|Home | About | Journals | Submit | Contact Us | Français|
Herpes simplex virus 1 (HSV-1) is a human pathogen that leads to recurrent facial-oral lesions. Its 152-kb genome is organized in two covalently linked segments, each composed of a unique sequence flanked by inverted repeats. Replication of the HSV-1 genome produces concatemeric molecules in which homologous recombination events occur between the inverted repeats. This mechanism leads to four genome isomers (termed P, IS, IL, and ILS) that differ in the relative orientations of their unique fragments. Molecular combing analysis was performed on DNA extracted from viral particles and BSR, Vero, COS-7, and Neuro-2a cells infected with either strain SC16 or KOS of HSV-1, as well as from tissues of experimentally infected mice. Using fluorescence hybridization, isomers were repeatedly detected and distinguished and were accompanied by a large proportion of noncanonical forms (40%). In both cell and viral-particle extracts, the distributions of the four isomers were statistically equivalent, except for strain KOS grown in Vero and Neuro-2a cells, in which P and IS isomers were significantly overrepresented. In infected cell extracts, concatemeric molecules as long as 10 genome equivalents were detected, among which, strikingly, the isomer distributions were equivalent, suggesting that any such imbalance may occur during encapsidation. In vivo, for strain KOS-infected trigeminal ganglia, an unbalanced distribution distinct from the one in vitro was observed, along with a considerable proportion of noncanonical assortment.
Herpes simplex virus 1 (HSV-1) is a pathogen that commonly leads to primary and recurrent orofacial and ocular lesions (32). Epithelial lesions result from massive production of virions with attendant immune responses. Primary infection is followed by the spread of viral particles to sensory neurons, especially those of the trigeminal ganglia, and establishment of a lifelong latent infection. Reactivation can take place sporadically, causing a new productive infection in tissues innervated by the neurons that harbored the reactivation process.
Although HSV-1 seroprevalence in the general population ranges from 60 to 90% depending on the country and age (9), HSV-1 infections are mostly asymptomatic. However, HSV-1 may lead to recurrent labial lesions in 15 to 45% of the adult population (18), encephalitis in one per 500,000 people per year (65), and corneal keratitis with an annual incidence of 30 cases per 100,000 people (30). The genome is a linear ~152-kb DNA molecule that is organized into two covalently linked segments: a long (L; 82% of the genome) and a short (S; 18% of the genome) fragment. Each segment is composed of a unique sequence, UL and US, bounded by inverted repeats referred to as b and c, respectively. They are both flanked by the a inverted repeats, so that the final canonical structure is ab-UL-b′a′c′-US-ca, also designated a-TRL-UL-IRL-a′-IRS-US-TRS-a, with TR and IR standing for terminal repeat and internal repeat, respectively (51, 63). During productive infection, HSV-1 DNA replication generates concatemeric DNA more than one unit in length that includes complex branched structures (26, 34, 50, 69). It has long been thought that the linear viral genome circularized and was replicated through a rolling-circle mechanism (7). However, this model is being challenged, as analysis of wild-type HSV-1 DNA on Gardella gels, which allow the differentiation of latent episomal from lytic linear viral DNA on the basis of electrophoretic mobility, has shown that its circularization may not occur after its entry into cells (25). Furthermore, it was suggested that synthesis of HSV-1 DNA proceeded through recombination-dependent replication (67).
There is clear evidence that recombination is intrinsically linked to replication (16, 37, 45, 64). Recombination events are frequent, as implied by the study of the products of cells coinfected with two types of HSV-1 genome (8, 21). During replication, homologous recombination may occur between the inverted repeats of the HSV-1 genome (54, 55, 64), resulting in the generation of four isomers differing in the relative orientations of their L and S genomic fragments. They are designated P, for prototype; IL, for inversion of the L fragment; IS, for inversion of the S fragment; and ILS, for inversion of the L and S fragments (19, 44, 51). Homologous recombination is not sequence specific (10, 11, 42, 54–56, 64). Although the a sequence is dispensable for isomerization (33), it seems to play an important role, as it supports recombination with high efficiency despite its short length (11, 54, 61, 62, 64). Moreover, it contains the packaging signal and the cleavage site of concatemeric DNA (35, 38, 56). The alternative cleavage events account for the generation of the four genome isomers (56, 62).
Recombination events seem to occur early during HSV-1 DNA replication, as genome isomers have been observed in concatemeric DNA in the earliest detectable products (69). They are probably initiated by double-strand breaks (45). Moreover, it has been shown that the cellular endonuclease G cleaves in the a sequences (23, 24, 68). Lastly, virion and cellular HSV-1 DNAs present nicks and gaps (3, 17, 26, 43, 52, 66) that could lead to double-strand breaks during the passage of the replication forks, favoring recombination (29).
Recombinant mutants of HSV-1 that are not able to undergo homologous recombination due to the deletion of part or all of the inverted repeats at the junction between the L and S fragments (a′b′/a′c′) showed lower efficiency of replication in vitro than the corresponding wild-type virus and an absence of acute infection in vivo (27, 28, 41, 46). These observations suggest that either the doubled rate of synthesis of proteins encoded by the b′a′/a′c′ sequences and/or the ability to undergo homologous recombination is required for efficient in vivo infection. It has been previously shown that the four genome isomers of strains MP and 17 of HSV-1 are randomly distributed in virion preparations obtained from infected Hep-2 and BHK cells, respectively (12, 13, 19). These data were obtained either by the use of endonucleases that cleave within a 32P-labeled HSV-1 DNA or by the analysis of partially denatured HSV-1 DNA by electron microscopy. The genome isomers were also detected in concatemeric DNA using techniques such as pulsed-field gel electrophoresis/field inversion gel electrophoresis coupled with HSV-1 DNA endonuclease restriction and Southern blotting (1, 2, 34, 49, 53, 69).
In the present study, we tested the hypothesis that the process of isomerization systematically results in a random distribution no matter what cells or strain is being considered. We applied molecular combing for the direct visualization and analysis of a large number of single HSV-1 DNA molecules through hybridization of uniformly and irreversibly stretched DNA on silanized glass surfaces (36, 48). Molecular combing analysis was performed on stretched DNA extracted from either Vero, COS-7, BSR, or Neuro-2a cell lines infected with the SC16 or KOS strain of HSV-1 or on corneas and trigeminal ganglia from mice experimentally infected with the same viral strains. Using fluorescence hybridization with a set of digoxigenin- or biotin-labeled probes covering the entire HSV-1 genome, surprisingly, we observed that the distribution of the genome isomers in both cell and viral-particle extracts varied significantly from a random distribution depending on the strain and of the cell lines used for production. A similar observation was made in vivo. As molecular combing allows the visualization of individual DNA molecules within a heterogeneous population, we also identified concatemers of up to 10 genome equivalents, along with a large proportion of noncanonical forms, many of them showing signs of recombination. Among the large numbers of HSV-1 genomes generated, some probably are noninfectious, indicating that although HSV-1 replication is a costly process, it is not incompatible with efficient transmission.
Wild-type HSV-1 strains SC16 (20) and KOS (M. Levine, University of Michigan, Ann Arbor, MI) were used. Stocks of HSV-1 SC16 and KOS were prepared and titrated in BSR cells as previously described (31). African green monkey kidney epithelial cell (Vero), African green monkey kidney fibroblast (COS-7), and subclones of baby hamster kidney cell (BSR) and albino mouse neuroblast (Neuro-2a) cell lines were cultured in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin and grown at 37°C in a humidified atmosphere with 5% CO2.
Cells were seeded in 6-well plates at 2 × 105 cells per well and incubated overnight at 37°C in Dulbecco's modified Eagle's medium. After 24 h, the cells were infected with HSV-1 strain SC16 or KOS at a multiplicity of infection of 0.5. Up to 16 h postinfection, culture supernatants were recovered and centrifuged for 5 min at 5,000 × g (Heraeus 7593 rotor) at room temperature. The viral particles and infected cells embedded in agarose plugs were prepared as previously described (48), except for the viral-particle preparation, which required a 1% (wt/vol) low-melting-point agarose (Nusieve GTG; Cambrex) solution in 1× phosphate-buffered saline (PBS).
All procedures were performed in compliance with the Association for Research in Vision and Ophthalmology guidelines for the use of animals in ophthalmic and visual research and the tenets of the European guidelines for protection and welfare of animals. The mice were infected with either HSV-1 strain SC16 (100 PFU per eye) or KOS (106 PFU per eye) by corneal scarification, based on preliminary experiments on the optimal inoculum to induce reproducible and nonlethal infection of the cornea and connected neurons. The mice were sacrificed on the sixth day postinoculation. The corneas and trigeminal ganglia were recovered, rinsed three times in 1× PBS for 15 min at room temperature, and then cut into small pieces. Dissociation of cornea and trigeminal ganglion tissues was performed for 3 h at 37°C in a 0.3-mg/ml collagenase type A (Roche) and 0.8-mg/ml Gibco dispase (Invitrogen) solution and in a 1-mg/ml collagenase type A and 0.25-mg/ml Gibco trypsin (Invitrogen) solution prepared in Gibco 1× Hanks' balanced salt solution buffer (Invitrogen), respectively. The lysates were pelleted by centrifugation (Heraeus 7593 rotor) at 5,000 × g for 10 min, resuspended in a 1× PBS solution, and mixed thoroughly at a 1:1 ratio with a 1.2% (wt/vol) solution of low-melting-point agarose prepared in 1× PBS at 50°C. Ninety microliters of the tissue lysate-agarose mixture was poured into a Chef plug-forming well (Bio-Rad) and left to cool for 30 min at 4°C.
Extraction of genomic DNA for all samples except viral particles was performed by digestion of the agarose plugs according to the procedure described by Schurra and Bensimon (48). Genomic DNA from HSV-1 particles was extracted either by a phenol-chloroform method, as previously described (4), or in agarose plugs with a slightly modified procedure. Lysis of viral particles was performed by dipping agarose plugs in 1% SDS-0.5 M EDTA, pH 8.0, solution at 50°C for 30 min. After three washing steps in digestion buffer (0.5 M EDTA, pH 8.0) for 10 min at room temperature, the plugs were digested by overnight incubation at 50°C with 2 mg/ml proteinase K (Eurobio) in 250 μl digestion buffer.
Molecular combing was performed using the Molecular Combing System (Genomic Vision S.A.) and Molecular Combing coverslips (20 by 20 mm; Genomic Vision S.A.) as previously described (48). After the combing process, the coverslips were dried for 4 h at 60°C. For direct visualization of combed HSV-1 DNA fibers, the coverslips were mounted with 20 μl of 1/1,000 (vol/vol) Yoyo-1 iodide (Molecular Probes)-Prolong Gold antifade reagent (Invitrogen).
HSV-1 genomic fragments ranging from 1,110 to 9,325 bp were obtained by either SacI or BspEI (New England BioLabs) enzymatic digestion of HSV-1 strain SC16 or by long-range PCR using LR Taq DNA polymerase (Roche), a set of HSV-1-specific primers (Table 1), and the DNA from HSV-1 strain SC16 as the template DNA. SacI and BspEI HSV-1 fragments were ligated in the SacI- and XmaI-digested pNEB193 plasmid (New England BioLabs). The PCR products were ligated in the pCR2.1 vector using the Topo TA cloning kit (Invitrogen). Both extremities of each cloned fragment were sequenced with the universal primer M13 forward primer −48, 5′-CGCCAGGGTTTTCCCAGTCACGAC-3′, and the M13 reverse primer −48, 5′-AGCGGATAACAATTTCACACAGGA-3′, for verification and compared to the reference HSV-1 genome sequence (GenBank accession number NC_001806). DNA labeling was performed using the conventional random-priming protocol with either biotin-14-dCTP or digoxigenin-11-dUTP and 200 ng of DNA as a template. For biotin-14-dCTP labeling, the BioPrime DNA kit (Invitrogen) was used according to the manufacturer's instructions, except that the labeling reaction proceeded overnight. For digoxigenin-11-dUTP, a deoxynucleoside triphosphate (dNTP) mixture composed of 2 mM dATP, 2 mM dCTP, 2 mM dGTP, 1 mM dTTP, and 1 mM digoxigenin-11-dUTP (Roche) was used in place of the dNTP mixture furnished with the BioPrime DNA kit.
Hybridization steps were performed as described by Schurra and Bensimon (48). Detection of labeled probes was carried out using three layers of antibodies in a 1:25 dilution. Biotin-14-dCTP-labeled probes were revealed with Alexa Fluor 594-conjugated streptavidin (Invitrogen) as the first step, followed by incubation with a biotinylated goat anti-streptavidin antibody (Vector Laboratories) and then with Alexa Fluor 594-coupled streptavidin. Digoxigenin-11-dUTP-labeled probes were consecutively revealed with an aminomethylcouramin acetate (AMCA)-conjugated monoclonal mouse anti-digoxigenin antibody (Jackson ImmunoResearch) and then AMCA-coupled rat anti-mouse (Jackson ImmunoResearch) and Alexa Fluor 350-conjugated goat anti-rat IgG (Invitrogen) as a final step. After the last washing steps, the hybridized coverslips were gradually dehydrated in 70%, 90%, and 100% ethanol solutions and air dried.
Yoyo-1 iodide-stained or hybridized-combed DNA coverslips were scanned with an inverted automated epifluorescence microscope equipped with a 40× objective (ImageXpress Micro; Molecular Devices). All images were uniformly contrasted using ImageJ (http://rsbweb.nih.gov/ij/) with identical parameters to allow comparison. The lengths of Yoyo-1 iodide-stained DNA fibers and HSV-1-specific hybridization signals were measured using a custom image analysis system (5) and converted to kilobases using an elongation factor of 2 kb/μm. The minimum number of fluorescent signals analyzed was determined using the central-limit theorem that leads to the following equation: n = [ê(1−ê)Zc2]/m2, with Zc as the quantile of the standard normal distribution in a 95% confidence interval, ê as the estimated proportion of a given isomer, m as the margin of error between the estimated proportion ê and the actual proportion of a given isomer, and n as the size of the sample. With a tolerated margin of error of 5% and the proportion ê estimated at 50%, 405 signals were randomly selected and analyzed. The distribution was tested by a chi-square test and accepted when the P value observed was above 0.05. The binomial error was calculated according to the equation , where Ntotal is the total number of signals analyzed and N is the number of a given isomer.
In order to characterize the structure and the organization of the HSV-1 genome, we applied molecular combing, a technology that allows the direct visualization and analysis of single DNA molecules at the kilobase level of resolution. Since the HSV-1 DNA is a 152-kb-long genome, high-resolution mapping of the genome requires very high-molecular-weight DNA. First, HSV-1 strain SC16 DNA was extracted from viral particles by the standard phenol-chloroform method and combed. Individual DNA fibers were revealed with the intercalating agent Yoyo-1 iodide (Fig. 1A), and 1,268 stained fibers were then measured. The majority of the fibers had lengths between 15 and 20 μm (30 and 40 kb, using the elongation factor of 2 kb/μm) (Fig. 1C). With a median size of 17 μm (34 kb), only 0.9% of the measured DNA molecules were over 70 μm (140 kb), indicating that full-length HSV-1 genomes were rare. These results suggest that the majority of DNA molecules are partly fragmented by several factors, for example, as a consequence of uncontrolled forces during phenol-chloroform extraction or the activity of nucleases. DNA extraction from viral particles embedded in agarose plugs, as described by Schurra and Bensimon (48), did not lead to detection of Yoyo-1 iodide-stained DNA molecules (data not shown), suggesting that the HSV-1 particles are not lysed within the agarose plug. We slightly modified the extraction method by decreasing the concentration of low-melting-point agarose to 1% and by performing lysis of viral particles in two consecutive lysis buffers containing 1% SDS (instead of 1% sodium N-lauroylsarcosinate) and 2 mg/ml of proteinase K, respectively. As shown in Fig. 1C, the analysis of 2,268 individual DNA molecules revealed a different type of DNA fiber length distribution than did phenol-chloroform extraction—the median size was 42 μm (84 kb), with 4% of DNA fibers having a length over 70 μm (140 kb). As 2.5% of the DNA fibers had a size greater than 76.5 μm (153 kb), this suggests that the viral-particle preparations were contaminated with cell genomic DNA.
The combed DNA fibers were then cohybridized with a set of 41 digoxigenin-11-dUTP-labeled cloned fragments that covered the entire HSV-1 genome. Continuous Alexa Fluor 594-fluorescent hybridization signals were obtained, showing that the observed signals corresponded to HSV-1 DNA (Fig. 1B). Measurement of 2,309 linear fluorescent signals yielded a median size of 39 μm (78 kb) (Fig. 1C). This time, 3% of the individual fibers were over 70 μm (140 kb). It was noted that 1.5% of the HSV-1 DNA fibers had lengths greater than 76.5 μm (153 kb), suggesting that the viral-particle preparations are partially contaminated with HSV-1 concatemeric DNA from infected cells. In conclusion, the data show that the method of extraction was improved in terms of DNA length, allowing analysis of the HSV-1 genome structure in viral particles by molecular combing.
In order to distinguish the four isomers of the HSV-1 genome, the 41 HSV-1 probes were labeled with either digoxigenin-11-dUTP or biotin-14-dCTP and detected with Alexa Fluor 350/AMCA- and Alexa Fluor 594-coupled antibodies, respectively, to create six groups of probes, referred as H1 to H6, that differ in size and color (Fig. 2). This constitutes the HSV-1-specific genomic Morse code. The order of the 6 probes within the genomic Morse code differs for each isomer; for example, the expected H1H2H3H4H5H6 fluorescent array corresponds to a P isomer, while an IL isomer is revealed by the theoretical H3H2H1H4H5H6 fluorescent array. Immunofluorescence microscopy of hybridized combed DNA extracted from particles of HSV-1 strain SC16 displayed a large diversity of multicolor linear patterns. Figure 2 shows an example of each full HSV-1 genome isomer that was successfully detected. Nevertheless, most of the canonical signals detected (95%) were found to be fragmented. However, they can be considered for further analysis, since we determined the minimal genomic Morse code pattern necessary to identify an isomer (Table 2). For example, a continuous signal composed of a 10-kb-long AMCA/Alexa Fluor 350 signal followed by a 21.5-kb-long Alexa Fluor 594 fluorescent signal, a 13-kb-long AMCA/Alexa Fluor 350 signal, and a 7.5-kb-long Alexa Fluor 594 fluorescent signal corresponds to a fragment of H2 followed by H1H4H5 and thus can only be interpreted as an IL isomer. On 1,082 measurements, the mean sizes of the H1, H3, H4, and H6 probes were 21.3 ± 1.46 kb (theoretical 21.5 kb), 45.5 ± 2 kb (theoretical 45.5 kb), 12.7 ± 0.7 kb (theoretical 13 kb), and 7.8 ± 0.5 kb (theoretical 7 kb), respectively, indicating that the designed genomic Morse code detects HSV-1 genomes that structurally correspond to what was theoretically expected. All these observations indicate that the designed genomic Morse code can discriminate the four HSV-1 genome isomers at the single-molecule level. To obtain the distribution of the isomers from viral-particle extracts of the HSV-1 strains SC16 and KOS produced in BSR cells, 405 fluorescent arrays that fulfilled the criteria for attribution of an isomer were counted. In both cases, the distributions were found to be equivalent (Fig. 3A), with chi-square P values of 0.96 and 0.80, respectively. Those viral-particle preparations were designated stocks. They were subsequently used to infect various cell lines and to analyze the HSV-1 isomer distribution during productive infection.
In order to perform a systematic analysis of the distributions of the four isomers of the HSV-1 genome, stocks of strains SC16 and KOS were used to infect Vero, COS-7, BSR, and Neuro-2a cells at a multiplicity of infection of 0.5. The distributions of the four genome isomers were determined for one genome equivalent from cell extracts and the corresponding supernatants. In the cell extracts, the distribution was analyzed globally, as HSV-1 genomes may have various origins: HSV-1 DNA that has just entered cells, fragmented concatemeric DNA, residues of A capsids that lack the CCSC complex (UL25/UL17 proteins) that failed to retain DNA (60), and encapsidated HSV-1 DNA. Cell supernatants contained the viral particles that had been released from infected cells. As shown in Fig. 1C, the preparations of cell supernatants are slightly contaminated with HSV-1 DNA longer than one genome equivalent. Here again, the analysis of the distribution was performed on the overall population. For strain SC16, both particles and infected cell samples showed that the frequencies of the four HSV-1 isomers were about 25% (Fig. 3B and andC),C), indicating that the distributions were statistically equivalent (chi-square P value, between 0.6 and 0.9). For the KOS strain, the frequencies of the four isomers were also about 25% for cell and cell supernatant DNA extracts of the COS-7 and BSR cells. On the other hand, the P and IS isomers were overrepresented in Vero and Neuro-2a cell extracts, with proportions between 30% and 35%. These differences in the distributions of the four HSV-1 isomers were statistically confirmed, as the P values of the chi-square tests were 6 × 10−8 and 10−4 with Vero and Neuro-2a cells, respectively (Fig. 3D). A similar statistically unbalanced distribution was noticed in the corresponding viral particles recovered from the culture supernatant. P and IS were again overrepresented, with proportions between 28% and 31%, and the chi-square P values were 6 × 10−3 and 3 × 10−4 with Vero and Neuro-2a cells, respectively (Fig. 3E). This means that both P and IS viral genome isomers from HSV-1 genomes in infected cells and from released viral particles are more highly represented than the IL and ILS genome isomers. Thus, the process of isomerization seems not to systematically result in a random distribution.
In infected Vero, COS-7, BSR, and Neuro-2a cells, fluorescent signal arrays longer than 152 kb (likely corresponding to intermediates of replication) were detected, in addition to the canonical HSV-1 fluorescent signals (one-genome-equivalent elements with the usual a-TRL-UL-IRL-a′-IRS-US-TRS-a organization). Fluorescence hybridization was performed using the genomic Morse code labeled with digoxigenin-11-dUTP on combed DNA from noninfected Vero cells and SC16-infected Vero cell DNA extracts. As expected, in noninfected Vero cells, used as a control, no fluorescence hybridization signals were detected, indicating that the probes do not cross-hybridize with host genome DNA and are specific to the HSV-1 genome (data not shown). Continuous Alexa Fluor 594-fluorescence hybridization signals were detected in infected-cell DNA extracts. The measurement of 1,031 signals revealed the presence of HSV-1-positive DNA fibers as long as 1,393 kb. Fibers were grouped into bins of one-genome-equivalent elements (76.5 μm or 153 kb) (Fig. 4A and Table 3). The majority of fibers (62%) was less than 153 kb and 29% were between 153 and 306 kb, corresponding to two genome equivalents, while 3.6% were between 306 and 459 kb long, as expected for three genome equivalents. The remaining 5.4% were between 459 and 1,530 kb, i.e., 3 to 10 genome equivalents. It is notable that fibers of three genome equivalents or more were detectable and were possibly underestimated, considering the breakage during the DNA extraction or the molecular combing processes, probably due to the presence of nicks and gaps reported in replicative HSV-1 DNA (17).
Using the validated genomic Morse code, the isomer distribution was analyzed within the concatemers that were produced in Vero, COS-7, BSR, and Neuro-2a cells infected with either strain SC16 or KOS of HSV-1. As during the encapsidation process unit-length genomes are cleaved in the a sequence, and as it is not possible to orient the DNA molecules, there are two possibilities for cleavage per molecule, generating two possible sets of isomers per molecule (Fig. 4B). The frequencies of the four isomers were about 25%, even for concatemers from strain KOS-infected Vero and Neuro-2a cell DNA extracts (Fig. 4C and andD).D). These observations suggested that the nonequivalent distributions of the one genome equivalents produced in cells or cell supernatants from KOS-infected Vero or Neuro-2a cells likely occur throughout encapsidation rather than during homologous recombination between the inverted repeats surrounding the UL and the US fragments.
During the analysis of isomer distribution, we noticed several types of unconventional DNA structures. First, there were individual molecules for which the sequence of the H1 to H6 probes corresponded to one isomer but with unexpected length or order of at least one of the probes. For example, the 104-kb AMCA/Alexa Fluor 350 array for the first molecule in Fig. 5A is 47 kb longer than the longest expected blue H2 signal. A second example, typified by the second molecule in Fig. 5A, corresponds to a 75-kb AMCA/Alexa Fluor 350 array disconnected by a 5-kb Alexa Fluor 594 signal. These examples suggest events like internal duplications, deletions, or rearrangements. Third, several fibers showed regular repeats (Fig. 5A) for which a 17.8-kb repetition composed of a 5.8-kb AMCA/Alexa Fluor 350 florescence signal and a 12-kb Alexa Fluor 594 signal are observed. Their repetitive structure is similar to those described by Schroder et al. (47) as parts of viral genomes built up from repeats of restricted regions of the standard genome and designated repetitive interfering particles. Among the genomes extracted from viral particles, the ratio between canonical and noncanonical events varied between viral strains and could be as high as 47% for the KOS strain of HSV-1 (Fig. 5B, ,C,C, and andDD).
To determine whether nonrandom isomer distributions are also observed in vivo, total DNA was extracted from corneas and trigeminal ganglia from mice infected with the SC16 or KOS strain. The analysis of the distributions of the canonical HSV-1 genomes revealed that the ILS isomer predominated with both strains and in both cornea and trigeminal ganglion tissues. This was never found in vitro. Although the numbers of isomers were low in both SC16- and KOS-infected corneas (n = 209 and n = 115, respectively), their distributions were random (P = 0.06 and P = 0.6, respectively). The distributions were similarly random in the trigeminal ganglion extracts (P = 0.06) (Fig. 6A) from mice infected with strain SC16, whereas the ILS isomers were overrepresented (frequency, 34%) in trigeminal ganglion extracts from mice infected with the KOS strain. This led to statistically nonequivalent distributions with a chi-square P value of 6 × 10−4 (Fig. 6B). This nonrandom distribution was distinct from those observed in in vitro experiments. Concatemeric DNA was rarely detected in infected cornea and trigeminal ganglion DNA extracts, suggesting a low rate of replication at 6 days postinoculation. As observed in vitro, 20% and 35% of the genomes detected in the corneas and trigeminal ganglia corresponded to noncanonical forms (Fig. 6C), suggesting that HSV-1 genome replication leads to relatively high levels of aberrant genomes that are probably noninfectious.
The distribution of HSV-1 genome isomers is commonly thought to be random (12, 13, 19). While herpesviruses genomes have been studied by electron microscopy (13) and Southern blotting (2) for many years, the uniform stretching provided by molecular combing, which essentially pulls genomes into fibers like molecules that can be readily detected by fluorescent probes (36), coupled with direct visualization by differently labeled probes provides unprecedented resolution for the statistical analysis of the genome isomer distribution. The mean size of DNA molecules extracted from virions was around 30 kb after phenol-chloroform extraction (Fig. 1C), much less than one genome equivalent. To allow the analysis of the distribution of the four HSV-1 isomers, the method of extraction was improved, reaching ~70 to 80 kb as an average fiber length for DNA extracted from virions in agarose blocks. However, it was still well below one genome equivalent. This may have several origins that are not mutually exclusive. First, it has been reported that members of the Herpesviridae randomly produce partial genomes during their replication (22, 47). If those elements, called defective interfering particles, contain packaging signals, they are encapsidated and thus create interference during serial infection. Indeed, at a given time, more copies of the fragments than of complete genomes are reproduced, leading to a reduction in infectious particles. It is then probable that part of the viral particles extracted corresponded to accumulated defective interfering particles. Another source of HSV-1 genome fragmentation could be the activities of endonucleases, which may have cleaved concatemers from lysed infected cells. A recent work has also shown that HSV-1 genomes are susceptible to cytidine deamination by APOBEC3C (58), a reaction very rapidly followed by uracil removal by uracil N-glycosylase, and it is possible that packaged DNA might have some abasic sites. Finally, it has been reported that virion and replicative HSV-1 DNAs contain frequent nicks and gaps (3, 17, 26, 43, 52, 66). During DNA extraction and/or molecular combing, these sites might prove to be fragile, giving rise to large numbers of subgenome fragments. What is clear is that phenol-chloroform DNA extraction results in considerable degradation of HSV-1 genomes, meaning that prior studies relying on the quality of HSV-1 DNA may have led to some erroneous findings.
The variety of HSV-1 DNA forms detected was considerable. In cell extracts, it represented subgenomic fragments, concatemers, and noncanonical genomes. The isomer distribution was uniquely addressed in the context of one genome equivalent and concatemers. The isomer distribution was found to be equimolar in vitro when using the strain SC16. In contrast, statistically significant overrepresentation of P and IS isomers was observed in cell and cell supernatant DNA extracts from a culture of Vero and Neuro-2a cells infected with strain KOS. Thus, the distribution was found to be host cell and strain dependent in vitro, suggesting a cell- and strain-dependent mechanism. As equivalent distributions were noticed in KOS concatemeric DNA in vitro in all cell lines, the unbalanced distribution observed in viral particles and cell extracts was likely related to the encapsidation process. The isomerization mechanism has been previously linked to the encapsidation process, as the alternative cleavage of concatemeric molecules is part of the mechanism leading to the four genome isomers (56, 62). Even if the L and S termini both contain the signals for DNA encapsidation (57), only L termini are observed in concatemeric DNA, suggesting that encapsidation starts at the L termini and proceeds toward S (34, 69). Here, the unbalanced distribution could be linked to increased efficiency in the cleavage of the a sequence toward P and IS isomers or a predisposition to degradation of IL and ILS isomers after cleavage. In vivo, the distributions were found to be random, except with strain KOS, where the ILS isomers are overrepresented in trigeminal ganglia, which was distinct from what was observed in vitro. Statistically unbalanced distributions were uniquely observed with strain KOS, a less virulent strain than SC16 (6). The KOS strain of HSV-1 is known to be modified in the ICP34.5 gene (39), which encodes a neurovirulence factor that maps in the TRL /IRL inverted repeats (15, 40). This, combined with the observation of the balanced distribution of putative isomers in concatemeric DNA for both SC16 and KOS, suggests that the genomic and/or epigenomic organization of the TRL/IRL regions could affect the isomerization process, likely at the encapsidation step. Finally, both in vitro and in vivo, the frequency of one isomer reached 40% from the null distribution, suggesting that the tendency is not to deviate too far from an equimolar distribution. However, the continued unbalanced distributions observed using a single-molecule approach suggest a controlled process. It is notable that molecular combing allowed the measurement of replication intermediates. It has been shown previously that most of the concatemeric DNA was comprised of 2 or 3 genome equivalents (14), which was confirmed here (Table 3). We also observed replication intermediates of up to 10 genome equivalents. With the power to visualize different regions of the genome, we could identify the composition of the concatemers, which suggests recombination. As they could represent ligation of monomers, dimers, or trimers or reflect nuclease isomerization within the concatemers, it is not possible to formally distinguish between the two models currently proposed to describe HSV-1 DNA replication, i.e., rolling-circle (7) and recombination-dependent replication (67) mechanisms.
We also observed large proportions of noncanonical isomers in vitro (30 to 60%). A large variety of patterns were detected. Overall, as suggested previously (59), recombination between HSV-1 genomes is a major event in the replication process allowing the creation of new genetic assortments and thus may play an important role in the evolutionary process. In vivo, the noncanonical assortments were less abundant than in vitro, which suggests that some mechanisms may occur to optimize the efficacy of genome assortment and to improve the quality of encapsidated viral genomes. Moreover, the proportion of noncanonical forms was 10 to 20% higher in trigeminal ganglia than in corneas. At this stage of infection, i.e., 6 days after viral inoculation by corneal scarification, the cornea is markedly infected for several days, and most of the detected HSV-1 genomes are likely already encapsidated, if not contained in mature viral particles. In contrast the trigeminal ganglia had been reached by viral particles for only 1 or 2 days, and the replicative infection within the neuronal somata was less complete than in the cornea (data not shown). Thus, most of the HSV-1 genomes detected were likely undergoing replication, which may account for our results.
While the success of HSV-1 has never been in doubt, it seems to be achieved at a high cost in terms of DNA synthesis. While experimental factors reduce the genome size to less than one unit length, the sheer proportions of subgenomic molecules and noncanonical forms is a little surprising. They may reflect anthropocentric perceptions, as the particle- to-PFU ratio for HSV-1 is on the order of 100:1. The present data introduce another cost aspect in viral ecology, the accuracy of DNA replication of large genomes, which does not show up in particle-to-PFU ratios. By extrapolation, it is tempting to suggest that a similar situation may pertain to the replication of other herpesvirus genomes, most of which are larger than that of HSV-1. It is an open question whether genome replication of DNA viruses is similarly affected, but unique molecular combing has the potential to explore this question.
This work is part of the project ACTIVE, which is supported by the Strategic Industrial Innovation program of OSEO (OSEO-ISI). C.M is supported by a grant from the Association Nationale de la Recherche et de la Technologie (ANRT).
Published ahead of print 6 June 2012