoreoviruses were most likely acquired from ancestral phytoreoviruses and then horizontally transferred repeatedly among these diverse dsRNA viral lineages. This suggests that HGT between different virus species is an important macroevolutionary mechanism in dsRNA viruses.

Wild type S. sclerotiorum isolate Sunf-M was obtained from sunflower (Helianthus annuus) in Hohhot, Inner Mongolia, China. Isolate was routinely cultured on PDA at 20°C and stored at 4°C in PDA tube slants. Small mycelial agar plugs were grown out on cellophane membranes on top of PDA at 20°C for 2 day, and then the mycelium was harvested from the cellophane membranes for dsRNA extraction.

A previously described procedure for dsRNA extraction 43 was used with minor modifications. Briefly, the mycelium was ground in liquid nitrogen with mortar and pestle. dsRNA was extracted from the mycelium and purified to remove traces of DNA and ssRNA by digestion with S1 nuclease and DNase I (TaKaRa). The dsRNA preparations were fractionated on 1.0% agarose gel and stained with ethidium bromide. The isolated dsRNA was agarose gel purified with an AxyPrep™ DNA Gel Extraction Kit (Axygen).

cDNA synthesis and molecular cloning of the L-dsRNA was performed as previously described 3 . A single-primer amplification method 44 with minor modifications was used to obtain the full-length clones of S-dsRNA. In brief, a 3′ amino-blocked ligase adaptor (5′-pGCATTGCATCATGATCGATCGAATTCTTTAGTGAGGGTTAATTGCC-NH₂-3′) was ligated to the 3′-end of the purified S-dsRNA segment using T4 RNA ligase (Fermentas) and the adaptor-ligated dsRNA was then reverse transcribed with a complementary primer 1 (5′-GGCAATTAACCCTCACTAAAG-3′). Following treatment with RNase H (TAKARA), the cDNA strands were annealed for 10 min at 80°C, for 16 h at 65°C and for 3 h at 30°C and the resulting hybrid was filled in with Klenow Fragment (TAKARA). The cDNA was amplified with another complementary primer 2 (5′-TCACTAAAGAATTCGATCGATC-3′). The resulting products were recovered and purified with a gel extraction kit (Axygen), cloned into the pMD18-T vector (TaKaRa). DNA was sequenced by Sanger methods at the Beijing Genomics Institute (BGI). The full genomic sequences of L- and S-dsRNA were completed and confirmed by overlapping of cDNA clones and each base was determined by sequencing at least two independent clones from both orientations. New sequences generated in this study were deposited in GenBank under accession numbers: JQ513382, GQ280377.1 and GQ280378.1.

Northern hybridization analysis was performed as previously described 25 . To verify the authenticity of the cDNA clones generated with the purified dsRNA, the cDNA clones from L-dsRNA, S-1 and S-2 dsRNA segments were labeled with ³²P] dCTP using a radio-labeling kit (TaKaRa) and used to probe the different RNA blots, respectively.

The genome and protein sequences of the dsRNA viruses used in this study were downloaded from viral genome databases at the NCBI website (http://www.ncbi.nlm.nih.gov/genomes/GenomesHome.cgi?taxid10239). The software package DNAMAN 7 (Lynnon Biosoft, USA) was used for sequence annotations, including nucleotide statistics and ORF searching. Similarity searches of NCBI GenBank database were conducted using the online BLAST program (http://blast.ncbi.nlm.nih.gov/Blast.cgi). Searches for protein domains were performed using NCBI conserved domain database (CDD) (http://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi). The pseudoknot structure of L-dsRNA was predicted by the KnotSeeker program 45 and was visualized using program VARNAv3-7 46 .

Multiple alignments of protein sequences were constructed using the M-Coffee web server (http://tcf_dev.vital-it.ch/apps/tcoffee/play?name=mcoffee), which combines the output of several popular aligners into one single multiple sequence alignments. Phylogenetic trees were estimated using two independent methods: Maximum-Likelihood (ML) and Bayesian inference (BI) on the aligned amino acid sequences. BI trees were constructed with MrBayes-3.1.2 47 ,using WAG models of amino acid substitution with invgamma (+I + Γ), performing two runs each of four Monte Carlo Markov Chains (MCMCs), sampling every 1000th iteration over 1.1 × 10⁶ generations after a burn-in of 101 samples. ML trees were inferred with PhyML 3.0 48 , using the best-fit model selected by ProtTest2.4 49 for each dataset, with SPRs algorithms and 8 categories of γ-distributed substitution rates. The reliability of internal branches was evaluated based on SH-aLRT supports. The resulting BI and ML trees were then used to construct the consensus tree for each alignment using TREE-PUZZLE5.2 50 .

Considering that conflicting phylogenetic signals can lead to tree reconstruction artifacts, we also constructed phylogenetic networks with SplitsTree4 program (http://www.splitstree.org/), using Neighbor-Net method and WAG model of substitution. The unresolved branches or conflicting phylogenetic signals in alignments could be indicated by the box-like structures in neighbor-nets graphs.