Roles and interactions of begomoviruses and satellite DNAs associated with okra leaf curl disease in Mali, West Africa

Geminiviruses are a family of plant-infecting viruses with circular single-stranded DNA genomes encapsidated into small twinned icosahedral virions. The family Geminiviridae is divided into four genera (Mastrevirus, Curtovirus, Topocuvirus and Begomovirus), which are differentiated based on their genome structure, the host plants they infect and the type of insect vector (Fauquet et al., 2008). Members of the genus Begomovirus have monopartite (one ∼2.8 kb DNA component) or bipartite (two ∼2.6 kb DNA components) genomes, infect dicotyledonous plants and are transmitted by whiteflies (Bemisia tabaci Genn.). Begomoviruses are one of the largest groups of plant viruses and cause economically important diseases of many vegetable and fibre crops (Rojas et al., 2005; Varma & Malathi, 2003).

Most Old World monopartite begomoviruses are associated with one or more betasatellite DNA(s), which are required for induction of typical disease symptoms and depend on the helper begomovirus for replication and movement (reviewed by Mansoor et al., 2006; Briddon & Stanley, 2006). Another type of satellite-like DNA component, termed DNA1, is also commonly associated with Old World monopartite begomoviruses. DNA1 satellites replicate autonomously, and may have originated from nanoviruses, based on the fact that they encode a nanovirus-like replication-associated protein (Rep) (Briddon et al., 2004; Mansoor et al., 1999; Saunders & Stanley, 1999). In contrast with betasatellites, DNA1 satellites have not been shown to play a role in the aetiology of begomovirus-induced diseases. Betasatellites and DNA1 are about half the size of a begomovirus component (∼1.4 kb) and require the helper begomovirus for systemic spread and insect transmission (Briddon & Stanley, 2006; Saunders et al., 2000).

In West Africa and the Nile Basin, diseases caused by begomoviruses have emerged in crops such as cotton, okra, pepper and tomato (Fauquet et al., 2005; Idris et al., 2005; Idris & Brown, 2002, 2005; Zhou et al., 2008). Evidence for the involvement of begomoviruses in these diseases includes leaf curling and crumpling symptoms, detection of begomovirus DNA and the presence of Bemisia whiteflies. Okra (Abelmoschus esculentus, family Malvaceae) is a widely grown vegetable crop in West Africa, where it originated. Okra leaf curl disease (OLCD) has become increasingly important in West Africa; it is characterized by stunted growth and leaf curling, distortion, mosaic, mottling and yellowing. A new begomovirus species, okra yellow crinkle virus (OYCrV) (Shih et al., 2006), was recently shown to be associated with OLCD in Mali. In other parts of the world, similar diseases of okra are also associated with begomovirus infection (De La Torre-Almaraz et al., 2006; Idris & Brown, 2002; Jose & Usha, 2003; Zhou et al., 1998).

In this study, we describe the molecular characterization of two begomovirus species, a betasatellite and a DNA1 satellite, from okra plants with OLCD in Mali. We have established that both begomovirus species, together with the betasatellite, can induce OLCD. We further demonstrate the complexity of the interactions of the OLCD-associated satellites with okra- and tomato-infecting begomoviruses.

Virus source and DNA extraction.
In June 2006, leaf samples from okra (BK2, BK5 and BK8) and Sida spp. plants (BK20 and BK30) showing OLCD symptoms were collected from around Bamako and Baguineda, Mali. Leaf samples were squashed onto nylon membranes and total plant DNA was extracted from dried tissue as described previously (Kon et al., 2002; Zhou et al., 2008).

Viral DNA detection and cloning.
Squash blot hybridization analysis with a general begomovirus probe was performed as described previously (Zhou et al., 2008). Total plant DNA was used as a template in a PCR with degenerate begomovirus primers (UPV1/UPC2; Briddon & Markham, 1994). PCR-amplified DNA fragments were cloned into the pCR2.1 TOPO vector (Invitrogen) and sequenced. Two pairs of overlapping primers, each with a BamHI site (underlined), were designed from these sequences to amplify full-length begomovirus DNA (OkBamHI-1 , 5'-GGATCCATTAGTCAACGAGTTC-3'/OkBamHI-2 , 5'-GGATCCCACATGTTGTAAATC-3' and OkBamHI-3 , 5'-GGATCCGTTATTGAACGACTTC-3'/OkBamHI-4 , 5'-GGATCCCACATAGTTATGGCGGAC-3'). PCR with degenerate primers (PBL1v2040/PCRc1; Rojas et al., 1993) was used for detection of the begomovirus DNA-B component.

Betasatellite DNA was detected by PCR with primers Beta01/02 (Briddon et al., 2002). PCR products were cloned into pCR2.1 and sequenced. An overlapping primer pair with a PstI site (underlined), OkBPstI-1 (5'-CTGCAGTCTATATGATCGTCTTTG-3') and OkBPstI-2 (5'-CTGCAGGATAGAGGTGACGGCAAC-3'), was designed to amplify the full-length betasatellite DNA.

The DNA1 satellite was detected by PCR with primers UN101/DNA102 as described by Bull et al. (2003). PCR products were cloned into pCR2.1 and sequenced. An overlapping primer pair with a BamHI site (underlined) was designed to amplify the full-length DNA1 component, OkD1BamHI-1 (5'-GGATCCCATAAAAGGAGACAAC-3') and OkD1BamHI-2 (5'-GGATCCAGCAGCACAGTATTC-3').

Sequence and phylogenetic analyses.
Begomovirus and satellite DNA sequences were determined by the dideoxynucleotide chain-termination sequencing method (for details see Supplementary Methods, available in JGV Online). Phylogenetic trees were generated by using MEGA software and recombination events were identified with the recombination detection program (RDP) (Supplementary Methods).

Production of constructs for infectivity studies.
To determine infectivity, begomovirus and satellite clones were generated in pCAMBIA1300 (Hajdukiewicz et al., 1994). For okra virus-1, an ∼1.2 kb BamHI–EcoRI fragment containing the intergenic region (IR) was cloned to generate a 0.4-mer (pCOk1B0.4). The full-length monomer was cloned into the BamHI-digested pCOk1B0.4 to generate a 1.4-mer (pCOk1B1.4). For okra virus-2, an ∼0.7 kb BamHI–EcoRI fragment containing the IR was cloned to generate a 0.2-mer (pCOk2B0.2). The full-length monomer was cloned into the BamHI-digested pCOk2B0.2 to generate a 1.2-mer (pCOk2B1.2). Clones of tomato-infecting begomoviruses were as described previously (Zhou et al., 2008) or generated as described in Supplementary Methods.

For the betasatellite, a tandem dimer was generated in pCAMBIA1300 (pCOkBP2.0) using PstI. For the DNA1 satellite, an ∼1.3 kb BamHI–EcoRI fragment was cloned to generate a 0.9-mer (pCOkD1B0.9). The full-length monomer was cloned into BamHI-digested pCOkD1B0.9 to generate a 1.9-mer (pCOkD1B1.9).

Inoculation of plants with cloned viral DNA.
The infectivity of the begomovirus and satellite clones was determined by particle bombardment and agroinoculation. DNA was coated onto gold particles as described by Gilbertson et al. (1991) and bombarded into 14-day-old okra (A. esculentus cv. Clemson Spineless) and cotton (Gossypium hirsutum cvs. STV474 and Pima) seedlings at 1550 p.s.i. with a PDS-1000 (Bio-Rad). Control plants were bombarded with gold particles. Bombarded plants were visually examined for disease symptoms up to 6 weeks post-bombardment.

For agroinoculation, the infectious cloned DNA in pCAMBIA1300 was transformed into Agrobacterium tumefaciens strain C58 by using the freeze–thaw method (Chen et al., 1994). Cotton, okra and other plant species (see Supplementary Methods) were agroinoculated by the needle puncture method, as described by Kon et al. (2003).

Detection of viral DNA by PCR and Southern blot hybridization analyses.
Total genomic DNA was extracted from leaf tissue as described by Kon et al. (2002). Viral and satellite DNA was detected by PCR with the previously described overlapping primer pairs. For Southern blot hybridization analyses, total genomic DNA was separated and transferred to membranes as described by Kon et al. (2003). Blots were probed with cloned begomovirus, betasatellite or DNA1 satellite DNA labelled with α³²P-labelled dCTP by nick translation. Blots were exposed to X-ray film and hybridization signals were quantified with ImageJ software ().

Transient replication assays.
An Agrobacterium leaf infiltration assay was used to determine the replication capacity of the cloned DNA1. A DNA1 mutant with an altered Rep gene initiation codon was generated to assess the role of this protein in satellite replication (Supplementary Methods).

Cloning and molecular characterization of two begomovirus species associated with OLCD in Mali
Squash blot hybridization analysis of the leaf samples from okra and Sida spp. with OLCD symptoms revealed begomovirus infection, and an ∼2.7 kb DNA fragment was amplified from all five samples by PCR with degenerate begomovirus primers. These DNA fragments were cloned and sequence analyses revealed two distinct begomoviruses (okra virus-1 and okra virus-2) in all five samples. PCR analyses failed to reveal a DNA-B component in these samples.

PCR and overlapping primers were used to amplify full-length begomovirus DNA components which were cloned to generate pOkML-1 (okra virus-1) and pOKML-2 (okra virus-2). The complete sequence of okra virus-1 is 2791 nucleotides (nt), whereas that of okra virus-2 is 2780 nt. The genome organization of both components is similar to that of Old World monopartite begomoviruses [i.e. two virus-sense open reading frames (ORFs) (V1 and V2) and four complementary-sense ORFs (C1–C4)]. The IR sequences (between the start codons of the C1 and V1 ORFs) are ∼300 nt (nt 2636–161 for okra virus-1 and 2633–161 for okra virus-2). Within each IR is the characteristic geminivirus stem–loop structure containing the conserved nonanucleotide sequence in the loop, putative Rep (C1) protein high affinity binding sites and the ORF C1 TATA box. The Rep high affinity binding sites for okra virus-1 are GGGGGAACTGGGGG (core binding sites underlined, nt 2683–2696), with direct (GGGGG, nt 2658–2662) and inverted (CCCCC, nt 2711–2715) repeats flanking the high affinity binding sites and C1 TATA box. The okra virus-2 Rep high affinity binding sites are GGTGTATTGGTAG (nt 2684–2696), with inverted (ACACC, nt 2645–2649) and direct (GGTGT, nt 2707–2711) repeats flanking the high affinity binding sites and C1 TATA box.

The complete and IR nucleotide sequences of okra virus-1 and okra virus-2 are 72.8 and 67.8 % identical, respectively, consistent with these being distinct begomovirus species. The okra virus-1 sequence is >99 % identical to those of OYCrV isolates from Mali, and <80 % identical to sequences of other begomoviruses (Table 1). Thus, okra virus-1 is an isolate of OYCrV and has been designated OYCrV-[Mali:2006] (OYCrV-[ML:06]).

Table 1. Nucleotide identities (%) for total nucleotide (nt) and IR and nucleotide and amino acid (aa) sequence identities for ORFs of two begomoviruses associated with OLCD in Mali and selected previously characterized begomoviruses The highest value for each sequence is underlined. TbLCZV, tobacco leaf curl Zimbabwe virus; ToLCKMV, tomato leaf curl Comoros virus; ToLCArV, tomato leaf curl Arusha virus; ToCSV, tomato curly stunt virus; TYLCMLV, tomato yellow leaf curl Mali virus; ToLCSDV, tomato leaf curl Sudan virus.

The okra virus-2 sequence was most identical (87–94 %) to those of cotton leaf curl Gezira virus (CLCuGV) isolates, and the highest identities were with two isolates from okra in Niger (Table 1). Sequences of the okra virus-2 V1, V2 (CP), C1 (Rep), C2 and C3 ORFs were generally >92 % [nt and amino acid (aa)] identical with CLCuGV homologues (Table 1), whereas identities were lower for IR (72–86 %) and C4 sequences (<90 % nt and <85 % aa). These results indicated a recombinant genome. This was confirmed by using RDP analysis, which indicated that the majority of the okra virus-2 genome was derived from an isolate of CLCuGV (e.g. CLCuGV-NE[NE:02:Ok]), whereas the remainder (nt 2389–2715) was from hollyhock leaf crumple virus (HoLCrV) (e.g. HoLCrV-[EG:Cai:97]; P-value 3.001x10^–5). Thus, the recombinant region includes the 5' portions of the IR and C1 and C4 ORFs. Consistent with this analysis, the left IR (5' end of the IR to the nicking site), including the Rep protein high affinity binding sites, as well as the Rep iteron-related domain of the Rep (Arguello-Astorga & Ruiz-Medrano, 2001) of okra virus-2 were nearly identical to those of HoLCrV (data not shown). Because the complete nucleotide sequence of okra virus-2 is >89 % identical to sequences of CLCuGV isolates, it has been designated CLCuGV-Mali[Mali:Okra:2006] (CLCuGV-ML[ML:Ok:06]).

Phylogenetic analyses placed okra virus-1 in a cluster with OYCrV isolates, and this cluster was included in a larger cluster that had monopartite tomato-infecting begomoviruses from Africa (Fig. 1). Okra virus-2 was placed into a different cluster, which comprised isolates of CLCuGV and HoLCrV, most of which are from sub-Saharan Africa.

(31K): Fig. 1. Phylogenetic consensus tree showing the relationship of two begomoviruses associated with OLCD in Mali with previously characterized begomoviruses based on an alignment of complete nucleotide sequences. The positions of OYCrV-[ML:06] and CLCuGV-ML[ML:Ok:06] are indicated with arrows. Branch strengths were evaluated by constructing 1000 trees in bootstrap analysis by step-wise addition at random; bootstrap values are shown. The horizontal line lengths are in proportion to genetic (mutation) distances, as indicated by the scale bar. The DNA-A component of the bipartite begomovirus cotton leaf crumple virus was used as an outgroup. Sequences were obtained from GenBank and virus abbreviations are as described by Fauquet et al. (2008).

Cloning and molecular characterization of OLCD-associated satellite DNAs
The expected-sized DNA fragment (∼1.4 kb) was amplified from all five samples by PCR with the degenerate betasatellite primers; sequence analysis indicated the presence of a single betasatellite. The full-length betasatellite component was amplified by PCR with overlapping primers, and cloned to generate pOkBML-1. The OLCD-associated betasatellite is 1346 nt and has characteristic features including a single complementary-sense ORF (βC1), an adenine (A)-rich region (nt 720–877 with 60 % A residues) and a satellite conserved region (SCR) with a predicted stem–loop structure containing the geminivirus nonanucleotide sequence (TAATATTAC). Within the SCR are sequence motifs (GGTDKN; D=A, G, T; K=G, T; N=A, C, G, T) with similarity to begomovirus Rep binding sites, including those of OYCrV and CLCuGV.

The sequence of the OLCD-associated betasatellite was most identical (87–99 %) to isolates of cotton leaf curl Gezira betasatellite (CLCuGB); it had near identity (99.2 %) with a recently described betasatellite from tomato in Mali (CLCuGB-[Mali:Tomato:2005], GenBank accession number DQ136001[GenBank] ). The predicted βC1 amino acid sequences of these two betasatellites are identical, whereas identities with βC1 sequences of other CLCuGBs were lower (∼90 %; data not shown). Phylogenetic analyses placed the OLCD-associated betasatellite in a cluster with CLCuGBs, most of which are from sub-Saharan Africa (Fig. 2a). Thus, the OLCD-associated betasatellite is an isolate of CLCuGB, and is designated CLCuGB-[Mali:Okra:2006] (CLCuGB-[ML:Ok:06]).

(32K): Fig. 2. Phylogenetic consensus trees showing the relationship of the betasatellite and DNA1 component associated with OLCD in Mali with other betasatellites and DNA1 components based on alignments of complete nucleotide sequences. The positions of the OLCD-associated betasatellite and DNA1 are indicated with arrows. (a) Betasatellite phylogenetic tree rooted with the sequence of the tomato leaf curl virus satellite DNA. Sequences were obtained from GenBank and betasatellite abbreviations are as described by Briddon et al. (2008). (b) DNA1 tree rooted with the sequence of the C2 component of the nanovirus faba bean necrotic yellow virus. Abbreviations and GenBank accession numbers for DNA1 satellites associated with their respective diseases are as follows: Ageratum yellow vein [AYVD1-IN01 (AJ512958[GenBank] ), -IN02 (AJ512959[GenBank] ), -KE01 (AJ512963[GenBank] ), KE02 (AJ512964[GenBank] ), -PK01 (AJ512951[GenBank] ), -PK02 (AJ512952[GenBank] ), -PK03 (AJ512948[GenBank] ), -PK04 (AJ512949[GenBank] ) and -SG (AJ238493[GenBank] )], cotton leaf curl [CLCuD1-EG (AJ512960[GenBank] ), -IN (AJ512957[GenBank] ), -PK01 (AJ132344[GenBank] ) and -PK02 (AJ132345[GenBank] )], hibiscus leaf curl [HLCD1-PK01 (AJ512950[GenBank] ) and -PK02 (AJ512953[GenBank] )], hollyhock leaf crumple [HLCrD1-EG (AJ512962[GenBank] )], okra leaf curl [OLCD1-PK (AJ512954[GenBank] )], okra yellow vein [OYVD1-EG (AJ512961[GenBank] )], Sida leaf curl [SiLCD1-VI (DQ641717[GenBank] )], tobacco curly shoot [TbCSD1-Y132 (AJ579349[GenBank] ) and -Y137 (AJ579351[GenBank] )], tobacco leaf curl Yunnan [TbLCYND1-Y276 (AJ888455[GenBank] )], tobacco leaf curl [TbLCD1-PK (AJ512956[GenBank] )] and tomato leaf curl [ToLCD1-PK (AJ512955[GenBank] )]. The branch strengths of trees in (a) and (b) were evaluated by constructing 1000 trees in bootstrap analysis by step-wise addition at random; bootstrap values are shown. The horizontal line lengths are in proportion to genetic (mutation) distances, as indicated by the scale bar.

The expected-sized DNA fragment (∼0.5 kb) was amplified from all five samples by PCR with the degenerate DNA1 primers; sequence analyses indicated the presence of a single DNA1. The full-length DNA1 component was amplified by PCR with overlapping primers and cloned to generate pOkD1ML-1. The OLCD-associated DNA1 is 1388 nt and has characteristic features, including a stem–loop structure, with the conserved nanovirus nonanucleotide sequence (TAGTATTAC) and a single viral-sense ORF, which encodes a predicted Rep protein of 36.5 kDa.

The sequence of this DNA1 satellite was divergent (<73 % identity) from those of previously described DNA1s; phylogenetic analyses placed it basal to previously described DNA1s (Fig. 2b). Identities for the Rep protein amino acid sequence were higher (84–87 %) (data not shown), consistent with previous reports (Briddon et al., 2004). Thus, this satellite has a typical DNA1 genome organization, but a divergent sequence.

Particle bombardment inoculation of cloned DNA of the OLCD-associated begomoviruses and satellites
The cloned DNA of OYCrV and CLCuGV bombarded individually or with the cloned DNA of the DNA1 satellite induced symptomless infections in a small number of okra plants (Table 2; Fig. 3b, d, e and i). In contrast, okra seedlings co-bombarded with cloned DNA of the CLCuGB and those of either begomovirus developed typical OLCD symptoms (leaf curling and crinkling) (Table 2; Fig. 3c and f). Okra seedlings co-bombarded with cloned DNA of CLCuGB and both begomoviruses also developed OLCD symptoms (Table 2, Fig. 3h). When cloned DNA of CLCuGB and DNA1 were co-bombarded with that of OYCrV, CLCuGV or both begomoviruses into okra seedlings, OLCD symptoms developed in 20, 66 and 53 % of plants, respectively (Table 2; Fig. 3j, k and l). The presence of begomovirus and satellite DNA in selected symptomatic plants was confirmed by PCR with specific primers (Table 2). These results indicate that: (i) OLCD in Mali is caused by a complex of two begomovirus species and a single betasatellite and (ii) the DNA1 satellite did not influence OLCD development.

Table 2. Infectivity of OYCrV-[ML:06] and CLCuGV-ML[ML:Ok:06] alone or in combination with CLCuGB-[ML:Ok:06] and/or the DNA1 satellite associated with OLCD following particle bombardment inoculation of infectious cloned DNA

(128K): Fig. 3. Disease symptoms in okra plants infected with various combinations of the infectious cloned DNA of begomoviruses and satellite DNAs associated with OLCD in Mali following particle bombardment inoculation. Okra plants bombarded with gold particles (a), OYCrV (b), OYCrV and CLCuGB (c), CLCuGB (d), CLCuGV (e), CLCuGV and CLCuGB (f), OYCrV and CLCuGV (g), OYCrV, CLCuGV and CLCuGB (h), DNA1 (i), OYCrV, CLCuGB and DNA1 (j), CLCuGV, CLCuGB and DNA1 (k) or OYCrV, CLCuGV, CLCuGB and DNA1 (l). Plants were photographed 28 days following bombardment.

A similar series of inoculations was performed with cotton seedlings. The DNA-A (1.8-mer) and DNA-B (1.6-mer) clones of the cotton-infecting bipartite begomovirus, cotton leaf crumple virus (CLCrV-CA) (Seo et al., 2006), were used as a positive control. Cotton seedlings (∼70 %) bombarded with the CLCrV-CA clones developed leaf crumple symptoms by 21 days post-inoculation (p.i.) (Table 2). In contrast, none of the seedlings bombarded with any combination of the cloned DNA of OYCrV, CLCuGV, CLCuGB and/or DNA1 developed obvious disease symptoms or became systemically infected, based on PCR analysis of newly emerged leaves (Table 2).

Agroinoculation of cloned DNA of the OLCD-associated begomoviruses and satellites
All Nicotiana benthamiana plants agroinoculated with OYCrV or CLCuGV developed mild downward leaf curling symptoms by 14 days p.i. Plants agroinoculated with CLCuGB or DNA1 alone did not develop symptoms, nor was satellite DNA detected in newly emerged leaves of these plants. N. benthamiana plants co-agroinoculated with DNA1 and OYCrV or CLCuGV developed symptoms indistinguishable from those of plants infected with the viruses alone; whereas plants co-agroinoculated with CLCuGB and OYCrV or CLCuGV developed severe downward leaf curling and stunting symptoms, which were considerably more severe than those induced by either virus alone (Supplementary Table S1, available in JGV Online). In all cases, DNA of the inoculated begomovirus and satellite were detected by PCR (Supplementary Table S1). Together, these results indicate that the CLCuGB, but not DNA1, increased the severity of symptoms induced by both begomoviruses in N. benthamiana.

Common bean, cotton, okra, tobacco and tomato plants agroinoculated with these begomovirus/satellite combinations did not develop symptoms. OYCrV, but not CLCuGV, induced symptomless infections in a small number of agroinoculated tobacco and tomato plants (Supplementary Table S1). Here, it is important to note that, in some cases, the failure of a plant species to be infected with a begomovirus by agroinoculation can be due to an incompatibility with A. tumefaciens (Saeed, 2008). Indeed, this was the case for okra, which was infected with the begomovirus/CLCuGB combinations following particle bombardment inoculation, but not agroinoculation (Table 2; Supplementary Table S1). However, in the case of the other hosts, including cotton, compatibility with the A. tumefaciens strain used in this study (C58C1) has been established (Supplementary Table S1; Hagen et al., 2008). Thus, these results suggest that the begomovirus/CLCuGB combinations have a relatively narrow host range.

OLCD-associated satellite DNA interacts with West African tomato-infecting begomoviruses
To determine whether CLCuGB could functionally interact with tomato leaf curl Mali virus (ToLCMLV-ML[ML:03]) and tomato yellow leaf crumple virus (ToYLCrV-[ML:03]), N. benthamiana plants were agroinoculated with various combinations of viral DNA, CLCuGB and DNA1 (Table 3). ToLCMLV induced upward leaf curling symptoms, whereas ToYLCrV induced downward leaf curling (Table 3; Zhou et al., 2008). Plants co-agroinoculated with CLCuGB and ToLCMLV or ToYLCrV developed severe downward leaf curling and stunting symptoms, which were more severe than those induced by either virus alone (Table 3).

Table 3. Infectivity of two tomato-infecting begomoviruses from Mali alone or in combination with CLCuGB-[ML:Ok:06] and/or the DNA1 satellite associated with OLCD following agroinoculation of infectious cloned DNA In each case, infection was detected in 15 of 15 plants inoculated (three experiments), based upon detection of viral and satellite DNA by PCR with virus-, betasatellite- and/or DNA1-specific primers.

Begomovirus and satellite DNA levels in N. benthamiana plants infected with different virus–satellite combinations
Southern blot hybridization confirmed that CLCuGB was replicated to high levels by all four begomoviruses (Fig. 4a–d), consistent with the increased symptom severity in the presence of the betasatellite. Furthermore, DNA levels of OYCrV and CLCuGV in plants co-infected with CLCuGB were considerably higher than those in plants infected with the viruses alone (247±24 % for OYCrV/CLCuGB and 174±18 % for CLCuGV/CLCuGB; Fig. 4a and b, compare lanes 4 and 5 with lanes 6 and 7). In contrast, levels of ToLCMLV and ToYLCrV DNA were increased only slightly in the presence of CLCuGB (116±9 % for ToLCMLV/CLCuGB and 137±10 % for ToYLCrV/CLCuGB; Fig. 4c and d).

(61K): Fig. 4. Southern blot hybridization analyses showing DNA levels of four begomoviruses, OYCrV (a), CLCuGV (b), ToLCMLV (c) and ToYLCrV (d), and satellite DNAs associated with OLCD in Mali in N. benthamiana plants infected via agroinoculation. Lanes: 1, empty vector; 2, CLCuGB; 3, DNA1; 4 and 5, begomovirus alone; 6 and 7, begomovirus and CLCuGB; 8 and 9, begomovirus and DNA1; 10 and 11, begomovirus, CLCuGB and DNA1. Blots in the upper, middle and lower panels were hybridized with the corresponding begomovirus, CLCuGB or DNA1 probes, respectively. Size markers are indicated (kb).

Southern blot hybridization also confirmed that DNA1 was replicated and moved systemically in the presence of all four begomoviruses (Fig. 4a–d). In contrast with the results for CLCuGB, the level of DNA of all four begomoviruses was reduced in the presence of DNA1 compared with that in plants infected with the viruses alone (OYCrV/DNA1, 62±15 %; CLCuGV/DNA1, 78±7 %; ToLCMLV/DNA1, 30±4 %; and ToYLCrV/DNA1, 43±8 %; Fig. 4a–d, compare lanes 8 and 9 with lanes 4 and 5). Interestingly, this reduction of viral DNA levels was not associated with an attenuation of disease symptoms (Table 3 and Supplementary Table S1).

In plants co-infected with each begomovirus and both satellites, DNA1 levels were substantially increased compared with those in plants infected with the begomoviruses and DNA1 (OYCrV/CLCuGB/DNA1, 441±53 %; CLCuGV/CLCuGB/DNA1, 154±15 %; ToLCMLV/CLCuGB/DNA1, 320±83 %; and ToYLCrV/CLCuGB/DNA1, 310±70 %; Fig. 4a–d, compare lanes 10 and 11 with lanes 8 and 9). Moreover, the level of viral DNA was only slightly reduced in these plants compared with that in plants infected with the begomoviruses and CLCuGB (OYCrV/CLCuGB/DNA1, 88±7 %; CLCuGV/CLCuGB/DNA1, 85±6 %; ToLCMLV/CLCuGB/DNA1, 52±10 %; and ToYLCrV/CLCuGB/DNA1, 80±4 %; Fig. 4a–d, compare lanes 10 and 11 with lanes 6 and 7). Finally, the level of CLCuGB DNA was slightly reduced compared with that in plants infected with the begomoviruses and CLCuGB (OYCrV/CLCuGB/DNA1, 72±6 %; CLCuGV/CLCuGB/DNA1, 62±7 %; ToLCMLV/CLCuGB/DNA1, 55±10 %; and ToYLCrV/CLCuGB/DNA1, 61±25 %; Fig. 4a–d, compare lanes 10 and 11 with lanes 6 and 7). Disease symptoms in plants infected with the begomoviruses and both satellites were indistinguishable from those in plants infected with the viruses and CLCuGB.

CLCuGB functionally interacts with tomato yellow leaf curl virus-Israel (TYLCV-IL) in tomato
To assess whether CLCuGB could functionally interact with TYLCV-IL in tomato plants, an agroinoculation system was generated with an isolate of TYLCV from the Dominican Republic (TYLCV-IL[DO]; Salati et al., 2002). Tomato plants agroinoculated with TYLCV developed typical yellowing and leaf curling (12 of 12 plants; Fig. 5a). In contrast, those co-agroinoculated with TYLCV and CLCuGB developed much more severe symptoms, including distorted and stunted growth and leaf curling and crumpling (10 of 12 plants; Fig. 5a). The level of viral DNA in plants co-infected with TYLCV and CLCuGB was considerably higher compared with that in plants infected with the virus alone (average 335 % for TYLCV/CLCuGB; Fig. 5b, compare lanes 7–10 with lanes 3–6). Southern blot hybridization confirmed that the CLCuGB was replicated to high levels in the presence of TYLCV (Fig. 5b; lanes 7–10), consistent with the increased symptom severity. Thus, these results indicate that TYLCV mediates replication and movement of CLCuGB.

(82K): Fig. 5. Disease symptoms and viral and satellite DNA levels in tomato plants infected with TYLCV alone or with TYLCV and CLCuGB via agroinoculation. (a) Plants infected with empty vector (left), TYLCV (centre) and TYLCV and CLCuGB (right). (b) Southern blot hybridization with TYLCV (upper panel) or CLCuGB (lower panel) probes. Lanes: 1, empty vector; 2, CLCuGB; 3–6, TYLCV; and 7–10, TYLCV and CLCuGB. Size markers are indicated (kb).

Autonomous replication of OLCD-associated DNA1 requires the DNA1-encoded Rep
Southern blot hybridization analysis of DNA extracts from leaves infiltrated with the Agrobacterium strain containing the DNA1 1.9-mer clone (pCOkD1B1.9) revealed single-stranded (ss) and double-stranded forms [e.g. covalently closed circular (ccc) and open circular (oc)], indicating autonomous DNA1 replication (Supplementary Fig. S1a). In contrast, DNA1 replication was not detected in leaves that were infiltrated with the strain containing the DNA1 which had a mutated Rep gene start codon or that were co-infiltrated with strains containing this mutant and OYCrV or CLCuGV (Supplementary Fig. S1b). Additionally, in leaves co-infiltrated with strains containing DNA1 and CLCuGB, no evidence of betasatellite DNA replication was detected (data not shown). Together, these results indicate that (i) DNA1, but not CLCuGB replicates autonomously, (ii) the DNA1 Rep protein is required for DNA1 replication, (iii) DNA1 replication is not mediated by the OYCrV or CLCuGV Rep and (iv) CLCuGB replication is not mediated by the DNA1 Rep. In West Africa, OLCD is an emerging disease that can cause significant yield losses. Here, we establish that OLCD is caused by a complex of two Old World monopartite begomovirus species and a promiscuous betasatellite (CLCuGB). Okra virus-1 is an isolate of OYCrV (OYCrV-[ML:06]), a begomovirus recently associated with OLCD in Mali (Shih et al., 2006). The placement of OYCrV in a distinct cluster with indigenous West Africa tomato-infecting begomoviruses (Zhou et al., 2008) suggests that it evolved locally to infect okra. Okra virus-2 is a recombinant begomovirus with >89 % sequence identity with isolates of CLCuGV; thus, it was designated CLCuGV-ML[ML:Ok:06]. This was further supported by the placement of this okra-infecting begomovirus from Mali in a distinct phylogenetic cluster with isolates of CLCuGV and HoLCrV, monopartite begomoviruses from sub-Saharan Africa and Egypt that infect various malvaceous hosts. This cluster represents another lineage of African begomoviruses, which may have emerged from a progenitor present in an indigenous host, such as the CLCuGV-like virus detected in Sida spp. in Mali. The recombinant region of CLCuGV-ML[ML:Ok:06] came from another malvaceous begomovirus, possibly HoLCrV, and includes the 5' portions of the IR and the C1 and C4 ORFs. This region is commonly exchanged among begomoviruses (Hou & Gilbertson, 1996; Lefeuvre et al., 2007) and may provide a selective advantage, such as enhanced replication or suppression of host defences.

Evidence that the OYCrV and CLCuGV isolates from okra in Mali have a monopartite genome was provided by the infectivity of their cloned DNA in N. benthamiana and the failure to detect an associated DNA-B component. However, neither virus induced OLCD symptoms in okra, indicating a more complex aetiology. The finding that a betasatellite, when co-inoculated with the infectious clones of OYCrV or CLCuGV, resulted in development of typical OLCD symptoms established that the disease is caused by a begomovirus/satellite complex. Further support for this notion came from the fact that this betasatellite increased viral DNA levels and disease symptoms in N. benthamiana. Thus, OYCrV and CLCuGV are monopartite begomoviruses that require a betasatellite for induction of typical disease symptoms, similar to complexes responsible for bhendi yellow mosaic disease of okra in India (Jose & Usha, 2003) and ageratum yellow vein, cotton leaf curl and tomato yellow leaf curl diseases in Asia (reviewed by Mansoor et al., 2006).

The sequence of the OLCD-associated betasatellite was >78 % identical to that of CLCuGB, which is above the recently established threshold for designating new betasatellite species (Briddon et al., 2008); thus, it is now designated CLCuGB-[ML:Ok:06]. This is also consistent with results of the phylogenetic analysis, in which the OLCD-associated satellite was placed in a cluster with previously characterized CLCuGBs from Africa (Fig. 2a). These results suggest that there is a long evolutionary history of begomoviruses that infect malvaceous hosts (Idris et al., 2005). The fact that two distinct begomovirus species, OYCrV and CLCuGV, were helper viruses for the CLCuGB revealed the promiscuous nature of the betasatellite. Further evidence for this came from the finding that three monopartite tomato-infecting begomovirus species (ToLCMLV-ML[ML:03], ToYLCrV-[ML:03] and TYLCV-IL[DO]) also served as helper viruses for CLCuGB. These results are fully consistent with previous reports of promiscuity in betasatellite replication (Mansoor et al., 2003, 2006). Moreover, the failure to identify the Rep protein high affinity binding sites of these five begomoviruses in the CLCuGB SCR supports the notion that these sequences are not required for betasatellite replication mediated by the begomovirus Rep protein (Lin et al., 2003).

It is possible that exchange of betasatellites could extend the virus host range, thereby leading to new diseases. Moreover, this could negatively impact management strategies that are designed based upon the biology of a given begomovirus. For example, a 2–3 month tomato-pepper free period, combined with the use of early maturing and moderately resistant varieties, has been an effective management strategy for tomato-infecting begomoviruses in Mali due to the narrow host range of these viruses (Zhou et al., 2008). The persistence of promiscuous betasatellites, such as CLCuGB in okra or other hosts (e.g. Sida spp.), could facilitate the emergence of highly pathogenic begomovirus/betasatellite complexes, which could cause more severe disease symptoms or overcome disease resistance.

The identification of a CLCuGV/CLCuGB complex in Mali is a concern because of the economic importance of cotton in West Africa. CLCuGV is a diverse and complex species composed of host-adapted strains that infect malvaceous hosts, such as cotton, hollyhock, okra and Sida alba (Bigarre et al., 2001; Idris & Brown, 2002). Evidence for host specificity in CLCuGV came from a study showing that an okra-infecting strain is infectious in okra, Malva parviflora and hollyhock, but not in cotton, whereas a cotton-infecting strain infected cotton and hollyhock, but not okra (Idris & Brown, 2002). The finding that CLCuGV-ML[ML:Ok:06] from okra did not infect cotton, alone or when co-inoculated with CLCuGB, is consistent with this host specificity. However, it will be important to monitor cotton for the emergence of a cotton-infecting CLCuGV variant in West Africa, given the genetic potential for such variants to adapt to new hosts.

DNA1 satellite-like molecules are associated with monopartite begomovirus/betasatellite disease complexes mostly in Asia, but also in Egypt and Kenya. These satellites are believed to have evolved from satellite-like, Rep-encoding components associated with nanoviruses (Briddon et al., 2004; Briddon & Stanley, 2006; Saunders & Stanley, 1999). The genome organization of the OLCD-associated DNA1 was similar to those of other DNA1s, but the sequence was highly divergent (<73 % identity) from previously characterized DNA1s (Briddon et al., 2004; Saunders et al., 2002). It is likely that this DNA1 has been geographically isolated for a long period of time and represents a distinct West African lineage of these satellites. Thus, our results extend the known geographical distribution and the genetic diversity of these satellites, and are consistent with a long-term association with monopartite begomoviruses (Briddon & Stanley, 2006).

Results from our transient replication assays established that the OLCD-associated DNA1 replicated autonomously. Moreover, DNA1 replication was specifically mediated by the DNA1 Rep, as the DNA1 mutant was not replicated by OYCrV or CLCuGV. On the other hand, the CLCuGB did not replicate autonomously and replication of this satellite was specifically mediated by the begomovirus Rep, as it was not replicated by the DNA1. These results probably reflect the difference in the begomovirus and DNA1 replication origins. However, our results did reveal promiscuity for the DNA1, as the okra- and tomato-infecting begomoviruses served as helper viruses. This must reflect a lack of specificity in factors involved in movement (e.g. encapsidation).

The role of DNA1 satellites in diseases induced by begomovirus/betasatellite complexes remains unclear. In the present study, the OLCD-associated DNA1 reduced helper begomovirus DNA levels, presumably due to competition for host factors involved in DNA replication, but this did not influence (attenuate) symptom development. The finding that the DNA levels of both the helper begomovirus and DNA1 were increased in the presence of CLCuGB indicates a beneficial effect for both components, perhaps due to increased cell division, enhanced movement (Saeed et al., 2007) or suppression of host defences (e.g. gene silencing). Support for the latter hypothesis comes from the finding that the βC1 of many betasatellites suppresses gene silencing (Cui et al., 2005; Kon et al., 2007). Thus, our results indicate that the DNA1 satellite is most likely a molecular parasite of the helper begomovirus.

In conclusion, OLCD in Mali is caused by a complex of begomoviruses and a promiscuous betasatellite. Results of PCR and dot blot analyses have also revealed the presence of OYCrV, CLCuGV and CLCuGB in OLCD samples from Burkina Faso, Ghana and other locations in Mali (e.g. Mopti, Sotuba and Sikasso) (data not shown), indicating that this complex is causing the disease throughout West Africa. The complex aetiology of this disease and the capacity of the CLCuGB to be replicated by other begomovirus species increases the possibility of recombination and reassortment events, which could lead to evolution of new recombinant viruses or begomovirus complexes with different biological properties (Lefeuvre et al., 2007; Rojas et al., 2005; Seal et al., 2006).

This research was funded by grants from the United States Agency for International Development (USAID) as part of the Integrated Pest Management-Collaborative Research Support Program (IPM-CRSP) and Agricultural Biotechnology Support Project II (ABSP II). We thank Nasrin Hakimi for excellent technical assistance.