|Home | About | Journals | Submit | Contact Us | Français|
It is well established that poxviruses are subjected to genetic recombination, but attempts to map vaccinia virus genes using classical genetic crosses were historically confounded by high levels of experimental noise and a poor correlation between physical and genetic map distances. These virus-by-virus crosses also never produced the 50% recombinant progeny that should be seen in experiments involving distant markers. Poxviruses replicate in membrane-wrapped cytoplasmic structures called virosomes (or factories) and we have developed a method for tracking the development of these structures using live cell imaging and cells expressing phage lambda Cro protein fused to enhanced green fluorescent protein (EGFP). The EGFP-cro protein binds nonspecifically to DNA and permits live cell imaging of developing vaccinia virus factories. Using this method, we see virosomes first appearing about 4 to 5 h postinfection. The early virosomes exhibit a compact appearance and then, after a period of exponential growth lasting several hours, blur and start to dissipate in a process presumably linked to viral packaging. During the growth period, the virosomes migrate toward the nuclear periphery while colliding and fusing at a rate dependent upon the numbers of infecting particles. However, even at high multiplicities of infection (10 PFU/cell), we estimate ~20% of the virosomes never fuse. We have also used fluorescence in situ hybridization (FISH) methods to study virosomes formed by the fusion of viruses carrying different gene markers. FISH showed that DNA mixes rather poorly within fused virosomes and the amount of mixing is inversely dependent on the time between virosome appearance and fusion. Our studies suggest that the intracellular movement and mixing of virosomes create constraints that reduce opportunities for forming recombinants and that these phenomena create outcomes reflected in classical poxvirus genetics.
Genetic recombination catalyzes the acquisition of new traits and plays a key role driving virus adaption to new hosts and ecological niches. It can also promote the reassortment of antigenic determinants and facilitate DNA repair and thus aids virus survival in response to evolving immune surveillance and other environmental hazards. Poxviruses provide several classic illustrations of the impact of recombination on viral evolution. For example, the attenuated South American form of variola virus (alastrim or variola minor) is probably a hybrid virus derived from recombination between the more virulent West African and Asian variola strains (10). Malignant rabbit virus is another example of a recombinant virus found in a facility propagating myxoma and Shope fibroma viruses (2). Genome analysis suggests that poxviruses can also slowly accrete homologs of host genes over long periods of evolutionary time. This process selects for acquisition of virulence factors, and each of the poxvirus genera carries a characteristic complement of such genes (23).
Fenner provided the first experimental demonstration of vaccinia virus (VAC) recombination in culture (12), and soon thereafter Dumbell and Bedson produced recombinants between different orthopoxviruses such as variola and rabbitpox viruses (1). Using a variety of methods, we and others have shown that poxvirus recombination uses a simple form of single-strand annealing reaction intimately linked to virus replication (4, 13, 15, 34). An interesting feature of these reactions is that poxviruses can catalyze very high frequency recombination between genetic markers on cotransfected DNAs. Such “four-factor crosses” have shown that among DNAs selected for having undergone at least one intermolecular recombination event between distantly spaced flanking markers, linkage is lost between internal markers when the spacing starts to exceed ~600 nucleotides (24).
The high frequencies of recombination that can be measured using DNAs transfected into poxvirus-infected cells are not consistent with the observation that classical virus-by-virus crosses rarely generate the hypothetical limit of 50% recombinant viruses. This is best illustrated by experiments conducted in the Ensinger and Condit laboratories (7-9, 11, 28). These groups had produced stocks of temperature-sensitive VAC strains and were using marker rescue methods and complementation studies to position the different mutations on the early VAC restriction map. As detailed in the Discussion, these virus-by-virus crosses produced only ~25% recombinant progeny despite using markers spaced up to ~80 kbp apart. In fact, classical virus-by-virus crosses never proved a useful way of mapping poxvirus genes. The best maps were first constructed using marker rescue methods (i.e., testing for reversion of a mutation using transfected restriction fragments) (31) and then later updated using DNA sequencing technologies. It has never been explained what feature of virus biology precludes mapping mutations in this manner or why recombinants are not recovered in the expected numbers.
Classical bacteriophage mating theory provides insights into what factors might limit production of virus recombinants. A key concept is the “mating room,” a place where two or more genomes can mix in a way that permits recombination (29). At its simplest, the cell is a mating room and high multiplicities of infection are needed to ensure coinfection with two or more virus genotypes. This is because any virus that replicates in isolation from virus of another genotype will contribute only the parental class of virus to the pool of progeny and will thus reduce the apparent overall recombination frequency. However, at the multiplicities of infection commonly used in virus crosses (~10 PFU per cell), this effect will reduce the 50% limit on recombination by only a few percent. The fact that few poxvirus crosses generate recombinant frequencies in excess of ~25% shows that some other constraint(s) limits cooccupancy of a “mating room.” What these constraints might be is not clear.
Poxvirus replication and virion assembly take place in cytoplasmic structures called “factories” or “virosomes” (3, 5). Each virosome likely derives from a single infecting particle and seems to be bounded by a membrane possibly derived from the endoplasmic reticulum (22, 30). Confocal microscopy has been used to detect virosomal substructures, specifically DNA-free channels, and the pattern of proteins expressed from each virosome suggests that most are composed of DNA encompassing a single virus genotype (17). This is consistent with the observation that few mixing events are seen in cells bearing more than one virosome and followed using static (17) or live cell imaging (27). In this communication, we examine in greater detail the extent of interaction between virosomes in coinfected cells and show that the intracellular milieu creates constraints that limit the fusion of coinfecting virus particles and the mixing of different viral DNAs. Virosome fusion would seem to be a logical prerequisite for mixing DNA prior to recombination, and thus these physical constraints on virus-virus interaction can quantitatively explain the long-standing mystery regarding the shortfall in the production of recombinant poxviruses. They also suggest that replicating poxviruses are subject to a previously unrecognized form of “purifying selection” that may help maintain the genetic diversity and integrity of virus populations.
Vaccinia virus (VAC) carrying the T7 RNA polymerase gene inserted in the thymidine kinase locus (vTF 7.5) was obtained from P. Traktman. VAC carrying a β-galactosidase gene inserted in the thymidine kinase (TK) locus was constructed using standard transfection and selection methods and using a lacZ gene excised from plasmid pSC66 (VAC TK::lacZ). All viruses were grown and titers determined on BSC-40 cells in modified Eagle's medium supplemented with nonessential amino acids, l-glutamine, antibiotics, and antimycotics (all from Gibco) plus 10% fetal calf serum (Sigma). Cells were tested and shown free of mycoplasma.
We also produced a reporter cell line constitutively expressing the bacteriophage λ cro repressor fused to enhanced-green fluorescent protein (EGFP). The cro gene was amplified from a phage λ DNA template using the PCR and two primers (forward, 5′ AAGCTTGTATGGAACAACGCATAACCCTGAAAG 3′; reverse, 5′ GGATCCTATTATGCTGTTGTTTTTTTGTTACTCGGGA 3′) and cloned into a Topo PCR 2.1 vector (Invitrogen). The DNA was excised using HindIII and BamHI and recloned into pEGFP-C1 (Clontech). This creates a gene encoding the λ cro protein (66 amino acids) fused to the C terminus of EGFP. The plasmid was transfected into BSC-40 cells using Lipofectamine 2000 (Invitrogen), and recombinants were recovered using G418 selection for neomycin resistance. Individual clones were isolated, and one designated Cro-2 C16 was selected for use in this project based upon a uniform pattern of expression of mostly nuclear green fluorescence. BSC-40 cells were also separately transfected with pEGFP-C1 to produce cloned control cell lines expressing unmodified EGFP.
The PCR and two sets of primer pairs were to amplify the T7 RNA polymerase (5′ ATGAACACGATTAACATCGCTAA 3′ and 5′ TTACGCGAACGCGAAGTCC 3′) and lacZ (5′ CTCGAGGAATGGGAGATCCCGTCGTTTTAC 3′ and 5′ AAGCTTGCGGCCGCTCAGCTGAATTCCGCCGATACTGAC 3′) genes using vTF 7.5 and pSC66 templates, respectively, and the PCR products were then cloned into Topo PCR 2.1 vectors. ARES kits (Alexa Fluor 488 for LacZ plasmids and Alexa Fluor 594 for T7 RNA polymerase plasmids) from Molecular Probes and nick translation were used to prepare DNA probes for fluorescence in situ hybridization (FISH) analyses. The labeled DNAs were purified using Qiaquick columns and quantified using a Nanodrop spectrophotometer.
For FISH, the VAC-infected Cro-2 C16 cells were fixed for 30 min with ice-cold 4% paraformaldehyde in phosphate-buffered saline (PBS) or overnight at 4°C. The samples and probes were denatured simultaneously at 95°C in hybridization buffer for 5 min in a water bath, chilled in an ice-water bath for 30 s, and then hybridized overnight at 42°C. The hybridization buffer contained 50% (vol/vol) deionized formamide, 2× SSC (300 mM NaCl, 30 mM sodium citrate), 50 mM sodium phosphate (pH 7.4), 1 mM EDTA, 400 μg/ml salmon sperm DNA, 1× Denhardt's solution (Invitrogen), 5% (wt/vol) dextran sulfate, 0.05% (wt/vol) sodium dodecyl sulfate, and 1 to 10 μg/ml fluorescent probe. The samples were washed twice with 50% formamide in 2× SSC at 42°C for 30 min each and then twice with 0.2× SSC at 55°C for another 30 min each. The samples were finally washed with PBS and counterstained with 10 ng/ml DAPI (4′,6-diamidino-2-phenylindole; Molecular Probes) in PBS.
All of the fluorescence imaging was performed using an Applied Precision DeltaVision microscope equipped with a temperature-regulated environmental chamber. For live cell imaging, the cells were cultured on optically clear glass-bottom dishes (Fluorodish from World Precision Instruments or gridded μ-dishes from Integrated BioDiagnostics) using phenol red-free modified Eagle's medium (Sigma) supplemented as described above, except that 10 mM HEPES (Gibco) replaced the bicarbonate buffer. The cells were infected with virus for 1 h in PBS at 37°C, and then the inoculum was replaced with fresh warmed growth medium. The dishes were sealed and mounted on the microscope stage, and the temperature was maintained at 37°C. At the end of the second hour, image data were collected at 5-min intervals using Resolve3D software and the fluorescein isothiocyanate (FITC) filter set for EGFP. Static imaging of DAPI, Alexa Fluor 488, and Alexa Fluor 594 reporters used the DAPI, FITC, and/or tetramethyl rhodamine isothiocyanate (TRITC) filter sets, respectively. Using a 100× objective, the pixel dimensions were x = y = 0.13 μm and z = 0.20 to 1.0 μm, depending upon the thickness of the imaged z plane. The thickness of cells varies over the course of infection, and the depth of the z plane was chosen to minimize the risk of particles moving beyond the image plane. All of the images were subsequently processed using the deconvolution algorithm built into softWoRx (v3.7).
Image data were exported as Photoshop files using softWoRx software and then assembled into composite images using Adobe Photoshop CS3 (v10.3). All of the images were subjected to the same background and scaling adjustments and used only a linear gamma factor. GraphPad Prism (v5.0a) was used to perform statistical analyses of the data. The images shown in Fig. 9C and D were created and analyzed using ImageJ (v1.42). To derive the distances shown in Fig. Fig.10,10, the map position of each mutation was determined from a description in the literature of either the exact map position (as determined by DNA sequencing), the midpoint of the gene location known from complementation analysis, or the midpoint of the restriction fragment known from marker rescue analysis. In the latter two situations, the location error was estimated as half of the distance between the boundaries of the gene or restriction fragment. The distance error in any given cross was approximated as the average of the location error for each of the two markers used in the cross.
To calculate what effect finite input would have on recombinant production, we used the formula derived by Lennox et al. (19). They demonstrate that the correction factor F(P) at any given multiplicity of infection can be calculated from the formula
where P is the sum of the average multiplicity of infection of each virus and N is the number of particles. We have not incorporated a correction for the number of virus particles that could replicate in a cell (19), as our studies use multiplicities of infection far lower than what has been reported to be achievable in culture (22).
A numerical simulation was used to determine how fusion events would be randomly distributed among coinfecting particles. Ten separate runs of 100,000 iterative simulations were performed and showed that if 5 fusion steps were distributed randomly among 10 interacting virosomes, it would leave an average of 2.20 ± 0.06 virosomes per cell unfused. A detailed description of the algorithm devised by Yin Li, Department of Mathematics and Statistics, University of Alberta, can be supplied upon request.
To follow the fate of replicating viruses, we devised a vital tagging system that uses cells constitutively expressing EGFP fused to the bacteriophage λ cro repressor protein. We had originally hoped to use the EGFP-cro DNA binding protein to selectively tag VAC modified by incorporating the cro repressor binding site. However, adding the six operator binding sites (OL1 to -3 and OR1 to -3) to the VAC genome did not significantly alter the amount of EGFP-cro labeling compared to that of an unmodified virus (data not shown). In fact, any of the virosomes formed in these cells can be labeled with the EGFP-cro protein, presumably through a nonspecific DNA binding activity (Fig. (Fig.1).1). This fluorescent label appears to be recruited from a pool of protein located predominantly in the cell nucleus and is sufficiently stable in fluorescence imaging experiments to track the fate of virus particles over time scales up to 10 h. No selective staining of the cell nucleus or of virosomes was detected in cells constitutively expressing just EGFP, showing that the cro peptide is responsible for DNA binding (data not shown). Vaccinia virus also seemed to plate equally well on EGFP- and EGFP-cro-expressing cells, suggesting that the cro-DNA binding interaction had little or no deleterious effect on virus growth.
This method provides a simple way of tracking the uncoating, movement, and growth of coinfecting VAC particles. Figure Figure22 shows a cell infected for 1 h at a high (10 PFU/cell) multiplicity of infection and where at least 7 virosomes can be seen. All of these particles “winked” into existence within a relatively short period of time (~20 min) and usually about 4 to 5 h after infecting the cells (i.e., 1 h of infection, 1 h of hold, 2 to 3 h into recording). The timing of virosome appearance varied somewhat from cell to cell, and sometimes viruses weren't seen until as late as ~7 h postinfection. We also noted a reduction in the proportion of infected cells within the imaging fields, compared with cells located in peripheral fields, suggesting that the imaging process may somewhat decrease the efficiency of establishing an infection in some fraction of cells. Because of this variability, the time of virosome appearance served as the most reproducible marker for synchronizing and comparing intracellular events. At first appearance, the size of each spot approximates the limits imposed by the pixel resolution (0.13 μm) and is thus consistent with the dimensions of a VAC particle (~0.2 μm). All of the particles also exhibited a similar initial fluorescence intensity that then increased with time. The initial doubling times ranged from 9 to 14 min for different particles over the first hour of tracking (Fig. (Fig.2).2). The number of particles detected in these experiments generally reflected the multiplicity of infection, suggesting that while the particle/PFU ratio is not known with certainty, these methods probably track the portion of viable plaque-forming virus in these stocks.
When one follows the fate of individual virosomes using live cell imaging, we see several examples of virosome fusions. Figure Figure33 shows a continuation of the image set initially tracked in Fig. Fig.2.2. In frame 70, one can see two equally bright particles, designated A and B, as well as a third particle that is distinguished from A and B by a slightly reduced fluorescence (C). Five minutes later, this arrangement has been replaced, in frame 71, by just two virosomes and one of these is now substantially brighter than any of the presumed antecedent factories. These two virosomes then continue to grow in size and brightness in the next frame (no. 72). If one integrates the intensity of these particles, it is apparent that the fluorescence of the fusion product (3,785 arbitrary units) represents more than the sum of the fluorescence exhibited by the two presumed parent particles (1,439 + 1,420 = 2,859 units). However, based upon the prior rate of growth of particles A and B (Fig. (Fig.2)2) it can be estimated that each would have increased in intensity by about 40% over a single 5-min frame. If one corrects for this fact, then the larger factory exhibits about 95% of the predicted fluorescence based upon the coalescence and continued growth (at the same rate) of two smaller particles. It should also be noted that a great many virosomes were tracked over the course of this study and if fusion was seen early in the replication cycle, then the resulting particles persisted as stable aggregates.
If one continues to monitor the behavior of these EGFP-tagged virosomes, they typically track toward the nuclear periphery while continuing to grow larger and becoming more asymmetric with time. They also start to lose the compact structure that characterizes early virosomes (Fig. (Fig.1).1). We presume that these visible changes in virosome structure reflect previously characterized transitions from an early DNA replication step in the virus life cycle to a later virus assembly and genome-packaging phase. Collisions between virosomes continued to occur throughout infection, but late in infection, these collisions rarely seemed to disrupt more mature virosomal boundaries. The question then arises regarding whether these fusion events can mix the DNAs of viruses of different genotypes. This is important for the purposes of our study, since the mixing of different coinfecting virus genomes is assumed to be an essential prequel to recombination.
We used fluorescence in situ hybridization (FISH) to investigate this question. Cro-2 cells were cultured on gridded dishes, infected with a 1:1 mixture of viruses carrying either the Escherichia coli lacZ or T7 RNA polymerase gene, and the distribution of the virus DNA was determined using EGFP fluorescence (Fig. (Fig.4,4, top panel). The dish was removed from the microscope, and the infected cells were fixed and then hybridized to a mixture of two gene-specific FISH probes. The specimens were then counterstained with DAPI and returned to the microscope, and the cells of interest were relocated using the recorded grid position. Although FISH methods cause some degradation of the image quality, it is apparent that most of the virosomes, in this example, are composed of DNAs that hybridize to just one of the two FISH probes (Fig. (Fig.4,4, bottom panel). However, one of the virosomes contained a mixture of DNAs, as illustrated by the costaining with both red and green fluorescent probes. Thus, virosome fusion can lead to the DNA mixing that is required for recombination.
Fig. Fig.55 illustrates the method used to determine how often coinfecting virus factories fuse during the process of virus replication. Cells were infected with VAC at multiplicities of infection of 2 or 10, and then a gridded field of 0.5 mm by 0.5 mm was scanned to map where the cells were located. These cells were then tracked over time, and the appearance and fate of any virosomes were followed using EGFP-cro fluorescence over the next ~6 h. Each cell was scored for the number of factories appearing over time, along with the timing and number of fusion events. For example, in this image (Fig. (Fig.5)5) at least 10 virosomes can be detected at the 1:45 time point, only 7 are seen 15 min later (2:00 h), and eventually only 4 mature virosomes are detectible within that cell (4:00 h). At the end of the experiment, each marked cell was recentered in the image field and a full set of z-stack images was recorded for each of the infected cells showing the final distribution of fused and intact factories. A composite image showing the relative grid position of each of the imaged cells was also collected. After FISH staining, this reference image was then used to relocate the subset of infected cells and was used to confirm that DNA mixing had occurred (see below).
It is important to note that this method is laborious, and it is impossible to accurately quantify all of the fusion events that might occur in cells infected with many viruses. A particular problem is that VAC-infected cells round up and move in an unpredictable manner during infection (27), and if some z planes are not captured in all of the scans over all of the time points, then some smaller virosomes could be missed and their disappearance possibly recorded as fusion. Thus, one cannot reliably track all of the fusion events as a function of virosome number. However, one can determine which fraction of infected cells show no evidence of virosome fusion by counting the number of virosomes at the beginning and end of the experiment. These data are still useful as the number of null events represents the P(0) term in the Poisson equation and can be used to determine the average number of fusion events across cells infected with different numbers of viruses. The results of this analysis are shown in Fig. Fig.66.
Figure Figure66 summarizes how often virosome fusion happens in VAC-infected cells. As might be expected, the proportion of infected cells exhibiting one or more virosome fusions goes up as the number of virosomes per cell increases. Only rarely are fusions seen in cells infected with two virosomes, and less than half of cells exhibit a fusion event with three infecting particles. However, at high multiplicities of infection, one or more fusions are almost always observed. The number approaches unity at multiplicities in excess of ~7. If one assumes that the frequency of seeing a cell enclosing a virosome fusion is a random event determined by a simple Poisson likelihood, one can estimate from the P(0) term the parameter μ, where μ is the average number of fusions per category of cell type (e.g., cells with 2 virosomes per cell). The parameter μ can be calculated from the formula
Figure Figure77 shows a plot of μ values versus the number of particles per cell. We plotted these values out to only 6 virosomes per cell, because P(0) cannot be measured accurately at higher multiplicities of infection due the rarity of such events. We observed a seemingly linear relationship between the average number of fusions per cell (μ) and the number of virosomes infecting these cells. If we extrapolate these data out to 10 particles/cell (a multiplicity of infection commonly used in recombination studies [see below]), one observes μ = 5 ± 1 fusion per cell (95% confidence interval [CI]). Collectively, the odds of no fusion occurring in cells infected with 10 viruses become vanishingly small at this multiplicity of infection. Parenthetically, these investigations also provide some insights into the timing of fusion. Although there is much variation from cell to cell, a trend is observed where fusion occurs much faster in cells containing multiple viruses (Fig. (Fig.8).8). For example, if zero is defined as the time of first appearance of the first virosome, it takes on average 65 ± 20 min (mean ± standard error [SE]) for the first fusion event to be detected in cells bearing two virosomes but only about 13 ± 2 min in cells bearing 10 virosomes.
Live cell imaging combined with FISH analysis can also provide insights into the extent of DNA mixing when virosomes do fuse. This is most conveniently illustrated by examining how the two fluorescent hybridization probes are distributed throughout the stack of pixels comprising each three-dimensional image. We used the Cro-2 cells to first track the fate of coinfecting viruses carrying lacZ and T7 RNA polymerase genes, located cells that exhibited one or more fusion events, and then used FISH to differentiate between the two infecting VAC genomes. Figure Figure9A9A shows an example of one of these FISH images along with a plot showing how the red fluorescence (617 nm) and green fluorescence (528 nm) are correlated within the bicolored image stack seen lying immediately adjacent to the cell nucleus (R = 0.43; Fig. Fig.9B).9B). The image shown in Fig. Fig.9A9A was also used to create a three-dimensional model of how the two signals were distributed within this fused virosome. These are shown in two views looking either down from the top of the image, as also seen in panel A (i.e., along the z axis, panels C1 to C3), or in the x-y plane looking toward the side of the factory facing away from the cell nucleus (Fig. 9D1 to D3). By separating the image into the two-component green and red FISH signals (Fig. 9C2 and D2 and C3 and D3, respectively) one can see that the virosome contains a complex distribution of the two hybridization targets, which overlap along the boundaries to generate the yellow signal seen in panels C1 and D1.
This analysis was extended to study the variation in the amounts of mixing following a number of different fusion events. We noted that the extent of fusion varied greatly from cell to cell, as judged by a range of correlation factors for the red-green overlap ranging from no overlap (R = 0) to R = 0.65 (Fig. (Fig.9,9, lower panel). Curiously one never sees a late virosome exhibiting a homogeneous yellow mixture of the two signals, suggesting that although fusion does occur at various points in the development of each factory, this rarely leads to complete mixing of the antecedent particles. Because each of these data points was associated with a live cell video, we could also test whether the time between the appearance of each virosome and its subsequent fusion affected the degree of mixing as determined by the spectral overlap of the two FISH probes (i.e., as shown in Fig. Fig.9B).9B). Although these data are noisy and the correlation is thus limited (r2 = 0.25), the slope of the curve is significantly negative and suggests that the longer it takes for virosomes to fuse, the less opportunity there exists over the course of the virus life cycle to mix the two viral DNAs.
We have used live cell imaging methods to track the fate of replicating vaccinia viruses. These studies show that when cells are coinfected with multiple viruses, only a subset of virosomes interact in a manner that produces stable fusions between virus factories. Furthermore, few of the fused particles exhibit an intimate mixture of different DNAs, as judged by FISH methods (Fig. (Fig.9).9). The timing, frequency, and extent of virus fusion are all affected by the multiplicity of infection, but even in cells infected by large numbers of viruses, a portion of the virosomes apparently avoid fusion. The fact that coinfecting viruses interact poorly, or not at all, within the intracellular milieu can explain why too few recombinant viruses are recovered from classical genetic crosses.
The method makes several simplifying assumptions, not all of which can be tested experimentally. In particular, it is not certain that virosome fusion is essential for recombination or how mixing relates to the efficiency of recombination. However, it would seem reasonable to assume that some mixing of virus DNAs is required prior to or during virus DNA replication. This is when replication-dependent virus recombination has been detected by other methods (13, 33), and the fact these virosomes are replicating while they are also fusing is illustrated by exponential growth we see in the amounts of EGFP-cro binding virus DNA (Fig. (Fig.2).2). We cannot also be certain that our methods capture every fusion event, especially early ones, which could be obscured by the background from nuclear EGFP-cro or because fusion happens at a time preceding or coincident with the appearance of uncoated viruses. However, the fact that viruses typically appear with a similar initial fluorescence (Fig. (Fig.2)2) suggests that there aren't many early fusion events missed by our approach. We are also assuming that each of these early EGFP-cro-labeled particles represents a single virus captured at the point where uncoating makes the viral DNA accessible to EGFP-cro protein. This is consistent with the dimensions of the signal, but our methods couldn't differentiate two particles fused within a space smaller than the limits of the microscope's resolution. Finally, in modeling these events, we are assuming that they are all randomly driven processes unbiased by the history or genetic properties of particular virosomes.
The EGFP-cro fusion protein offers many advantages for live cell imaging of poxviruses. We had hoped that cro's high affinity for repressor binding sites could be used in combination with a virus encoding multiple binding sites (OL1 to -3 and OR1 to -3), to provide a tool for selectively labeling VAC encoding these sites. We produced such a virus, but although the cro-operator dissociation constants (Kds) are at least as favorable as for the LacI-based systems that have previously been used to tag replicating poxviruses (6, 14, 18), this strategy still cannot provide sufficient selectivity to clearly differentiate viruses encoding six operator binding sites from wild-type viruses. In hindsight, the ~200-fold difference in Cro's affinity for operator-containing versus nonspecific DNA (6) is probably insufficient to create the specificity we need. However, the reporter still serves as an excellent marker for virus DNA and thus provides a convenient tool for live cell imaging of replicating viruses. Although the EGFP-cro fusion protein has no nuclear transport signal, it is smaller than the nuclear pore limits and much of the protein is found in the nucleus in uninfected cells. It appears to traffic outward and bind to virus DNA over the course of infection. Many other infected-cell proteins are also recruited from the cell nucleus by poxviruses—e.g., barrier to autointegration factor (BAF) protein (32) and topoisomerase II (20)—but whether by an active or passive transport process remains unclear. The gradual loss of fluorescence over time occurs selectively in infected cells and presumably reflects a combination of protein turnover and virus destabilization of host cell mRNAs (25, 26). The rate of increase in virosomal fluorescence will be determined by several factors (e.g., EGFP-cro turnover, rates of recruitment to virosomes, and rates of DNA synthesis) and probably cannot be directly correlated with rates of viral replication. However, the initial estimates of doubling time as deduced from the rate of fluorescence increase (Fig. (Fig.2;2; 9 to 14 min) do approximate the rates that can be measured using Southern blotting for virus DNA (21).
These methods clearly show that some coinfecting virus particles never seem to mix, or mix only poorly, and this can partially explain the aforementioned shortfall in recombinant virus production. Figure Figure1010 shows a meta-analysis of the classical VAC intergenic cross data. These data were culled from the published literature and updated (where possible) to include estimates of the marker positions on the VAC genome. These data represent many different crosses between temperature-sensitive markers and span distances up to about one-third of the VAC genome (8, 9, 11, 16, 28). Two features of these data are immediately obvious. The first is the substantial experimental scatter, and the second is the difficult to define, but clearly not 50%, recombination limit. We suggest that the experimental scatter is actually an intrinsic feature of virus biology and can be explained by variation in the timing of fusion (Fig. (Fig.8)8) combined with variable degrees of mixing within fused virions (Fig. (Fig.9).9). We also suggest that the ~25% recombination limit reflects a combination of effects caused by poor mixing of virus DNAs and by the failure of fusion of some coinfecting virosomes.
One classical explanation for why phage and virus crosses don't produce 50% recombinants arises from the manner in which genomes distribute during infection (19, 29). For example, if a cell were infected by just one of the two parental viruses (virus A or virus B), it could never produce a recombinant. Furthermore, in cells infected by two virus particles, half of the cells will receive two parental viruses (A + A or B + B) and half a mixture of both genotypes (A + B or B + A). In any such infections, replication will contribute only the two parental classes of virus to the pool of progeny and thus reduce the proportion of recombinants. However, if one tests for what effect a finite input of virus would have on Rf, it is apparent that at high multiplicities of infection, the capacity to reduce the yield of recombinants is not large enough to explain the discrepancies. For example, most of the crosses shown in Fig. Fig.1010 employed a multiplicity of infection of ~10, and under these conditions, the maximum Rf, Rf (max), would be reduced to ~45% (19).
This limit for Rf (max) would be reduced further by the failure of some virosomes to fuse, even at these high multiplicities of infection. We estimated that an average of 5 ± 1 fusion events happen over ~3 h in a cell initially bearing 10 virosomes (Fig. (Fig.77 and and8).8). If we assume these 5 events are randomly distributed among 10 coinfecting particles, a numerical simulation can be used to calculate that on average 22% of the particles will never fuse during a period spanning the onset of replication through what appears (based upon a lack of further growth and the dissolution of the virosome) to be the start of virus packaging. This 22% of the virus will replicate but can produce only parental-type progeny, while the other 78% of viruses could presumably generate up to equal numbers of parental (P) and recombinant (R) progeny if the gene markers are unlinked. If we assume that the yields of progeny virus are equal and proportional regardless of any interaction history, then Rf = R/(P + R) = 39R/(22P + 39P + 39R) = 39%. This effect would reduce the Rf (max) to 39%.
The revised prediction for Rf (max) is still higher than the 21 to 27% seen in classical crosses (Fig. (Fig.10),10), and we suggest that the limited mixing of DNA within the virosomes (Fig. (Fig.9)9) could cause a further reduction in recombinant formation. For example, in this particular case only 44% of the red (VAC vTF 7.5) volumetric pixels (voxels) also exhibit a signal in the green (VAC TK::lacZ) channel and 16% of the green voxels also exhibit a red signal. It is difficult to deduce what this measurement actually means in biological terms, because FISH methods may alter the structure of the virosome and thus there is no assurance that the distribution of the fluorescent probes necessarily reflects the original path or distribution of the DNA. However, at multiplicities of infection of 10, the time from viral appearance to fusion is only ~15 min (Fig. (Fig.8)8) and one can crudely estimate from Fig. Fig.99 (bottom panel) that the average amount of mixing would then resemble what is seen in the example image shown in Fig. Fig.99 (top panel). In this image, only ≈30% [30% = (44% + 16%)/2] of the FISH-labeled voxels contain a mix of two genomes. If the example is representative of general trends, and only 30% of the DNAs mix in a way that could produce recombinants at high multiplicities of infection, then Rf (max) would be reduced to ≈15% [15% = 15R/(70P + 15P + 15R)]. Although the magnitude of this effect is uncertain, for the reasons outlined above, the potential for also reducing recombination frequencies is clear.
In conclusion, these studies suggest that the intracellular movement and mixing of virosomes create constraints that reduce opportunities for generating recombinants and that these phenomena create outcomes reflected in classical poxvirus genetics. Perhaps more intriguingly, these observations generate new insights into factors affecting poxvirus evolution. Katsafanas and Moss have used VAC strains encoding core proteins tagged with cyan and yellow fluorescent proteins to show that some virus-encoded gene products aren't randomly distributed throughout infected cells (17). Instead, these particular late gene products are synthesized either close to, or within, the virosome presumed to encode them. Our data suggest that even in cells infected with many virus particles, a significant portion never fuse or fuse only late in the replication cycle and thus likely replicate partly in genetic isolation. These constraints on complementation and recombination would impose a previously unrecognized form of purifying selection on replicating poxviruses, which could help maintain the genetic integrity of virus populations and laboratory stocks. For example, it would prevent the accumulation of defective interfering particles since a “mutant” virosome could have difficulties recruiting complementing factors from other coinfecting particles. Thus, a virosome may be more than just a site of poxvirus replication and assembly. These observations raise the intriguing possibility that virosomes represent an adaption used by poxviruses to compete, in a Darwinian sense, with other poxviruses for intracellular resources.
We thank Nick Li and Michelle Shih for help producing the Cro-2 C16 cell line, Chad Irwin for measuring the plating efficiency of virus on Cro-2 cells, and Chad Irwin and other laboratory members for helpful comments. Yin Li (Department of Mathematics and Statistics, University of Alberta) kindly devised the algorithm used to calculate the proportion of nonfused virions in cells.
This project was funded by awards from the CIHR, NSERC, CFI, and ASRIP (to D.H.E.).
Published ahead of print on 23 December 2009.