Interaction of DNA with the Movement Proteins of Geminiviruses Revisited

Abstract

Geminiviruses manage the transport of their DNA within plants with the help of three proteins, the coat protein (CP), the nuclear shuttle protein (NSP), and the movement protein (MP). The DNA-binding capabilities of CP, NSP, and MP of Abutilon mosaic virus (AbMV; family Geminiviridae; genus Begomovirus) were scrutinized using gel mobility shift assays and electron microscopy. CP and NSP revealed a sequence-independent affinity for both double-stranded and single-stranded DNA, as has been previously reported for other begomoviruses. MP interacted selectively with dimeric supercoiled plasmid DNA in the electrophoretic assay. Further apparent size- and form-selective binding capacities of MP have been previously reported for another geminivirus (Bean dwarf mosaic virus), but in the case of AbMV, they have been identified as the result of electrophoretic interference rather than of complex formation. Without these complications, electron microscopy confirmed the assembly of double-stranded supercoiled DNA with NSP and MP into conspicuous structures and provided the first direct evidence for cooperative interaction of MP, NSP, and DNA. Based on these results and previous ones, a transport model of geminiviruses is discussed in which NSP packages DNA and MP anchors this complex to the protoplasmic leaflets of plasma membranes and microsomes for cell-to-cell movement.

Plant viruses differ from animal viruses in their mechanisms of transport within the host. Once injected into a plant, they move predominantly symplastically from cell to cell by using phloem cells for long-distance transport. Viral movement proteins have evolved to manage this task, especially to open the gates (plasmodesmata) between cells (27). What mechanism plant geminiviruses use to mediate the transfer of their DNA from cell to cell is still an open and challenging question. Most of our knowledge about this process is based upon indirect evidence, mainly derived from in vitro assays, because the transport complex itself has not been identified during natural infection. Geminiviruses (36) package circular single-stranded DNA (ssDNA) into twinned particles, hence their name. Depending on the particular geminivirus, three different proteins are involved in mediating transport through host plants, the coat protein (CP), the nuclear shuttle protein (NSP), and the movement protein (MP). These protein designations have been widely used in the literature (25) and will be used here also, although in a strict sense all three proteins (CP, NSP, and MP) may be called movement proteins, as will be discussed below.

Geminiviral capsids are composed of a single CP, encoded by the gene V1, V2, or AV1 (also known as AR1), depending on the geminivirus (56). The geminivirus genome may be segmented (bipartite) or not (monopartite), and in both cases one DNA molecule occupies the complete twinned particle (3, 56). For monopartite geminiviruses, CP is essential for systemic spread through the plant (5, 7, 26, 28, 34, 52). For bipartite geminiviruses, CP is not absolutely necessary for this task, and NSP (encoded by the gene BV1, also known as BR1, on DNA B) substitutes for its function in transport (16, 18, 44, 46). However, if NSP is mutated, CP can still complement this defect (20). Hence, CP and NSP share, in part, one redundant function and probably have a common evolutionary origin (R. Kikuno, H. Toh, H. Hayashida, and T. Miyata, Letter, Nature 308:562, 1984). All geminiviruses express a further protein, designated MP, which is essential for viral movement from cell to cell. This protein is encoded by different genes, dependent on the virus: V1 or V2 for monopartite geminiviruses (4, 9, 34, 52) and BC1 (also known as BL1) for bipartite geminiviruses (10, 11, 13, 23, 25, 30, 39, 48, 49).

Over the past decade, consensus has developed that MP cooperates with NSP or CP during viral transport (15, 25). However, the viral DNA conformation which participates in the natural transport complex is unknown.

Geminiviruses replicate in cell nuclei and, consequently, have to cross two borders during spread, the nuclear envelope and the plasma membrane. Both CP and NSP are targeted to the nucleus, and at least NSP, hence its name, is able to shuttle between the nucleus and the cytoplasm (19, 24, 33, 38, 46, 47, 50). Protein domains responsible for DNA binding and intracellular localization have been analyzed for several geminiviruses. A consensus exists that CP and NSP bind ssDNA as well as double-stranded DNA (dsDNA) in a sequence-independent manner (25).

By comparison, the role of MP in DNA binding and movement is still controversial. Depending on the geminivirus analyzed, MP has been localized in “cell wall-enriched” (48) and microsomal (30) fractions as well as in the lumen of particular endoplasmic reticulum-derived tubules of the protophloem (51). More detailed information is available for the MP of Abutilon mosaic virus (AbMV). Cellular imaging of AbMV MP translationally fused to green fluorescent protein or glutathione-S-transferase as reporter proteins showed that MP is targeted to the cell periphery in plants (55) or to microsomes and the plasma membranes in fission yeast cells (1), respectively. Sequence analysis of AbMV MP revealed no plausible domain which may function as a transmembrane helix. However, a membrane-binding domain of AbMV MP has been delimited, and this domain harbors a putative amphipathic helix which may allow the insertion of MP into one leaflet of a membrane (54). It has been proposed that MP serves as a membrane anchor at the protoplasmic face of microsomes and plasma membranes, thus facilitating the movement of the transport complex along these membranes (54). Moreover, MP can shuttle between the nuclear envelope and the cellular periphery in order to fulfill its task, because green fluorescent protein-MP was also observed in the vicinity of nuclei in some cells (55).

Less is known about the viral DNA participating in the transport process. Microinjection experiments employing Bean dwarf mosaic virus (BDMV) (29) provided evidence that NSP shuttles plasmid DNA between the nucleus and the cytoplasm and that MP transports dsDNA, but no ssDNA, to the neighboring cell. Consistent with these findings, Rojas et al. (35) have concluded from gel mobility shift assays that ssDNA and dsDNA are bound in a form- and size-selective manner by NSP as well as by MP. Upon these findings, a “relay race model” of transport was proposed in which NSP transfers viral dsDNA from the nucleus to the cytoplasm, from which it is delivered to MP for plasmodesmata crossing. Further support for DNA size selection during cell-to-cell transport has been provided by recent infection experiments (17).

A second model has been established which may be called the “couple-skating model.” Applying DNA-cellulose chromatography and protein overlay techniques for a distinct begomovirus (Squash leaf curl virus [SLCV]), Pascal et al. (31) observed stronger binding of NSP to ssDNA than to dsDNA but only a weak interaction of MP with ssDNA and no interaction with dsDNA. These authors favor a transport model in which viral ssDNA is shuttled between the nucleus and the cytoplasm in an NSP-containing complex, which then interacts with MP to move from cell to cell. Rojas et al. (35) tried to explain the differing results which led to the relay race and the couple-skating models with the tissue tropism of the particular virus under investigation: whereas BDMV is able to spread into all leaf tissues, SLCV is phloem limited. Alternatively, the apparent discrepancy in the interpretations of the transport processes may have resulted from the different techniques used.

To resolve this debate, we initiated gel mobility shift experiments with a phloem-limited begomovirus (AbMV) and followed the technical approach of Rojas et al. (35) as closely as possible. In general, we found the same results for AbMV as for BDMV, showing that viral tissue tropism is not the essential factor. Scrutinizing the experimental approach, however, led us to a different, more cautious interpretation which is closer to the couple-skating model.

MATERIALS AND METHODS

Protein expression and purification.

AbMV proteins CP, NSP, and MP were expressed in Escherichia coli BL21(DE3) and purified by differential centrifugation, urea washing, and sodium dodecyl sulfate (SDS)-polyacrylamide gel electrophoresis as described previously (53). Protein-containing bands were excised from the gels and electroeluted. Buffer exchange against 0.2 M NaHCO₃ and 0.02% SDS (pH 8.1) and concentration was carried out by ultrafiltration (Centricon 10; Amicon/Millipore) (53), yielding stock solutions of 11 μg of CP/μl, 3.5 μg of NSP/μl, and 4 μg of MP/μl.

Purification of dsDNA.

Phagemid pBluescript II KS(+) or pBluescript-AbA 1.0V (full-length AbMV DNA A PstI clone [14] with the orientation to produce viral sense ssDNA from the F1 origin) was purified by cesium chloride-ethidium bromide centrifugation from E. coli JM 83 according to standard protocols (37). The plasmid was linearized with BamHI or the DNA A fragment was released with PstI and BglI before separation on 1% agarose gels and elution of the DNA with the GFX-Kit (Pharmacia). Sheared fish sperm DNA (molecular biology grade) was purchased from Boehringer Mannheim. Marker DNA was obtained by digesting λ DNA with HindIII or with HindIII and EcoRI.

Preparation of ssDNA.

pBluescript-AbA 1.0V was used to transform E. coli XL1-Blue MRF′ (Stratagene) by using standard protocols (37). For replication and packaging of ssDNA, helper phage R408 was inoculated with a multiplicity of infection of 10 according to the Promega guidelines (1996 protocols and applications guide, Promega, Madison, Wis.). Infected cells were grown under rotation (180 rpm) overnight at 37°C. This longer incubation time was used to accumulate more recombined and deleted ssDNA molecules for the test system (37). Bacterial cells were pelleted by centrifugation (15 min at 12,000 × g at room temperature in a Sorvall centrifuge). The supernatant was digested by DNase I (2 U/ml) and RNase A (10 μg/ml) for 15 min at 37°C. Phages were precipitated with polyethylene glycol (5%), incubated on ice for 60 min, and centrifuged (15 min at 12,000 × g). The pellet was resuspended in 400 μl of buffer (10 mM Tris-HCl [pH 8.0], 1 mM EDTA) and extracted with phenol and chloroform. DNA was precipitated with ethanol and stored at −20°C.

Protein-DNA binding assays.

Cytochrome c (from horse heart; catalog no. 9007-43-6; Sigma) was used for comparison. Binding buffer as described in reference 35 (10 mM Tris-HCl [pH 8.0], 100 mM NaCl, 1 mM EDTA, 10% glycerol, 1 mM phenylmethylsulfonyl fluoride, 1 mM dithiothreitol) was supplied with cytochrome c, CP, NSP, or MP and DNA in various concentrations as indicated in the figure legends. To this aim, protein stock solutions were diluted to 1 μg/μl with binding buffer, mixed with DNA and further binding buffer to obtain the indicated concentrations in a reaction volume of 20 μl, and incubated for 30 min at room temperature. The whole sample was loaded into gel slots, giving the indicated concentrations of proteins and nucleic acids per lane.

One- and two-dimensional gel electrophoresis.

One-dimensional gel electrophoresis was performed in 0.7% agarose in 1× Tris-borate-EDTA containing 0.5 μg of ethidium bromide/ml (21). Samples were run at 5 V/cm for 1 to 2 h. The same gel system was used as the first dimension for two-dimensional analysis but run at 0.5 V/cm for 22 h. For the second dimension, samples were run perpendicular to the former direction in 1.4% agarose, 1× Tris-borate-EDTA, and 50 μg of chloroquine/ml with application of 45 V for 19 h. DNA was blotted onto Hybond N⁺ membrane (Amersham) and hybridized with a digoxigenin-labeled pBluescript probe as described previously (21).

Western blotting using agarose gels.

One-dimensional agarose gels were run as described above, results were documented by photography under UV illumination, and DNA-protein was blotted onto nitrocellulose membrane as for Southern transfers. Membranes were analyzed using antibodies against MP as described previously (53).

DNA-protein complex spreading and electron microscopy.

The spreading technique was performed as described in reference 8. Ten microliters of in vitro binding assay mixture was diluted with 90 μl of binding buffer and supplemented with 2.5 μl of 8% glutaraldehyde solution to fix the DNA-protein complexes for 10 min at 37°C. Ethidium bromide solution (12.5 μl; 1 mg/ml) was added, and 40-μl drops were left on Parafilm at room temperature for 10 min. Parlodium-carbon-coated grids were attached to the surface of these drops and to further drops of water, uranyl acetate, and ethanol for 30 s each. After drying at room temperature, evaporation from the grids was carried out with 2 nm of platinum (1.6 kV/60 mA; 8° angle) and the grids were analyzed by transmission electron microscopy as described previously (21).

RESULTS

AbMV CP, NSP, and MP proteins were expressed in E. coli, purified using differential sedimentation of inclusion bodies, limited urea washing, SDS gel electrophoresis, and electroelution (53), and renatured by dilution in binding buffer as described in reference 35. These proteins, in various concentrations, were incubated with different DNA samples of either viral or nonviral origin and analyzed by gel mobility shift assays as described in reference 35. Since CP and NSP are highly basic proteins, all assays included cytochrome c, which is known to bind DNA due to its positive charge although it is not a proper DNA-binding protein. This additional control served to estimate the contribution of nonspecific interference with the electrophoretic behavior of DNA in the particular gel system. The following results underscored the importance of this control and unraveled the narrow gap between true binding and electrophoretic interference.

For linear dsDNA of plasmid including full-length AbMV DNA A (Fig. 1a), for excised AbMV DNA A (Fig. 1b), for vector (Fig. 1c), and for fish sperm DNA (Fig. 1e), as well as for undigested plasmid DNA (Fig. 1d), CP (AV1) and NSP (BV1) shifted the DNA in a concentration-dependent manner. Remarkably, the shift was more pronounced for lower than for higher concentrations of the proteins. This is comprehensible if few bound protein molecules initially reduce the netto charge of an extended complex and DNA is then compacted into a smaller volume with increasing amounts of protein as expected for virus-like particles (42, 43). The increasing bends of the bands with higher protein concentrations indicate reconstructions of the complexes during electrophoretic migration. In comparison, cytochrome c shifted the DNA but the retardation increased with larger amounts of protein (Fig. 1) until the DNA was ultimately moved to the cathode and was lost from the gel image. Interestingly, the migration behaviors of monomeric plasmid covalently closed circular DNA (cccDNA) (Fig. 1d) and of fish sperm DNA with a size below 2 kb remained unaffected by the viral proteins but not by cytochrome c. As for BDMV (35), AbMV CP and NSP bound dsDNA in a form- and size-selective manner without any sequence preference.

FIG. 1.

Agarose gel analysis of the interaction of various DNA samples with AbMV CP (AV1), NSP (BV1), MP (BC1), and horse cytochrome c (Cytc). (a) Linearized pBluescript with inserted DNA A. (b) Linearized DNA A. (c) Linearized pBluescript. (d) Plasmid preparation of pBluescript. (e) Fish sperm DNA. Protein concentrations were 0, 1, 2, and 4 μg/lane, respectively, for the AbMV proteins and 0, 2.5, 5, and 7.5 μg/lane (a to d) or 0, 5, 10, and 20 μg/lane (e) for cytochrome c. Increasing concentrations are indicated by triangles. Conformations of DNA are indicated as linear (lin), open circular (oc), or covalently closed circular (ccc), and 1 and 2 indicate monomeric and dimeric forms, respectively. DNA concentrations were 50 ng/lane (a to c), 1.5 μg/lane (d), and 1 μg/lane (e). M, lambda HindIII fragments as size markers (in kilobase pairs). Asterisks indicate shifted bands under discussion.

A more subtle and different effect was found for AbMV MP (BC1). The migration behavior of linear dsDNA was hardly changed (Fig. 1a to c). A slight transient shift may be inferred from Fig. 1a, but this was not observed in all gels. Similarly, a small window of low fluorescence was observed at the highest concentration of MP with fish sperm DNA (Fig. 1e), and this window was qualified by the following experiments. In contrast, a transient concentration-dependent shift of certain DNA conformations was observed with undigested plasmid DNA and MP (Fig. 1d) and was reproducible in several gels. The fluorescence of the bands corresponding to dimeric cccDNA (2cccDNA) or monomeric open circular DNA (1ocDNA) and dimeric open circular DNA (2ocDNA) was diffuse with 1 and 2 μg of protein/lane but reappeared as sharp bands at the highest concentrations. Although this result together with some influence of AbMV MP on the migration of fish sperm DNA (Fig. 1e) may be interpreted to support DNA size selection of ABMV MP, further experiments described below showed that this conclusion is not reliable.

To monitor the interaction of the AbMV MP with ssDNA (Fig. 2), we used a helper phage to produce ssDNA from a construct of pBluescript and AbMV DNA A in the viral sense. We took advantage of the fact that, upon longer multiplication of phages, recombinant and deleted DNA molecules accumulate (37). Plasmids with inserted DNA A were found to be more prone to such recombination and deletion than the vector alone, possibly because the geminiviral plus strand origin of replication was active to a certain extent in bacteria, as has been reported for other geminiviruses (41). In consequence, a ladder of ssDNA, circular and linear forms, was produced. To make sure that only ssDNA was monitored in this assay, phages were polished with DNase I before DNA release and the resulting preparation was checked by S1 nuclease digestion (data not shown). DNA corresponding to a major band comigrated with viral monomeric or slightly smaller ssDNA (Fig. 2) as far as can be deduced from nondenaturing gels by comparison with dsDNA size standards (Fig. 2) and pBluescript ssDNA (data not shown). The DNA corresponding to this band consisted of plasmid and DNA A sequences since it hybridized with a pBluescript-specific probe (Fig. 2) and DNA A (data not shown). The migration behavior of this small DNA was not affected by CP, NSP, or MP (Fig. 2a to c). On the contrary, a window of retardation of DNA was exhibited for CP and NSP and, to a lesser extent, MP, showing the characteristic retardation and reacceleration of the DNA with increasing protein concentrations (Fig. 2).

FIG. 2.

Southern blot of helper phage-produced ssDNA (pBlue with inserted DNA A; 50 ng/lane) in the presence of AbMV CP (AV1) (a), NSP (BV1) (b), and MP (BC1) (c) with 0, 1, 2, 4, and 6 μg/lane, respectively, from left to right. Agarose gels were run as those described in the legend to Fig. 1, which did not allow an exact size reference with double-stranded markers, but the lowest band (1ss) comigrated roughly with monomeric ssDNA of AbMV. Hybridization was done with a pBluescript-specific probe. For comparison, part of the ethidium bromide-stained source gel with the fifth lane of panel c and λ markers (in kilobase pairs) is shown at the same magnification. Asterisks indicate shifted bands under discussion.

To summarize the results of electrophoretic assays with dsDNA and ssDNA, it appears that DNA molecules running ahead of the λ standard reference band of 2 kb are not affected in migration by the viral proteins in contrast to cytochrome c. Whereas the retardation effect of AbMV CP and NSP is very prominent for all DNA partners, it is weak and transient for some plasmid forms, fish sperm DNA, and ssDNA with AbMV MP.

Because of the peculiar behavior of AbMV MP, its interaction with plasmid DNA was analyzed in greater detail (Fig. 3). A two-dimensional gel was used to decide whether the MP-retarded DNA was predominantly 1ocDNA or 2cccDNA, which comigrate under the chosen gel conditions. Samples in the absence and presence of MP were run on a first agarose gel (as that in Fig. 1) but with lower voltage to better resolve the two comigrating plasmid bands (Fig. 3a). Subsequently, the DNA was run perpendicularly on a second gel in the presence of chloroquine, which intercalates into dsDNA and thereby resolves the topoisomers of cccDNA (Fig. 3b and c). For a detailed description and interpretation of two-dimensional gels, see references 21 and 32. The double band in the first dimension (Fig. 3a) was well separated in the second dimension (Fig. 3b). Since the upper band exhibited a series of topoisomers and the lower band exhibited a typical oval spot, it can be concluded that the upper band results from dimeric supercoiled DNA and not from ocDNA.

FIG. 3.

One-dimensional (a) and two-dimensional (b and c) gel analysis of plasmid DNA (pBluescript; 1.8 μg) in the absence (−) or presence (+) of 1 μg of MP (BC1). M, size marker as shown in Fig. 1. One-dimensional gels were run as those described in the legend to Fig. 1 but with lower voltage and longer time, and two-dimensional gels were run with the same first dimension (1st) as in panel a and a second dimension (2nd) in the presence of 50 μg of chloroquine/ml to resolve the topoisomers of supercoiled DNA. DNA conformations are indicated as in Fig. 1. Note that the dimeric supercoiled form runs slightly more slowly than the open circular form in the first dimension (b) and that the migration of both forms is influenced by the presence of MP (c). 2xoc, dimeric open circular; 1ccc, monomeric covalently closed circular.

In most gels examined so far (Fig. 1d and 4, lane 7, and data not shown), MP retarded 2cccDNA preferentially in comparison with 1ocDNA. In some gels, however (Fig. 3a and 4, lanes 8 to 10), the migration of 1ocDNA was also changed. Moreover, MP selectively retarded additional plasmid DNA, presumably from replicative intermediates, in the center of the gel in Fig. 3b. This retardation is represented by the smear between 2cccDNA and 1ocDNA forms as indicated in Fig. 3c. By comparison, the electrophoretic behavior of 1cccDNA, 2ocDNA, and other multimeric forms of plasmids was not changed in the presence of MP.

FIG. 4.

Comparison of the migration behaviors of plasmid DNA (pBluescript; ethidium bromide-stained agarose gel) (a) and MP protein (Western blot of the above-mentioned gel developed with anti-AbMV BC1 antibody) (b). Lanes 1 and 2: 1 μg of MP (BC1) was incubated in the presence of SDS at 65 and 95°C, respectively, before loading. Lanes 3 to 5: 1, 2, or 3 μg of MP (BC1) was loaded without plasmid DNA and SDS. Lanes 6 to 10: 0, 1, 1.5, 2, or 3 μg of MP (BC1), respectively, was combined with 1.8 μg of plasmid DNA. M, size marker as shown in Fig. 1. Brackets in lanes 7 are of the same size for comparison of the migration distance of the protein in panel b with that of the corresponding DNA in panel a. +, presence; −, absence. DNA conformations are indicated as in Fig. 1.

Before drawing final conclusions from the electrophoretic assays, we noticed that the sizes of the DNA retarded by MP varied with different electrophoretic conditions. Thus, we suspected that electrophoretic interference might play a role and wanted to scrutinize this possible artifact. We reasoned that electrophoretic interference would retard the movement of DNA but not necessarily that of protein on account of the different charges and sizes of protein and DNA. On the other hand, a true complex should be recognized by the comigration of protein and DNA. Therefore, MP was analyzed in the absence and presence of plasmid DNA on agarose gels as before but was additionally blotted onto nitrocellulose to be detected by a specific antibody (Fig. 4). For comparison, SDS-denatured MP was run and blotted in parallel. Most importantly, MP migrated within the previously detected fluorescence window (Fig. 4b), irrespective of whether DNA was present or not. This finding lends additional weight to the suspicion that the nonfluorescing window may be caused by electrophoretic interference. Otherwise the protein should also be retarded and comigrate with the shifted DNA. The only exception to this general observation was a minor amount of MP which appeared as a tiny protein band (Fig. 4b, lanes 7 and 8) comigrating with the position of 2cccDNA ahead of the free MP protein's front (compare Fig. 4a and b). Because the ethidium stain (Fig. 4a) is less sensitive than Western blotting (Fig. 4b), the respective band of 2cccDNA is seen only in the Western blot but with the lowest protein concentrations applied (1 and 1.5 μg of protein versus 1.8 μg of DNA) (Fig. 4b, lanes 7 and 8). Since no protein comigrated with ocDNA, we attribute its retardation to electrophoretic interference which may be indirectly enhanced by the formation of a complex of MP and 2cccDNA. With this explanation, it is comprehensible why the results of the electrophoretic assays depended on the voltage and the running time (compare Fig. 1d, 3a, and 4a; also data not shown).

To avoid the conflicts presented by electrophoretic interference, complex formation was visualized by electron microscopy (Fig. 5). Plasmid DNA was incubated with NSP, MP, or both and spread directly onto the drop surface (21). Without the addition of cytochrome c as a spreading agent, as it is used in the classical Kleinschmidt technique (6), naked DNA is poorly resolved in the electron microscope at an acceleration voltage of 60 kV compared to 40 kV (21). Moreover, the tested proteins contribute to the background grains according to whether they are free or bound to DNA. Under these conditions, samples with NSP and DNA exhibited mainly collapsed amorphous bodies and only a few circular DNA molecules (Fig. 5a). The presence of MP resulted in pronounced spreading of DNA with a relaxed appearance (Fig. 5b). The combination of NSP and MP produced conspicuous beads-on-a-string structures of open extended rings harboring characteristic bubbles at one site of the complex (Fig. 5c and d). Such a type of structure is expected if supercoiled DNA was packaged by NSP and MP proteins and concomitantly unwound so that part of the dsDNA had to be molten to compensate for the topological stress.

FIG. 5.

Electron micrographs of plasmids (pBluescript) spread after incubation with AbMV NSP (a), MP (b), or NSP and MP (c and d). The samples were positively stained with uranyl acetate and shaded with platinum-carbon as described previously (21). Note the beads-on-a-string appearance of the complex in panels c and d. Typically the complexes open a bubble at one site (arrows). Bar represents 1 μm.

DISCUSSION

In summary, we conclude that size and form selection is probable for CP and NSP of AbMV as shown for those of other geminiviruses before. In the case of AbMV MP, we can detect only a very low level of form-specific binding to 2cccDNA. However, electron microscopy has provided the first evidence for formation of complexes of NSP, MP, and DNA, consistent with the couple-skating model. The results confirm the previously identified role of CP and NSP as sequence-independent DNA-binding proteins (Fig. 1 and 2) and resolve a discrepancy in the literature concerning the DNA-binding capacity of MP (Fig. 3 and 4).

In contrast to the basic cytochrome c as a control for a nonspecific and biologically nonfunctional DNA-binding protein, CP and NSP did not change the migration behavior of 1cccDNA of plasmids (Fig. 1d), of linear fish sperm dsDNA smaller than 2 kb (Fig. 1e), and of monomeric ssDNA of viral genome size (Fig. 2). Similar results were previously reported for BDMV and dsDNA and were interpreted as indicative of the DNA size selection capability of CP and NSP (35). In the case of ssDNA binding, Rojas et al. analyzed only M13 DNA as a binding partner. M13 DNA is more than twice as large as geminivirus ssDNA, so a direct comparison with our results is impossible.

The in vitro DNA-binding results obtained for BDMV CP and NSP have been reproduced here for the proteins of the phloem-limited AbMV. We do not know whether the proposed size selection capability of NSP is biologically functional until the transport complex has been isolated from infected plants. On the other hand, CP must be able to package viral ssDNA of genomic and subgenomic sizes to form twinned particles or half-size solitary particles (12, 56). These ssDNA samples were obviously excluded from the protein binding in the gel mobility shift assay (Fig. 2a). Several explanations could hold for the failure of CP to bind genome-size ssDNA. Either there is still some, until now undetected, sequence-specific CP binding, or CP protein may be posttranslationally modified in plant cells in order to fulfill its primary biological task. Finally, a gel mobility-based detection system may be inappropriate for monitoring such an interaction on account of its electrophoretic properties. DNA of higher velocity may be removed from complexes by electrical forces during the run in agarose. The last argument is strengthened by the observation that all DNA samples, irrespective of their nature, which migrated faster than 2-kb λ standard fragments remained unaffected.

Sequence-independent formation of complexes of CP and NSP with larger ssDNA and dsDNA is substantiated by the reversion of electrophoretic mobility with increasing protein concentrations. In an approach different from that of Rojas et al. (35), who used His-tagged proteins in most experiments, we have employed proteins lacking fusion domains throughout the study. The results, however, were essentially the same, confirming that the proteins themselves, and not the fusion partner, were responsible for the binding activity. For MP, we rely only upon a form-specific transient interaction with 2cccDNA (Fig. 2d, 3, and 4). This interaction was verified by the results of the reciprocal assay, the change in mobility of the protein (Fig. 4b). In contrast, all indications for size-selective DNA binding of MP have to be judged with caution. Due to the coincidental migration behavior of MP and DNA in the presumed size window, electrophoretic interference cannot be excluded as the main cause for mobility changes.

A very low binding affinity of MP for ssDNA, but not for dsDNA, was found for SLCV by using chromatographic techniques (31). Interaction with 2cccDNA, as found in our study, is not necessarily proof per se of dsDNA-binding capacity. Under topological stress, dsDNA may be molten to expose ssDNA stretches. Correspondingly, cccDNA becomes accessible for the ssDNA-specific S1 nucleases as has been shown for geminiviruses (45). Interaction of MP with such regions of dsDNA molecules may change the topology of the circular DNA and thereby its mobility in gels (Fig. 1, 3, and 4) as well as its appearance in electron micrographs (Fig. 5b to d). The presence of bubbles in the circular complexes (Fig. 5c and d) is the most suggestive representation of such type of interaction.

Until now, indirect evidence has suggested that begomovirus movement proteins NSP and MP cooperate in cell-to-cell transport. Wild-type SLCV MP redirected NSP from the nucleus to the cell periphery in protoplasts (38-40), whereas specific point mutations in MP (39) or NSP (38) prevented this migration. Similarly, in plant tissues, AbMV MP redirected NSP to the cell periphery and into the next cell's nucleus (55). Further neighbor cells received NSP only if the NSP-reporter construct was coinoculated with fully infectious viral DNA. This result is up to now the strongest available evidence in a biological context that NSP is transported intercellularly. The cooperation of MP and NSP within the cell seems to be independent of the host species, because the same redirection of NSP has been observed in fission yeast upon ectopical expression of unfused AbMV MP and NSP (unpublished data). To study localization of MP at the cell periphery in greater detail, freeze-fracture immunolabeling was carried out in yeast and showed an intimate association of this protein with the protoplasmic leaflets of microsomes and plasma membranes (1). The membrane-binding domain of AbMV MP has been delimited (54) and may serve as a membrane anchor for NSP or NSP-DNA complexes. Here, we have presented the first direct evidence of a physical interaction of MP, NSP, and DNA by using electron microscopy. The beads-on-a-string appearance of the complexes differs from that of viral chromatin as it has been visualized previously (2), but the regular organization of the structures in Fig. 5c and 5d resembles linear structures, which were isolated from begomovirus-infected Sida micrantha plants in the early days of geminivirus research (22). The excellent spreading of the MP-NSP-DNA complexes for electron microscopy may be attributed to the hydrophobic portion of MP, which may orient the otherwise collapsed NSP-DNA complex to the drop surface and provide improved binding to the mainly hydrophobic carbon layer of the grid film. Perhaps this property of MP also allows spreading of NSP-DNA complexes at the surface of cellular membranes, which could be suitable for threading of the complexes through small pores such as plasmodesmata.

In conclusion, we prefer the couple-skating model of geminiviral transport as described in reference 25 in which viral DNA (whether single or double stranded, circular or linear remains to be shown) is packaged by NSP, attached to membranes by MP, and transported along the membranes through plasmodesmata to the next cell.