Coalescing replication compartments provide the opportunity for recombination between coinfecting herpesviruses

Abstract

Homologous recombination (HR) is considered a major driving force of evolution because it generates and expands genetic diversity. Evidence of HR between coinfecting herpesvirus DNA genomes can be found frequently both in vitro and in clinical isolates. Each herpes simplex virus type 1 (HSV-1) replication compartment (RC) derives from a single incoming genome and maintains a specific territory within the nucleus. This raises intriguing questions about where and when coinfecting viral genomes interact. To study the spatiotemporal requirements for intergenomic recombination, we developed an assay with dual-color FISH that enables detection of HR between different pairs of coinfecting HSV-1 genomes. Our results revealed that HR increases intermingling of RCs derived from different genomes. Furthermore, inhibition of RC movement reduces the rate of HR events among coinfecting viruses. Finally, we observed correlation between nuclear size and the number of RCs per nucleus. Our findings suggest that both viral replication and recombination are subject to nuclear spatial constraints. Other DNA viruses and cellular DNA are likely to encounter similar restrictions.—Tomer, E., Cohen, E. M., Drayman, N., Afriat, A., Weitzman, M. D., Zaritsky, A., Kobiler, O. Coalescing replication compartments provide the opportunity for recombination between coinfecting herpesviruses.

Recombination is considered to be a major driving force in evolution of most organisms because it accelerates adaptation (1, 2). The architecture of eukaryotic nuclei is suggested to regulate many DNA-mediated processes, including replication, gene expression, and recombination. DNA viruses that replicate inside the nucleus clearly change the nuclear architecture; however, they are likely to be subjected to similar spatial constraints as host DNA. The rate of mutation accumulation is lower for DNA viruses than that of viruses with RNA genomes (3, 4). It has been hypothesized that high rates of recombination can facilitate genetic adaptation to the changing environment (5). Indeed, homologous recombination (HR) among coinfecting herpes simplex virus type 1 (HSV-1) genomes is very frequently observed in both in vitro genetic assays (6–12) and in sequence analysis of clinical isolates (13–15). Herpesvirus infection therefore provides a system to study spatial features that promote or constrain recombination in the eukaryotic nucleus.

Like all other herpesviruses, HSV-1 viral gene expression, replication, and capsid assembly all occur in the host nucleus of infected cells. Viral genomes enter the nucleus through the nuclear pore complex as naked DNA molecules (16), and these rapidly recruit several host and viral proteins to the viral genomes (17–25). Expression of the immediate early viral genes allows initiation of viral DNA replication (26). HSV-1 DNA replication proceeds at distinct foci within the nucleus known as replication compartments (RCs) (27, 28). The formation of the viral RCs was suggested to initiate from small pre-RCs (29, 30). Live cell imaging of viral DNA binding proteins suggested that the pre-RCs migrate toward nuclear speckles, sites of RNA processing and come into contact with other pre-RCs where they seem to coalesce into large, mature RCs (31). On the other hand, direct visualization of the viral DNA suggested that each RC usually emerges from a single incoming genome (32, 33). Our previous study with the swine alphaherpesvirus pseudorabies virus (PRV) suggested that although viral RCs are found in close proximity, they retain distinct territories for each individual genome (33). Earlier experiments with HSV replicons also supported this notion (32). A recent study showed that viral genomes entering the nucleus are observed as condensed foci and suggested that viral expression and DNA replication allow decondensation of these genomes and formation of RCs (34). Interestingly, some genomes remain highly condensed at the edge of newly developing RCs (34, 35). Here, we visualized coinfecting HSV-1 genomes and confirmed that alphaherpesviruses RCs initiate from single genomes.

Viral DNA recombination is facilitated by both viral and cellular proteins. Two viral proteins have been suggested to work as a complex to facilitate viral recombination and have been shown to catalyze strand exchange in vitro: the single-strand binding protein ICP8 and an exonuclease UL12 (36). Single-strand annealing was found to be a recombination mechanism up-regulated during viral infection and is thus considered as the mechanism by which the viral recombinase induces recombination (37). Although ICP8 is required for viral DNA replication, UL12 is not essential for DNA replication per se, although it is required for formation of infectious viral genomes that can be packaged into capsids (38). Recent observations suggest a complex relationship between HSV and the host DNA repair machinery. Some components of the DNA damage response machinery are recruited to replicating viral DNA, where they may function in ways that are beneficial for viral progeny (12, 21, 23, 39–42). In contrast, DNA damage response pathways as a whole may be considered to be antiviral and are suppressed in HSV-infected cells (19, 43–48). These findings suggest that the processes of viral recombination and viral replication are intimately associated (46). Although knowledge regarding the molecular aspects of HR has been accumulating over the last few years, little is known regarding spatiotemporal constraints on intergenomic recombination.

The compartmentalization of coinfecting genomes at different RCs raises questions about where and when recombination takes place during the course of infection. To tackle these questions, we used a FISH-based assay to differentiate between de novo synthesized variants of viral genomes. Our results suggest that multiple intergenomic recombination events occur at later stages of infection following DNA replication and that intergenomic recombination takes place at the interface between mature RCs. We also found that the number of RCs correlates with nuclear size, suggesting a possible spatial restriction to the number of viral genomes that initiate replication.

MATERIALS AND METHODS

Cell culture

African green monkey kidney cells [Vero ATCC CCL-81; American Type Culture Collection (ATCC), Manassas, VA, USA] and human female osteosarcoma cells (U2OS cells ATCC HTB-96; ATCC) were grown in DMEM (DMEM ×1; Thermo Fisher Scientific, Waltham, MA, USA), supplemented with 10% fetal bovine serum (Thermo Fisher Scientific) and 1% penicillin (10,000 U/ml) and streptomycin (10 mg/ml; Biological Industries, Beit HaEmek, Israel).

Viruses

All viral recombinants are derivatives of HSV-1 strain 17. Each viral recombinant contains 1 or 2 tag sequences for specific staining by FISH. To facilitate isolation, both tag sequences are expression constructs for fluorescent proteins. The red fluorescent protein mCherry driven by the human cytomegalovirus promoter (CMVp) and the yellow fluorescent protein YPet driven by the simian virus 40 promoter (SV40p). Tag sequences were inserted into the viral genome by HR. Viral DNA was cotransfected along with a plasmid containing the tag sequences flanked by sequence homologies to the viral site of insertion (synthetically generated by GenScript, Piscataway, NJ, USA). Recombinant viruses were isolated from the progeny by plating lysate from transfected Vero cells and picking fluorescent plaques using a Nikon (Tokyo, Japan) Eclipse Ti-E epifluorescence inverted microscope. Viral stocks were prepared by growing purified plaques for each recombinant virus on Vero cells. The viral recombinants were validated by PCR. Viral titers were measured by plaque assay. An additional viral recombinant containing both tag sequences was isolated by crossing the recombinant OK26 to the previously described OK11 (49) and selecting for plaques containing 2 fluorescent proteins by plaque assay. All viral recombinants constructed for this paper are described in Table 1.

TABLE 1.

List of viral recombinants used in this work

Viral recombinant	Tag sequence	Locus of insertion	Insertion start point
OK25	SV40p:YPET	Between the ORFs UL3 and UL4	11736
OK26	SV40p:YPET	Between the ORFs UL55 and UL56	116153
OK32	SV40p:YPET	Between the ORFs UL37 and UL38	84252
OK31	SV40p:YPET	Between the ORFs UL55 and UL56	116153
	CMVp:mCherry	Between the ORFs UL37 and UL38	84252
OK35	CMVp:mCherry	Between the ORFs UL3 and UL4	11736

Fluorescent probes

A set of 20 fluorescent probes was designed to correspond to each one of the 2 tag sequences. The probes for the CMVp:mCherry and sequences SV40p:YPet were conjugated on their 5′ end to the fluorescent dyes cyanine 3 (Cy3) and Cy5, respectively. An additional probe was designed to stain HSV1 viral DNA nonspecifically. This probe corresponds to the viral a’ sequence and conjugated to the fluorescent dye Alexa Fluor 488 (Thermo Fisher Scientific) on its 5′ end. Fluorescent probes were synthesized by Integrated DNA Technologies (IDT; Coralville, IA, USA). The probes were dissolved in TE buffer [Tris-HCl (MilliporeSigma) pH 8.1 10mM, EDTA (MilliporeSigma) 1mM] to a stock concentration of 10 μM. Probes from each set were pooled together in equal ratios and kept in −20°C before hybridization. Probes sequences are detailed in Table 2.

TABLE 2.

List of fluorescently labeled probes for FISH

Primer sequence, 5′–3′
SV40p:YPET	HSV-1 a’ sequence	CMVp:mCherry
CTTGCATCTCAATTAGTCAGCAACC	CTTAAGCCCATATATGGAGTTCCGC	CCCCCCGCTCCTCCCCCCGCT
GCTACGGCGTGCAGTGCTTCGCCAG	TAACTTACGGTAAATGGCCCGCCTG
TCCGCCCATTCTCCGCCCCATCGCT	AACGACCCCCGCCCATTGACGTCAA
ATTTATGCAGAGGCCGAGGCCGCCT	GACGTATGTTCCCATAGTAACGCCA
CCTCTGAGCTATTCCAGAAGTAGTGAGG	GACTTTCCATTGACGTCAATGGGTG
GGAGGCCTAGGCTTTTGCAAAAAGC	CGGTAAACTGCCCACTTGGCAGTAC
GACACAACAGTCTCGAATTTAAGGCTAG	ATGGGCGTGGATAGCGGTTTGACTC
CTGTTCACCGGAGTGGTGCCTATCC	CCAAAATGTCGTAACAACTCCGCCC
CGACGCTACCTACGGAAAGCTGACC	ATGGGCGGTAGGCGTGTACGGTGGG
GTGCCTTGGCCCACCCTTGTGACCA	GCAGAGCTGGTTTAGTGAACCGTCA
ACCACATGAAGCAGCACGACTTCTT	GATAACATGGCCATCATCAAGGAGT
AACATCCTGGGCCACAAGCTGGAGT	GTGGATAGCGGTTTGACTCACGGG
CATCACCGCCGACAAGCAGAAGAAC	GTCCCCTCAGTTCATGTACGGCTCC
CCCATCGGCGACGGTCCCGTGCTGC	GTGGTGACCGTGACCCAGGACTCCT
CCTGACAACCACTACCTGAGCTACCAG	GGACGGCGAGTTCATCTACAAGGTG
TTCAAGGACCCCAACGAGAAGCGGG	CCCGTAATGCAGAAGAAGACCATGG
CACGACTGTGCCTTCTAGTTGCCAG	CTCCGAGCGGATGTACCCCGAGGAC
GACCCTGGAAGGTGCCACTCCCACT	CGGCGCCTACAACGTCAACATCAAG
GGAAATTGCATCGCATTGTCTGAGT	GTCTCCACCCCATTGACGTCAATGG
GGAAGACAATAGCAGGCATGCTGGG	CATCGCTATTACCATGGTGATGCGG

Fluorescent in situ hybridization

Cells were seeded and grown to confluence inside 12-well plates underlined with a glass coverslip. Cells were infected at multiplicity of infection (MOI) 20 with 200 µl inoculums. The cells were incubated on ice for 1 h following the addition of the inoculums. The inoculums were then removed and replaced with DMEM + 10% fetal bovine serum. The infected cells were incubated on 37°C for 6 h. Medium was then removed and cells were washed with PBS. Cells were then fixed with 1 ml 4% paraformaldehyde in PBSX0.3 for 10 min at room temperature. Cells were washed 3 times with 0.05% Triton X-100 in PBS at 5-min intervals. Cell nuclei were permeabilized by incubating the coverslips with 0.5% Triton X-100 in PBS for 20 min. The 0.5% Triton was then removed and 20% glycerol (MilliporeSigma, Burlington, MA, USA) in PBS was added and incubated at room temperature for 30–60 min. Following incubation, coverslips were repeatedly frozen and thawed by dipping them in liquid nitrogen 3 times. Before refreezing, the coverslips were dipped in the 20% glycerol solution to prevent cells from drying. Coverslips were then washed with 0.05% Triton X-100 in PBS 3 × 5 min. Coverslips were briefly washed in 0.1 N HCl and incubated in fresh 0.1 N HCl for 8 min. Subsequently, the coverslips were washed with 0.05% Triton X-100 in PBS 3 × 5 min and then transferred to 50% formamide (Merck, Kenilworth, NJ, USA) in saline sodium citrate buffer ×2 [SSCX2; diluted from SSCX20 prepared as follows: 175.3gr sodium chloride (BioLab, Lawrenceville, GA, USA), 88.2gr of sodium citrate (Merck) in 1 L of deuterium depleted water, adjusted to pH 7.0 and filtrated]. Viral genomes were hybridized using hybridization solution [10% dextran sulfate (MilliporeSigma) in SSCX2] containing the fluorescent probes (Integrated DNA Technologies, Coralville, IA, USA; final concentration of 0.7 µM for each probe). For each coverslip, 5 µl of probe was placed on a clean slide. Each slide was lifted out of the 50% formamide in SSCX2 solution and coverslips were placed directly on the probe drop, cell side down, and excess fluid was removed. Rubber cement was applied to the edges of the coverslip to form a temporary seal for hybridization. The slides were set aside to dry completely at room temperature. The denaturing of cellular and probe DNA was achieved by placing the slides on a hot slide block (Eppendorf, Hamburg, Germany) for precisely 2 min at 95°C. Following denaturation, slides were sealed with parafilm and placed in a 37°C incubator for 72 h. After sufficient hybridization, the rubber cement was removed from the slides. Each coverslip was carefully removed and washed with SSCX2 3 × 5 min at 37°C then washed with SSCX0.1 3 × 5 min at 63°C. For each coverslip, 1 drop of fluoroshield mounting medium containing DAPI (Abcam, Cambridge, MA, USA) was placed on a fresh slide, and each coverslip was placed cell side down on the drop. Coverslips were sealed with nail polish. FISH protocol was adapted from Cremer et al. (50).

Microscopy

Viral plaques were visualized under a Nikon Eclipse Ti-E epifluorescence inverted microscope (Nikon, Tokyo, Japan). Single cell analysis by FISH assay was performed using a Nikon Eclipse Ti microscope equipped with Yokogawa CSU X-1 spinning disc confocal system (Yokogawa, Musashino, Japan).

Image analysis

Nuclei segmentation: Nuclei were automatically segmented from the DAPI channel as follows. Gaussian smoothing (σ = 2) was applied to the image followed by Otsu thresholding (51) to partition the image to nuclei/background regions. The nuclei segmentation was refined by morphologic operators: opening (width = 3 pixels) to exclude small excesses; closing and an additional opening (width = 20 pixels) to unify regions that belong to the same nucleus; filling holes; considering only connected regions of area exceeding 900 pixels, the minimal nucleus size. The output of this stage is a set of masks, each corresponding to a single nucleus.

RCs were segmented from each fluorescent channel independently. Three groups of intensities were observed within each nucleus: background, at intensities close to those outside the nucleus; intermediate, above the background levels; and bright, which are the replication centers. First, Gaussian smoothing was applied (σ = 2) to the image. Second, background statistics were calculated (mean intensity, sd) from the pixels outside the nuclei masks. A threshold of 2.5 sd above the background mean intensity level was calculated. To exclude pixels with background intensities from further analysis, the threshold was applied for each nucleus mask. Otsu thresholding was then applied to the remaining pixels, pixels below the threshold were pooled, and a new threshold of 2 sd above the mean was calculated. The pixels above these thresholds were defined as the RCs. The segmentation refined by applying morphologic operators: opening, excluding small excess; closing, filling gaps; and opening again, disconnecting independent replication centers; all with a square kernel of 3 pixels width. Last, holes were filled to define the RCs. Parameters were optimized by visual assessment of RC segmentation accuracy.

Regions for statistical analysis were manually selected based on the accurate segmentation of RCs. Regions of accurate RC segmentation were selected independently for each channel; the intersection (implying accurate segmentation on both channels) was used for quantifications.

The program outputs required measurements such as nuclei area, number, and area of RCs in each channel of each nucleus and the size of overlapping area for each RC. The data were then transferred into Microsoft Excel (Microsoft, Redmond, WA, USA) for subsequent processing. The data were collected for each nucleus for further analysis. Cells without colocalization were removed from further analysis. The ratio of colocalization area out of the total area of RCs was calculated per cell (see Supplemental Data). Outliers were removed using the interquartile method (1.5 times interquartile distance from either the first of third quartile). Two-tailed Student’s t test was performed to determine significance. The data collected from the 3 experiments are available in Supplemental Table S1. The Matlab source code used for analysis is publicly available at https://github.com/assafzar/RecombinationHSV1.

Fluorescent plaque assay

Cells (either Vero or U2OS) were grown to confluence in 12-well plates. Infection was carried out at an MOI of 20 with a mix of the coinfecting viruses in 4°C. Cells were harvested 9 h postinfection (HPI). In latrunculin B experiments, the 1 HPI medium was replaced by a medium containing 2.5 μM latrunculin B (MilliporeSigma). The infected cells were lysed by freeze thawing and sonicating. The lysate was centrifuged to remove cell debris and plated on 6-well plates containing Vero cells. Following infection, cells were overlaid with growth medium containing 0.5% methylcellulose and incubated for 72 h. The plates were then inspected under fluorescent microscope for plaque counting. The percent of dual-color and colorless plaques out of the total fluorescent plaques are indicated. A total of 4 wells, collected from 2 individual plaque plating, from 2 technical repeats in 2 separate experiments, were obtained. Two-tailed Student’s t test was performed to determine significance.

RESULTS

A FISH-based assay designed to detect intergenomic recombination events between coinfecting viruses

Recombination among coinfecting HSV-1 strains is a frequent event that can be detected by the progeny viruses released from infected cells (8–10, 12). To study the spatiotemporal constraints of these intergenomic recombination events, we developed a FISH-based experimental assay that enables visualization of 2 different viral genomes within a coinfected nucleus. For this assay, we constructed a series of viral isolates isogenic to each other except for 2 unique tag sequences. These tag sequences, YPet or mCherry encoding genes (yellow or red fluorescent proteins, respectively), were inserted into various loci across the viral genome (Fig. 1A and Table 1). We designed 2 sets of fluorescent probes, 1 set for each tag sequence (Fig. 1B). Each probe set was conjugated to a distinct fluorophore (Cy3 or Cy5) to enable visual identification of the genomes containing each tag sequence. We hypothesize that using different mixtures of viruses should lead to distinct patterns within the infected nucleus (Fig. 1C). As was shown previously for PRV (33), we expect that coinfection with 2 HSV-1 viruses containing tag sequences at the same genomic locus cannot result in a new recombinant genome containing both tags. Thus, each RC will react to a single fluorophore, including at the contacting edges of proximate RCs (Fig. 1C example I). Infection with a viral recombinant containing 2 tag sequences within 1 genome is expected to result in RCs stained with both fluorophores (dually labeled RCs) (Fig. 1C example IV). Infection with 2 viral recombinants containing tag sequences at different genomic loci could result in progeny genomes that either contain both tag sequences on a single genome or contain neither tag sequence. One of 2 types of spatial patterns could be expected to dominate under these conditions. Dually labeled RCs (fully covered by both probe sets) would imply that intergenomic recombination takes place early during infection before the viral DNA replicates and RCs mature. In this situation, the reciprocal recombinant genome would contain no tag sequences and will generate RCs that are not detected because they are not covered by any of the probes (Fig. 1C example II). Alternatively, partially overlapping RCs that fuse to each other at their periphery, would suggest that intergenomic recombination occurs later during the infection cycle following viral DNA replication (Fig. 1C example III). Therefore, this FISH assay is designed to detect viral intergenomic recombination events, with visual readouts for when and where it takes place.

Figure 1.

Schematic representation of dual-color FISH assay. A) A series of viral recombinants with 2 unique tag sequences (designated red and green) inserted into various loci of the parental genome. Illustration of the insertion sequences (color boxes) and viral genomes (black line with repeats marked in gray boxes). Illustrations are not to scale. B) Two sets of probes conjugated to distinct fluorophores; each set corresponds to 1 tag sequence. C) Illustration of expected results, in which the gray nucleus contains viral RCs originating from genomes with either 1 tag sequence (red or green), 2 tag sequences (yellow), or no tag sequence (black). The expected results in the case of infection with tag sequences on the same locus of 2 coinfecting genomes (I), with tag sequences on separate loci in 2 coinfected genomes where recombination takes place before DNA replication (II) or following DNA replication (III) and with 2 tag sequences in the same genome (IV). D) Representative images of viral plaques initiating from progeny viruses collected from Vero cells coinfected with different viral recombinants as marked above each image. White arrows indicate early colorless plaques. Scale bars, 100 µm. E, F) Progeny viruses from either Vero (E) or U2OS (F) cells coinfected with the viral recombinants OK35 and OK25 (blue), OK35 and OK32 (yellow), and OK35 and OK26 (orange) were plated for individual plaques and colors quantitated. The percent of dual-color plaques (full) and colorless plaques (stripes) are indicated. The averages are shown from 2 technical repeats in 2 separate experiments. Error bars represent sd. *P < 0.05, **P < 0.01, ***P < 0.001 by t test.

Patterns of RC interactions dependent on HR

To identify intergenomic recombination events, we coinfected Vero cells with different pairs of viruses listed in Table 1, as detailed below. All infections were carried out at MOI 20 to increase the likelihood of interactions between incoming viral genomes (8). Cells were either infected with 1 virus that contained both mCherry and YPET tag sequences (OK31) or were coinfected with 2 viruses, each carrying a different tag sequence (OK35 together with either OK25, OK32, or OK26). First, we collected the progeny viruses released from the infected cells at 9 HPI and plated them for single plaques. We found that progeny viruses from coinfections can result in dual-color (yellow) plaques as well as colorless plaques under our different infection conditions (Fig. 1D); both are likely to be the results of HR events. The number of dual-color and colorless plaques, which presumably arise from HR events, was measured for each infection pair and compared to the total number of plaques (Fig. 1E). The percent of dual-color plaques was higher compared to colorless plaques because dual-color plaques can also result from coinfection and colorless plaques are likely underdetected (because they are colorless). We found that the rate of colorless and dual-color plaques was increased by ∼9-fold for infections with viral genomes carrying fluorescent protein genes at different genomic sites when compared to infections with viruses where the genes are located at the same genomic site. We note that there was no statistical difference between the 2 infections with the viruses carrying the reporter genes at alternative sites in the genomes (OK35 together with either OK32 or OK26). Our results support the assumption that most progeny viruses producing dual-color or colorless plaques are the outcome of intergenomic HR events (8).

To verify further that the dual-color plaques are a result of recombination events, we randomly picked dual-color plaques resulting from each coinfection and replated the progeny viruses from these plaques onto new cell monolayers (Fig. 2A). We found that dual-color plaques arising from the coinfection with either OK26 and OK35 or OK32 and OK35 (reporter genes at different sites in the viral genomes) produced mostly new dual-color plaques. In contrast, dual-color plaques arising from coinfection with OK25 and OK35 (reporter genes at the same site in the viral genomes) produced almost exclusively single-color plaques (Fig. 2B). These results indicate that dual-color plaques from coinfection with OK25 and OK35 are likely to result from either coinfections or unstable gene duplication, whereas dual-color plaques arising from coinfections with either OK26 and OK35 or OK32 and OK35 result from stable HR events. We conclude that in our experimental system, ∼17% of progeny viruses from coinfection with either OK26 and OK35 or OK32 and OK35 stem from HR events.

Figure 2.

Progeny viruses arising from dual-color plaques. Progeny viruses from Vero (A, B) or U2OS (C) cells coinfected with the viral recombinants OK35 and OK25 (blue), OK35 and OK32 (yellow), and OK35 and OK26 (orange) were plated for individual plaques. From each infection, 10 individual plaques with both fluorescent proteins were picked and replated for individual plaques. A) Representative image from each of the progeny plaques arising from a single individual plaque picked from each coinfection. Scale bars, 100 μm. B, C) For each of the 10 individual plaques, the percent of dual-color progeny plaques is plotted.

To visualize intergenomic HR events at the single-cell level, Vero cells were coinfected and fixed at 6 HPI for hybridization with the appropriate fluorescent probes. Using confocal fluorescent microscopy, we imaged the viral RCs and identified 4 distinct patterns of interactions between the RCs (Fig. 3A–D). The first interaction type (A: minimal overlap) is of 2 RCs that come into close contact but without evidence of mixing between the 2 RCs [i.e., no visible colocalized pixels and no intersection between RCs margins (Fig. 3A)]. These interactions were observed in all coinfections; however, they are significantly more pronounced when the tag sequences were at the same relative genomic locus (OK25 and OK35, P < 0.05, Fig. 3E). The second interaction type (B: periphery overlap) is of 2 RCs that come into close contact and clearly mix with each other, as defined by the presence of dual-color pixels at the site of contact (Fig. 3B). Note that the 2 channels do not necessarily colocalize at the single pixel resolution at the contact sites (i.e., no yellow pixels) but do colocalize at the region scale. This was the most common interaction observed in all coinfections (Fig. 3E), with significantly increased frequency when the 2 tag sequences were located at different sites of the viral genome: either OK26 and OK35 (P < 0.005) or OK32 and OK35 (P < 0.05). In the OK25 and OK35 dual infection (tags located at the same genomic locus), this pattern cannot be due to HR and in this case reflects RC intermingling that is independent of HR events. In the third interaction type (C- Full overlap within larger RC), a small RC (smaller than half the size of all other RCs in that nucleus) is fully contained in a much larger RC (Fig. 3C). This pattern was detected in ∼5% of all coinfections (Fig. 3E). The fourth interaction type (D: full overlap) is a dual-color RC, in which the entire area of the RC contains both fluorophores (Fig. 3D). The relative frequencies of these events can be quantitated during single or dual infections (Fig. 3E). As expected, the overlapping pattern accounted for ∼90% of RCs in cells with single infection of virus carrying both tag sequences (OK31) and in ∼3% of interactions in all dual coinfections. We deduce that our assay can be used to read out intergenomic recombination because detectable HR events change the overall distribution of the interactions among the RCs. In all coinfection assays, partial overlaps at the point of contact are more frequent, compared to entirely overlapping RCs. The rare occurrence of entirely overlapping RCs compared to the high frequency of HR observed in the coinfection plaque assay and the relative similar rate of appearance in all 3 coinfections suggest that these RCs capture infrequent events and cannot reflect the majority of HR between RCs. We therefore suggest that intergenomic HR occur at points of physical interaction between coinfecting genomes after viral DNA replication has initiated.

Figure 3.

Observed patterns of RC interactions. A–D) Representative images of Vero cells infected with viral isolates containing 2 genomic tags, each corresponding to a set of fluorescent probes. The probe sets are imaged separately in red (I), green (II), and merged (III). Four observed patterns of interaction between RCs are represented. A) RCs containing different tag sequences come into contact without overlap (orange arrowhead), imaged from cells infected by 2 viral recombinants containing tag sequences on the same genomic locus (OK25 and OK35). B) RCs containing different tag sequences overlap at their periphery (white arrowhead), imaged from cells infected with 2 viral recombinants containing tag sequences on separate genomic loci (OK32 and OK35). C) RC containing 1 tag sequence overlaps completely within a larger RC with the second tag (black arrowhead) imaged from cells infected with 2 viral recombinants containing tag sequences on separate genomic loci (OK26 and OK35). D) Entirely overlapping RCs, imaged from cells infected by a viral recombinant containing 2 tag sequences on the same genome in separate loci (OK31). DAPI nuclear stain is presented in gray. Scale bars, 5 µm. E) Percent of each of the observed RCs interactions in A–D from the total number of RCs interactions. Manually counted at the different coinfection conditions. More than70 RCs interactions per infection type per experiment were counted. An average of 3 experiments and the sd in brackets are shown.

We have previously shown that infections of U2OS cells result in more HSV-1 genomes initiating gene expression per cell when compared to infections of Vero cells (52). To test if the number of initiating genomes has an effect on the interactions between RCs, we repeated our dual-color infection assay in U2OS cells. Although overall higher levels of dual-colored plaques were observed in U2OS cells when compared to Vero cells, the relative levels of dual-colored plaques attributed to HR (i.e., coinfection with either OK26 and OK35 or OK32 and OK35 relative to coinfection with OK25 and OK35) was not different between the 2 cell types (Fig. 1F). Similar to Vero cells, the dual-color plaques from U2OS coinfections with either OK26 and OK35 or OK32 and OK35 represent stable HR events (Fig. 2C). In both U2OS and Vero cells, there was no statistical difference between HR rates when comparing infections with viruses that carry the second reporter gene at different locations in the viral genome (OK26 vs. OK32).

In FISH-based experiments with infections of U2OS cells, we observed all 4 patterns of RCs interactions that we defined for infections of Vero cells (Fig. 4). However, in U2OS cells, multiple RC interaction patterns in a single cell were more common than in Vero cells (Fig. 4A, B), probably due to the higher number of RCs observed per cell in this cell type (See below). The distribution of patterns observed for the different infections in U2OS cells (Fig. 4D) was similar to the distribution detected in Vero cells (Fig. 3E), although in U2OS cells, lower rates of type C and D interactions were observed in all coinfections tested.

Figure 4.

Patterns of RC interactions on U2OS cells. A–C) Representative images of U2OS cells infected with either OK25 and OK35 (A), OK32 and OK35 (B), or OK31 (C) viral recombinants. The probe sets are imaged separately in red (I), green (II), and merged (III). Arrowheads are color coded as in Fig. 2. DAPI is presented in gray. Scale bars, 5 µm. D) Percent of each of the observed RCs interactions out of total RCs interactions. Manually counted at the different coinfection conditions. More than 70 RCs interactions per infection type per experiment were counted. An average of 3 experiments and the sd in brackets are shown.

Overlap between RCs is enhanced by HR

Merely identifying the patterns of overlap does not estimate the relative proportion of the overlapping area between RCs. Therefore, we developed a quantitative evaluation of the RC overlapping areas in each cell. Infection with 1 virus carrying the 2 tag sequences (OK31, see Figs. 3DIII and 4CIII) showed limited colocalization at the pixel level, probably due to the relaxed form of the viral DNA during transcription and replication (34). Therefore, we estimated that common pixel-based colocalization methods do not provide an accurate estimate of the degree of overlap between HSV-1 RCs. We applied a semiautomated object-based method to quantify the degree of overlap for RCs in cells infected with viruses that have 2 distinct genomes (see Materials and Methods). For each cell nucleus, we independently segmented the RCs in each channel. The relative overlapping area was calculated as the ratio between the overall area of the overlap and the total overall area occupied by viral DNA (total area of all RCs minus the total overlap area). The images analyzed were collected from 3 infections carried out on different days from separate viral stocks. Over 200 cells from each cell type were analyzed for each coinfection. The dual-tagged virus (OK31) allowed us to analyze fewer cells in this infection (n = 120 Vero, n = 99 U2OS) but still with sufficient statistical power.

We analyzed all the cells in which contacts between red and green RCs were observed (>70% of cells in all coinfections). For both cell types, we found a small but significant increase in the relative overlapping area in each nucleus for cells coinfected by viruses with tags in different viral loci compared to cells coinfected by viruses with tags in the same locus (>16% increase for OK32 and OK35 infection on Vero cells and >30% increase in all other coinfections, Fig. 5A, B). As with plaque assays (Fig. 1E, F) and manual count (Figs. 3E and 4D), we did not observe significant differences in overlap between coinfections with viruses that have tags in different loci. To confirm that cellular parameters do not bias our measurements, we compared the relative overlapping area to the total nuclear area (Fig. 4C, D) or to the number of RCs per nucleus (Fig. 4E, F). We found no evidence of correlation between these 2 parameters and the relative overlapping area in all infection conditions (Pearson correlation <0.3 for each of the infections). We conclude that the observed significant increase in overlapping areas indicates the readout of HR events. As mentioned above, the relatively high background levels (OK25 and OK35 infection) of overlapping areas (∼20%) are probably due in part to RCs intermingling during viral replication that is independent of HR events.

Figure 5.

HR enhances overlap between viral genomes. Vero cells (A, C, E) and U2OS cells (B, D, F) were infected with the viral isolates OK35 and OK25 where tags are inserted into the same genomic locus (blue), the viral isolates OK35 and OK32 (yellow) or OK35 and OK26 (orange) containing tags on separate genomic loci, and the viral isolate OK31 containing 2 tag sequences on a single viral genome (gray). A, B) Each column represents the average relative overlapping area calculated from all cells (with detectable overlap) infected under the same condition. C–F) Individual cells were plotted to compare the relative overlapping area to the nuclear area (C, D) or to the number of RCs per cell (E, F). A trend line (color coded as above) was calculated using the ordinary least squares (OLSs) method is presented for each infection condition. All analyses were conducted on images generated from 3 independent experimental repeats done on different days with viral stocks prepared and tittered separately. AU, arbitrary units. *P < 0.05, **P < 0.01, ***P < 0.001 by t test.

Triple-color FISH indicates that intergenomic recombination occurs at the edge of coalescing RCs

We speculated that if there are RCs that originate from a single recombinant genome and contain both tag sequences, there must be reciprocal recombinant RCs that contain none of the tags and are undetectable in our FISH assay (Fig. 1CII). To determine whether these RCs exist, we designed an additional probe conjugated to a third fluorophore corresponding to the genomic HSV a’ sequence, a repetitive sequence found in 4 copies within the HSV-1 genome (53). This probe stains all HSV-1 viral DNA regardless of the tag sequence inserted. Both Vero and U2OS cells were coinfected with 2 isolates containing tags at different genomic loci and were fixed and hybridized to all probes. We inspected >600 RCs from each cell type and found that 99.6% of RCs that were stained by the HSV nonspecific probes reacted to either of the 2 specific probe sets (example in Fig. 6A). The existence of at least 1 tag sequence for all RCs suggests that RCs are not generated from recombinant genomes. This result supports the hypothesis that recombination occurs at the edge of mature coalescing RCs at later stages of infection and is coupled with replication. Similarly, all coalescing areas between mature RCs respond to both the specific probe sets in contrast to our prediction that half of these areas should not respond to either specific probe (Fig. 1CIII). This finding suggests that these sites of recombination contain a mixture of genomes that contain either both tags or none of them. We therefore hypothesize that these sites of recombination do not arise from a single recombination event (because this would have led to the phenomenon of both tags or none) but rather from multiple events along the contact front of replicating RCs.

Figure 6.

HR occurs where RCs coalesce. A) Representative images of U2OS cells infected with 2 viral isolates containing tag sequences at different genomic loci (OK35 and OK32). The cells were hybridized with 2 probes conjugated to 3 distinct fluorophores. I) Detection of the OK35 tag sequence is shown in red. II) Detection of the OK32 tag sequence is shown in green. III) Detection of the viral a sequence (detect HSV DNA nonspecifically) is shown in blue. IV) Overlays of 3 colors are shown. DAPI is presented in gray. Scale bars, 5 µm. B, C) Vero (B) or U2OS (C) cells were coinfected with the viral recombinants OK35 and OK25 (blue), OK35 and OK32 (yellow), and OK35 and OK26 (orange) in the presence (empty bar) or absence (full bar) of latruncalin B. Progeny viruses from each infection condition were plated for individual plaques and colors were quantitated. The percent of dual-color plaques (full) and colorless plaques (stripes) out of the total observed plaques were measured. The means are shown from 2 technical repeats in 2 separate experiments. Error bars represent sd. *P < 0.05, **P < 0.01, by t test.

Inhibition of RC movement reduces the rate of HR events

It was previously shown that RC movement toward each other relies upon the formation of actin filaments (31). Because our results suggest that HR events occur between coalescing RCs, we tested whether disruption of RC movement will result in reduction of recombination rates. We treated coinfected cells with latrunculin B, an inhibitor of actin polymerization, and tested for recombination rate among the progeny viruses by fluorescent plaque assay. Vero and U2OS cells were infected with the 3 sets of pairs of viral constructs (OK25 and OK35, OK32 and OK35, or OK26 and OK35) and were treated with latrunculin B. At 9 HPI, progeny viruses were collected and plated for single plaques. The rate of dual-color and colorless plaques was measured for each infection condition and compared to the rate in the absence of latrunculin B (Fig. 6B, C). The presence of latrunculin B reduced dual-color and colorless plaques compared to untreated cells for both infection pairs in both cell types. Statistically significant decreases were observed in Vero cells during OK26 and OK35 coinfection and in U2OS cells for OK32 and OK35 coinfection. In comparison, latrunculin B did not change the ratio of dual-color and colorless plaques following OK25 and OK35 coinfection. These results suggest that RC movement enhances intergenomic HR, corroborating our initial conclusion that HR events occur at the contact point of coalescing RCs.

The number of RCs correlates with nuclear size

We previously demonstrated that the number of HSV-1 genomes replicating per cell can influence the outcome of infection (54) and that host conditions prior to infection impact the outcome of infection at the single-cell level (54, 55). To test the effect of cellular parameters on RC formation and growth, we took advantage of our assay that provides an opportunity to quantitate the number of mature RCs within individual cells because we can distinguish between the colors of coalescing RCs. We note that OK31 infection was omitted from this analysis due to our inability to distinguish between coalescing RCs in this infection. We therefore tested association of the number of RCs to other parameters of the infected cell. First, we tested whether nuclear area correlates with the number of RCs per nucleus. We observed that the number of RCs correlates to the nuclear area for individual cells in each of the cell types (Pearson correlation: R > 0.538 for all coinfections, P < 0.0001 for all coinfections, Fig. 7A, B). We hypothesize that the increase in the number of RCs per nucleus will result in more total viral DNA (RCs area) in larger nuclei. As expected, the nuclear area correlates with the total area of the RCs per nucleus (Pearson correlation: R = 0.530 for OK32 and OK35 coinfection on Vero and R > 0.625 for all other coinfections, P < 0.0001 for all coinfections, Fig. 7C, D). We speculated that the increase in total RCs area could also result from the possibility that RCs expand faster to a larger size in larger nuclei. We therefore compared the mean RC area (per cell) to the total nuclear area (Fig. 7E, F). We found a much weaker correlation between these parameters (Pearson correlation: <0.5 for U20S cells and <0.3 for Vero cells in each of the infections). These results suggest that the increase in RC area in larger nuclei results from higher numbers of RCs rather than increased RC size. Taken together, our results suggest that nuclear size could be a restricting factor, limiting the number of incoming genomes that are able to initiate replication.

Figure 7.

Nuclear size correlates with the number of RCs per cell. Cells were coinfected with the viral recombinants OK35 and OK25 (blue), OK35 and OK32 (yellow), and OK35 and OK26 (orange). A–F) Individual Vero (A, C, E) and U2OS (B, D, F) cells were plotted to compare the number of RCs per cell (A, B), the total area of RCs per cell (C, D), and the average area of each RCs (E, F) to the nuclear area. A trend line (color coded as above) was calculated using the ordinary least squares (OLSs) method is presented for each infection condition. G, H) Comparing the mean number of RCs per cell (G) and the mean nuclear area (H) between Vero (bright colored columns) and U2OS (dark colored columns). All analyses were conducted on images generated from 3 independent experimental repeats done on different days with viral stocks prepared and tittered separately. Error bars represent se of the means. AU, arbitrary units.

DISCUSSION

Recombination among coinfecting herpesviruses is fundamentally important for understanding viral evolution and pathogenesis (5). It is also important to consider intergenomic recombination when developing vectors for vaccination and oncolytic viral therapy because evidence for recombination between a vaccine strain and a wild-type strain has already been found in herpesviruses (56). Here, we developed a unique experimental system to identify the spatiotemporal constraints of intergenomic recombination. Our results corroborate the hypothesis that each RC initiates from a single incoming genome. Insertion of fluorescent tags in 2 separated places in 2 viral recombinants, increases HR events (visualized by plaque assay) and the overlapping areas between RCs (visualized by FISH assay) during coinfection, compared to coinfection with viral recombinants with tags at the same place. We suggest that most viral intergenomic recombination events occur at the edges of developing RCs where they coalesce with others. Areas in which overlapping RCs are detected probably result from multiple independent recombination events. Inhibition of RCs movement reduced the rate of recombination, further supporting our suggestion that HR events occur after coalescence of RCs. Finally, we found a correlation between nuclear size and the number of RCs detected per cell.

We infected cells with different combinations of viral isolates: OK35 and OK25 where tags are inserted into the same genomic locus, OK35 and OK32 containing tags on separate genomic loci spaced ∼72 Kbp apart, and OK35 and OK26 where they are spaced ∼105 Kbp apart. In our analysis, we did not detect any significant differences between the 2 combinations of viruses tagged at separate genomic loci (i.e., OK35 and OK32 as well as OK35 and OK26). This supports the hypothesis that recombination events occur between replicating genomes that are in either branched or circular concatemeric states (7). Because HSV-1 maintains the 4 genome isomers at equivalent distributions (57), the 105 Kbp distance between the markers is actually ∼50 Kbp in half of the concatemeric forms. This may explain the inability to distinguish differences in recombination rates when the markers are not closely linked, as previously suggested in Honess et al. (7).

Recent work from the Law et al. (8) article suggested that the rate of recombination is dependent on MOI. We therefore chose a relatively high MOI (i.e., MOI 20) for all our experiments. This high MOI (or even higher) is probably occurring in nature within the herpetic lesion (cold sores) where the concentration of viral PFU is extremely high (58).

We observed 4 recognizable patterns of interactions between adjacent RCs: 1) no mixing, 2) partial mixing, 3) one within another, and 4) complete overlap (Figs. 3 and 4). From our previous results with PRV (33), we expected that the lack of mixing of RCs would be the dominant interaction following coinfection with 2 viruses carrying tag sequences at the same location in the viral genomes (OK25 and OK35). We found that this readout is most commonly observed in this coinfection, although to our surprise the majority of RCs (∼ 60%) showed some degree of mixing. Revisiting the images obtained during the Kobiler et al. (33) article suggested that there is no major difference between the images from the 2 alphaherpesviruses coinfection (i.e., the overlap between signals was also observed in the PRV experiments but was attributed to image resolution and noise). We note that robust quantitative analysis of PRV images was not carried out.

The partial mixing of RCs was the most common interaction observed among the different coinfection conditions tested (Figs. 3E and 4D). This is in part due to the categorization of all partial mixing interactions into 1 single pattern, regardless of the proportion of mixing. To overcome this problem, we developed an image analysis code that calculates the relative area of mixing. We found a significant increase in the relative area of mixing when coinfections were carried out with viruses in which HR can result in the mixing of the colors. These measurements were significant both at the single-cell level (Fig. 5A, B) and at the single-RC level, indicating that a significant proportion of the mixing we observe is due to HR. On the other hand, even when HR cannot allow for signal mixing (i.e., OK25 and OK35 coinfection), we observed RCs and cells in which a high proportion of mixing occurred (Fig. 5). Similar findings were found in PRV coinfection assays (33). These results suggest that some color mixing can occur without recombining both tag sequences into the same genome. Alternatively, they could reflect non-HR events during replication, which were previously observed both in HSV-1 replication (57) and in other herpesviruses (59). These explanations are not mutually exclusive, and both may contribute to the mixing of colors observed without HR.

The interaction in which 1 RC is fully mixed into a larger RC (Fig. 3C) can be explained by multiple mechanisms. First, it could originate from a small RC that coalesces with a larger one. Second, the smaller RC could interact with 2 larger RCs that flank both sides of the small RC and eventually unite into a single, larger RC. A third mechanism could involve the recently observed condensed viral genomes at the edge of RCs (34). Sekine et al. (34) showed that at 3 HPI several incoming genomes remain condensed at the periphery of an enlarging RC. These condensed genomes might be silent throughout the lytic infection or may serve as templates for recombination at later time points (which will result in our observed third interaction). The latter hypothesis fits our finding that this interaction is less frequent in U2OS (Figs. 3E and 4D). Because U2OS cells lack ATRX and have compromised expression of other intrinsic immunity proteins (60–62), they are less likely to silence incoming genomes (52), and we speculate that they may have lower amounts of condensed genomes. Further experiments are required to distinguish between these different possibilities.

Complete overlap among RCs, which is the dominant phenotype of OK31 infection, was also observed in all coinfections (Figs. 3 and 4). However, because the ratio of these RCs is low and is comparable in all coinfections (both when the tags are both in the same place in the genome or at separate places), it is unlikely to be the source of most HR events.

In mammalian nuclei, translocations between chromosomes that can lead to tumorgenicity and other illnesses are thought to occur among adjacent chromosomes (63, 64). Our results support the requirement of proximity (contact first) among DNA molecules prior to recombination. Thus, maintaining genome integrity may require physical separation between DNA molecules. Our results with the 3 color FISH suggest that once the physical separation between RCs is breached, multirecombination events can occur (Fig. 6A). It will be interesting to determine what defines the location of viral RCs and the extent to which liquid phase separation (65) and other mechanisms regulate their mingling.

Our analysis of the dual-color FISH experiments led us to hypothesize that most viral intergenomic HR events occur at the edge of the RCs and only after they coalesce. Further support for this hypothesis was obtained from the latrunculin experiment (Fig. 6B, C). Because viral RCs movement in the nucleus is dependent on actin filaments (31), the inhibition of actin polymerization by the latrunculin B inhibitor should reduce RCs coalescence. Thus, the reduction in intergenomic HR rates observed with latrunculin treatment can be attributed to a reduction in interaction between growing RCs. We note that when RCs enlarge, they will eventually probably interact with each other without direct motion.

Our experimental system provides the opportunity to calculate accurately the number of functional RCs within a nucleus because the separation by color increases our ability to differentiate between coalesced RCs. We observed that in U2OS cells more RCs were detected per individual nucleus than in Vero cells. We detected an average of ∼5.1 RCs per nucleus in U2OS cells compared to ∼3.7 RCs per nucleus in Vero cells (1.375 fold increase, Fig. 7G). This corroborates our previous finding that in U2OS cells on average a higher number of incoming viral genomes per cell initiate expression and replication compared to Vero cells (52). These observations also further support the emerging view that individual RCs initiate from single incoming genomes (33). U2OS nuclei were larger (fold increase of ∼1.38 in nuclear area) compared to Vero cells (Supplemental Fig. S7H). However, in both cell types, we observed correlations between the nuclear area and the number of detectable RCs. Previously, we detected a similar correlation between cell size (as measured by flow cytometry) and the number of replicating genomes (54). Because our images represent snapshots during the infection process, we cannot distinguish between the possibilities that larger nuclei allow a larger number of incoming viral genomes to initiate replication or that the nuclei in which more genomes initiated replication expand faster. The use of new methods to visualize incoming genomes using click chemistry (34, 35, 66) should be able to resolve these 2 alternatives.

We note that U2OS cells are known to have larger nuclei and indeed have more RCs per cell than Vero cells. Similarly, HeLa cells that are relatively small cells with small nuclei have fewer incoming HSV-1 genomes that initiate replication (54). However, both U2OS and HeLa cells originated from tumors and carry multiple abnormalities that might regulate the number of herpes genomes that initiate replication. However, we are not suggesting that the average nuclear size of a cell type is the main factor in determining the number of herpes genomes initiating replication among cell types.

In summary, our results support a model in which incoming viral genomes establish RCs at distinct sites in the nucleus. The vast majority of RCs initiate from a single genome and only at later stages coalesce but maintain a detectable separation between RCs. These areas of interactions between the expanding RCs are predicted to be the site in which HR occurs in tight association with viral DNA replication. Our results suggest that in these sites, intergenomic recombination is a frequent event, although not all areas of interactions between the RCs are supportive of close contact that can lead to recombination. Further research will shed light on the differences between these types of RC interactions.

Supplementary Material

This article includes supplemental data. Please visit http://www.fasebj.org to obtain this information.

Glossary

CMVp

cytomegalovirus promoter

Cy3/5

cyanine 3/5

HPI

hours postinfection

HR

homologous recombination

HSV-1

herpes simplex virus type 1

MOI

multiplicity of infection

PRV

pseudorabies virus

RC

replication compartment

SSCX2

saline sodium citrate buffer ×2

SV40p

simian virus 40 promoter

AUTHOR CONTRIBUTIONS

E. Tomer and O. Kobiler designed research; E. Tomer, E. M. Cohen, N. Drayman, A. Afriat, A. Zaritsky, and O. Kobiler analyzed data; E. Tomer performed research; E. Tomer, M. D. Weitzman, A. Zaritsky, and O. Kobiler wrote the manuscript; and N. Drayman and A. Zaritsky developed software necessary to analyze experiments.