Conventional satellites are subviral agents which lack genes that encode functions needed for replication, depending instead for their multiplication on the co-infection of a host cell with a helper virus (Briddon et al., 2012). Two major classes of satellites may be distinguished, satellite viruses that encode a structural protein that encapsidates their genome, and satellite nucleic acids that encode either non-structural proteins, or are non-coding, and are encapsidated by the coat protein (CP) of helper viruses. The genetic material of satellites is distinct from that of the genome of their helper viruses. Replication of the satellites interferes with the replication of the helper virus and may affect disease symptoms, ranging from attenuation to exacerbation depending on the satellite, the helper virus and the host plant (Hull, 2002; Simon et al., 2004).

Viruses of the genus Begomovirus (family Geminiviridae) have circular single-stranded DNA genomes composed of one or two components each of ∼2.7 kb. They are transmitted exclusively by the whitefly Bemisia tabaci (Hemiptera: Aleyrodidae) and cause important diseases of dicotyledonous crops worldwide (Navas-Castillo et al., 2011). The sweepoviruses constitute a monophyletic group of begomoviruses which have been identified infecting sweet potato (Ipomoea batatas) and other species of the family Convolvulaceae. Typical of monopartite begomoviruses originating from the Old World (OW), the genomes of sweepoviruses encode two genes (the CP and V2) in the virion-sense and four [the replication associated protein (Rep), the transcriptional activator protein (TrAP), the replication-enhancer protein (REn) and the C4 protein] in the complementary-sense. Virion- and complementary-sense genes diverge from an intergenic region (IR) that contains a predicted stem–loop structure with the conserved (between most geminiviruses) nonanucleotide sequence TAATATTAC forming part of the loop. The IR also contains repeated short sequence motifs (known as iterons) surrounding the TATA box of the Rep promoter that are Rep binding sites and, together with the stem–loop structure, form the origin of virion-sense DNA replication (ori). The sweepoviruses cluster as a branch in the begomovirus phylogenetic tree below the divergence between the OW and New World (NW) branches, corresponding to the major clades in the genus (Lozano et al., 2009).

Two classes of circular single-stranded DNA molecules have been described in association with geminiviruses, the alphasatellites and the betasatellites (Briddon and Stanley, 2006; Zhou, 2013). The alphasatellites (previously known as DNA-1; Briddon et al., 2004) are not strict satellites, since they are capable of autonomous-replication in plant cells, but are dependent on their helper begomoviruses for movement within plants and insect transmission between plants (Mansoor et al., 1999; Saunders and Stanley, 1999). Alphasatellites are widespread in the OW and usually occur in association with monopartite begomoviruses and betasatellites. Alphasatellites have also been identified in the NW in association with bipartite begomoviruses and in the absence of betasatellites (Paprotka et al., 2010b; Romay et al., 2010). Recently an alphasatellite and a betasatellite were shown to be associated with a mastrevirus (genus Mastrevirus, family Geminiviridae) in wheat, the first time either of these components has been identified in a monocot (Kumar et al., 2014).

Betasatellites (previously known as DNA-β) have so far only been identified in the OW and are usually, but not exclusively, associated with monopartite begomoviruses (Saunders et al., 2000; Briddon et al., 2001). They are also half the size of begomovirus components (∼1360 nt) and contain three conserved regions: an A-rich sequence, a sequence conserved between all betasatellites known as the satellite conserved region (SCR) and a single gene in the complementary-sense that encodes a small protein (∼118 amino acids) known as βC1. Betasatellites increase the accumulation of their helper begomoviruses and enhance the symptoms induced in some host plants (Briddon et al., 2003; Zhou et al., 2003), likely due to the suppressor of RNA interference activity of the βC1 protein (Cui et al., 2005; Yang et al., 2011). The only sequence similarity between betasatellites and their helper begomoviruses is the nonanucleotide sequence TAATATTAC within the stem–loop structure forming part of the SCR. Betasatellites and alphasatellites are promiscuous with respect to helper virus in that they may be maintained in plants by several, but possibly not all, geminiviruses (Saunders et al., 2002a,b, 2008; Briddon et al., 2003).

The first DNA satellite was identified in association with a plant virus in tomato plants infected with the OW monopartite begomovirus tomato leaf curl virus (ToLCV) originating from Australia (Dry et al., 1997). ToLCV-sat is 682 nt, about one quarter the size of a begomovirus genome/genomic component, is non-coding and has no significance sequence similarity with the helper virus. The satellite contains a predicted stem–loop structure containing the geminivirus-like nonanucleotide sequence TAATATTAC, a sequence with similarity to the SCR of betasatellites, an A-rich sequence and a second predicted stem–loop structure that contains a sequence identical to the predicted iteron sequence of ToLCV within the loop. In common with betasatellites, ToLCV-sat depends upon a helper virus for replication and, based upon sequence and structural similarities with betasatellites, it is believed that ToLCV-sat originated as a defective betasatellite (Saunders et al., 2000).

A novel class of DNA satellites has recently been identified in association with begomoviruses infecting malvaceous plants in Cuba (Fiallo-Olivé et al., 2012). These satellites are approximately one quarter the size of a begomovirus genome/genomic component and have all the features of ToLCV-sat; containing a stem–loop structure with the nonanucleotide TAATATTAC forming part of the loop, an A-rich region and are non-coding. They also contain a putative second predicted stem–loop structure situated close to begomovirus iteron-like sequences and a TATA motif, and a short region with sequence identity to the SCR of betasatellites. Similar satellites have also been identified using vector-enabled metagenomics (VEM) in B. tabaci adults collected in Florida (Ng et al., 2011).

Although an increasing number of sweepovirus species have been described in the last few years infecting sweet potato crops worldwide (Luan et al., 2007; Lozano et al., 2009; Albuquerque et al., 2011, 2012; Wasswa et al., 2011), only recently have satellites been shown in association with a sweepovirus. Geetanjali et al. (2013) showed the association of croton yellow vein mosaic betasatellite and papaya leaf curl betasatellite with sweet potato leaf curl virus (SPLCV) infecting Ipomoea purpurea in India. The study presented here has identified ToLCV-sat-like DNA satellites in association with sweepoviruses originating from Spain and Venezuela. On the basis of these findings it is proposed that the small non-coding DNA satellites associated with begomoviruses be considered a class of satellites distinct from betasatellites.

The study described here has identified small DNA satellites associated with sweepoviruses occurring in both the OW and NW (Lozano et al., 2009; Albuquerque et al., 2012). Although satellites from the two regions have related helper viruses, sweepoviruses, and infect related hosts (members of the genera Ipomoea and Merremia, both belonging to the family Convolvulaceae), they are genetically only distantly related. The structures of these DNA satellites share properties with other small non-coding DNA satellites identified in association with begomoviruses; ToLCV-sat, the first satellite identified in association with a DNA virus (Dry et al., 1997), satellites identified with bipartite NW begomoviruses infecting species of the Malvaceae in Cuba and satellites isolated from B. tabaci originating from Florida (Ng et al., 2011; Fiallo-Olivé et al., 2012).

Geminiviruses have a stringent size surveillance mechanism that operates during virus movement within the plant and there are also size constraints for encapsidation (Etessami et al., 1989; Elmer and Rogers, 1990; Frischmuth et al., 2001; Gilbertson et al., 2003). For defective interfering DNAs (diDNAs) associated with geminiviruses, which are typically half (∼1400 bp) the size of a begomovirus genome/DNA component, encapsidation has been shown to be in isometric particles (thus half a geminate particle; Frischmuth et al., 2001; Casado et al., 2004). Although betasatellites, which are typically 1350 bp in size, have been shown to be encapsidated in CP of the helper virus (Tabein et al., 2013) and insect transmissible (Saunders et al., 2000), the nature (multiplicity) of the particles has not been investigated. However, it would seem likely that, in common with diDNAs, they are encapsidated in isometric particles. The nature of virus particles encapsidating quarter unit-length DNAs, such as the ToLCV-sat-like satellites, remains unclear. However, encapsidation of ToLCV-sat DNA in the CP of ToLCV has been demonstrated by immunocapture PCR (Dry et al., 1997).

The sizes of the satellites characterized here range from 633 to 750. Thus, these molecules are about half the size of betasatellites and alphasatellites and about a quarter the size of begomovirus DNA components. The presence of an A-rich region is a common feature for all the DNA satellites associated with begomoviruses so far, including betasatellites, alphasatellites and the small non-coding satellites such as ToLCV-sat (Dry et al., 1997; Mansoor et al., 1999; Briddon et al., 2003; Fiallo-Olivé et al., 2012). For alphasatellites, it has been suggested that the A-rich region may be a “stuffer” required to increase the size of a nanovirus DNA component (∼1000 nt), from which alphasatellites are believed to have evolved, to that required for encapsidation by a begomovirus (∼1400 nt, about half the size of a begomovirus DNA component) (Mansoor et al., 2003). The betasatellites also have a size corresponding to half a begomovirus genome/DNA component with an A-rich region. By analogy to alphasatellites, this has been taken to indicate that betasatellites possible also have their origins as a component of another ssDNA virus, although no virus from which the component could have originated has yet been identified. Defective versions of betasatellites that are approximately half the size of full-length betasatellites are frequently identified in association with begomovirus-betasatellite infections (Briddon et al., 2003; Akhtar et al., 2014). Such defective molecules retain both the A-rich region and the SCR, but lack the βC1 coding sequence and resemble the satellite molecules identified in this study. Retention of the A-rich region and the SCR suggests that these are required for maintenance of these molecules by the helper begomovirus and thus is more than just a “stuffer,” since random mutation would otherwise rapidly reduce the A content. It has been suggested that the A-rich region may have a function in complementary-strand DNA replication, although evidence in support of this is lacking (Briddon et al., 2003).

For all the satellites discussed here, including the betasatellites, the presence of a second hairpin structure upstream of the SCR/SCR-like sequences appears to be a common feature; although for NW satellites the position differs from the remaining satellites (Figures 1B and 3C). Only for some satellites (ToLCV-sat and some of the sweepovirus-associated satellites) is this structure also associated with sequences that match the predicted iteron sequence of the helper virus. Lin et al. (2003) showed that mutation of the “iteron” sequence in ToLCV-sat, which abolished interaction of the satellite with ToLCV Rep in vitro, did not abolish trans-replication of the satellite in planta. Additionally, ToLCV-sat could be trans-replicated and maintained in plants by other begomoviruses, and even a curtovirus, which do not have the same iterons as ToLCV (Dry et al., 1997). These findings suggest that, although the iteron sequence of ToLCV-sat may be important for interaction with ToLCV, other sequences, or possibly the hairpin structure, can act as a Rep binding site for both ToLCV and other viruses. Nawaz-ul-Rehman et al. (2009) put forward two hypotheses to explain the promiscuous interaction of betasatellites with begomoviruses. The “universal Rep” hypothesis proposed that begomoviruses that interact with betasatellites have Rep proteins with more relaxed origin recognition properties. The “universal iteron” hypothesis proposed that betasatellites contain sequences that allow them to be recognized by a greater range of Rep proteins (so-called “iteron-like” sequences). Although it remains unclear which hypothesis may be correct, the weight of evidence lies with the “universal iteron” hypothesis at this time. Saunders et al. (2008) showed that sequences important for trans-replication of betasatellites lie in the vicinity of the secondary hairpin and Nawaz-ul-Rehman et al. (2009) showed that, in adapting to a new helper begomovirus, these sequences may change to improve interaction with the virus. Also, in common with betasatellites, some of the small satellites have iteron-like sequences adjacent to the secondary hairpin structure (Figures 3B,C). This suggests that, due to long association with a single helper begomovirus, satellites such as ToLCV-sat and the sweepovirus-associated satellites have evolved a more close relationship with their helper viruses, by encoding helper virus iterons, but nevertheless maintain the iteron-like sequences of their betasatellites progenitors allowing them to interact with other begomoviruses (Saunders et al., 2002b; Lin et al., 2003).

It is evident based on the analyses presented here that the ToLCV-sat-like satellites are distinct from betasatellites. They lack the βC1 gene and have sequences that have diverged significantly from the SCR of betasatellites, including nonanucleotide sequence-containing hairpin loop structures that are distinct from those of betasatellites. Additionally they do not occur in the presence of betasatellites, unlike the defective betasatellite that have been identified in association with many betasatellites. On the basis of these findings it is proposed that the ToLCV-sat-like satellites be considered a class of satellites distinct from betasatellites, for which the name “deltasatellites” is proposed – the Greek letter delta often being used in molecular biology to indicate a deletion or mutation. The deltasatellites include the two groups of satellites identified here, the satellites associated with sweepoviruses infecting Ipomoea sp. in Spain and Merremia in Venezuela, ToLCV-sat, the small molecules identified in Croton from India and Malvastrum from the Philippines, the satellites identified by Fiallo-Olivé et al. (2012) isolated from Malvastrum and Sidastrum, as well as a group of small molecules identified in B. tabaci whiteflies originating from the United States of America (Ng et al., 2011).

Although the deltasatellites and betasatellites are clearly related, it is unclear whether one group evolved from the other or they have distinct origins. Based on the premise that the simplest explanation is frequently the correct one, it would seem most likely that the delatasatellites evolved from the betasatellites. This possibility is supported by the apparent ease with which betasatellites lose their βC1 coding sequence (Briddon et al., 2003; Akhtar et al., 2014). Thus a betasatellite would have lost its βC1 gene, becoming approximately one quarter the size of a begomovirus genome/genomic component, and the SCR-derived sequences diverged from the parent betasatellite once the betasatellite was lost by the virus. What forces might have driven the divergence of the SCR sequences is unclear since the precise function of these sequences remains unknown. Since, the SCR occupies a position analogous to the common region shared by components of bipartite begomoviruses, which ensure the integrity of the split genome by virtue of encompassing the ori, it has been suggested that the SCR sequences may be involved in transreplication by the helper begomovirus (Briddon et al., 2003). However, for transreplication of betasatellites, specificity has been shown to be mediated by sequences between the A-rich region and the SCR (Saunders et al., 2008). If the SCR is involved in interactions with the helper begomoviruses, the divergence of the SCR sequences of deltasatellites might have been driven by the need to interact with the helper begomovirus.

The fact that the nonanucleotide-containing hairpin structures of delatasatellites differ from those of betasatellites suggests that the betasatellite(s) from which deltasatellites evolved has yet to be identified or is extinct. Additionally, the difference in the nonanucleotide-containing hairpin structures between distinct deltasatellites that each lineage has a distinct origin (convergent evolution). The identification of deltasatellites in a vegetatively propagated species provides a possible explanation of the present wide geographic distribution of these satellites. Sweet potato has its evolutionary origins in tropical South and Central America and has been transported to all the warmer parts of the World for cultivation (Roullier et al., 2013).This is not, however, evidence for a NW origin of deltasatellites since sweet potato germplasm has been moved both to and from its geographic origin (Paprotka et al., 2010a) and the sweepoviruses are typical OW begomoviruses (Brown et al., 2012). Further deltasatellites will need to be identified and characterized to more precisely determine the origins and evolution of these satellites.

Despite the fact that ToLCV-sat was identified in the 1990s (Dry et al., 1997), little is known concerning the maintenance of deltasatellites by begomoviruses and the effects they have on virus infections. So far only ToLCV-sat has been experimentally introduced into plants and these studies did not look at the effects of the satellite on symptoms and virus replication (Dry et al., 1997; Lin et al., 2003). There is thus a need to investigate the effects of deltasatellites on begomovirus infections and their interactions with helper begomoviruses, for comparison to the better characterized, and more numerous, betasatellites and alphasatellites. These aspects will be the focus of future studies.

JN-C designed and supervised the study. DC and FG-P provided and identified plant samples. GL, HT, and EF-O performed the experiments and bioinformatic analyses. JN-C and RWB performed bioinformatic analyses and wrote the main manuscript text. All authors reviewed the manuscript.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.