|Home | About | Journals | Submit | Contact Us | Français|
Intermolecular recombination between the genomes of closely related RNA viruses can result in the emergence of novel strains with altered pathogenic potential and antigenicity. Although recombination between flavivirus genomes has never been demonstrated experimentally, the potential risk of generating undesirable recombinants has nevertheless been a matter of concern and controversy with respect to the development of live flavivirus vaccines. As an experimental system for investigating the ability of flavivirus genomes to recombine, we developed a “recombination trap,” which was designed to allow the products of rare recombination events to be selected and amplified. To do this, we established reciprocal packaging systems consisting of pairs of self-replicating subgenomic RNAs (replicons) derived from tick-borne encephalitis virus (TBEV), West Nile virus (WNV), and Japanese encephalitis virus (JEV) that could complement each other in trans and thus be propagated together in cell culture over multiple passages. Any infectious viruses with intact, full-length genomes that were generated by recombination of the two replicons would be selected and enriched by end point dilution passage, as was demonstrated in a spiking experiment in which a small amount of wild-type virus was mixed with the packaged replicons. Using the recombination trap and the JEV system, we detected two aberrant recombination events, both of which yielded unnatural genomes containing duplications. Infectious clones of both of these genomes yielded viruses with impaired growth properties. Despite the fact that the replicon pairs shared approximately 600 nucleotides of identical sequence where a precise homologous crossover event would have yielded a wild-type genome, this was not observed in any of these systems, and the TBEV and WNV systems did not yield any viable recombinant genomes at all. Our results show that intergenomic recombination can occur in the structural region of flaviviruses but that its frequency appears to be very low and that therefore it probably does not represent a major risk in the use of live, attenuated flavivirus vaccines.
RNA viruses are able to undergo rapid genetic changes in order to adapt to new hosts or environments. Although much of this flexibility is due to the error-prone nature of the RNA-dependent RNA polymerase, which generates an array of different point mutations within the viral population (23), recombination is also a common and important mechanism for generating viral diversity (18, 31, 42, 58). Recombination occurs when the RNA-dependent RNA polymerase switches templates during replication, an event that is favored when both templates share identical or very similar sequences. Three types of RNA recombination have been identified: homologous recombination occurs at sites with exact sequence matches; aberrant homologous recombination requires sequence homology, but crossover occurs either upstream or downstream of the site of homology, resulting in a duplication or deletion; and nonhomologous (or illegitimate) recombination is independent of sequence homology (31, 42).
When the same cell is infected by viruses of two different strains, or even different species, recombination between their genomic RNAs can potentially lead to the emergence of new pathogens. A case in point is the emergence of Western equine encephalitis virus, a member of the genus Alphavirus, family Togaviridae, which arose by homologous recombination between Eastern equine encephalitis virus and Sindbis virus (14).
Some mammalian RNA viruses can recombine at a frequency that is detectable in experimental settings (1, 2, 55), and phylogenetic analysis of partial or complete genome sequences suggests that RNA recombination is a widespread phenomenon. Naturally occurring recombinant viruses have been identified in almost every family of positive-stranded RNA viruses (31, 58).
Flaviviruses are members of the genus Flavivirus, family Flaviviridae, a family that also includes the genera Pestivirus and Hepacivirus. Several of the flaviviruses are important human pathogens, such as Japanese encephalitis virus (JEV), West Nile virus (WNV), the dengue viruses, yellow fever virus, and tick-borne encephalitis virus (TBEV).
Although there has never been a report of a pathogenic flavivirus strain arising due to recombination involving attenuated vaccine strains (39), the urgent necessity to develop tetravalent vaccines containing all four serotypes of dengue virus—two such vaccines are currently undergoing clinical testing (45)—has recently brought the recombination issue to the forefront of discussion among researchers, regulators, and vaccine producers (39). It has been suggested that recombination, either between the strains present in a multivalent vaccine or between an attenuated vaccine strain and a wild-type strain, could lead to the emergence of new viruses with unpredictable properties (49).
So far, recombination between flavivirus genomes has not been demonstrated directly in the laboratory. However, phylogenetic analysis of partial genome sequences available in the GenBank database has suggested that homologous recombination may have occurred between closely related strains of dengue virus (20, 52, 54, 59). An experimental approach for assessing the ability of flavivirus genomes to recombine is therefore urgently needed.
Flavivirus virions are composed of a single-stranded, positive-sense RNA genome that, together with the capsid protein C, forms the viral nucleocapsid. The nucleocapsid is covered by a lipid envelope containing the surface glycoproteins prM and E. These glycoproteins drive budding at the membrane of the endoplasmic reticulum during the assembly stage and mediate entry of the virus into host cells (41). Replicons, defined as self-replicating, noninfectious RNA molecules, can be generated by deleting parts or all of the region coding for the structural proteins C, prM, and E from the viral genome but maintaining all seven of the nonstructural proteins and the flanking noncoding sequences, which are required in cis for RNA replication (25). By providing the missing structural protein components in trans, replicons can be packaged into virus-like particles that are capable of a single round of infection (10, 15, 24, 47).
Typically, researchers developing novel replicating vaccines, especially ones that involve multiple components, make an effort to come up with strategies to prevent recombination, for example by “wobbling” codons, i.e., replacing codons in homologous regions with synonymous ones encoding the same amino acid but consisting of a different nucleotide triplet (50, 57). In this study, in order to assess the propensity of flavivirus genomes to recombine, we took an opposite approach, establishing a “recombination trap” that favors the selection and sensitive detection of recombination products. This system takes advantage of the ability of replicon pairs containing deletions in their structural protein genes to complement each other in trans and thus be propagated together in cell culture, and by passage at limiting dilutions, it allows infectious RNA genomes arising by recombination between the two replicons to be preferentially selected.
Using the recombination trap, we have now obtained the first direct evidence of recombination between flavivirus genomes in the laboratory. Aberrant homologous recombination was observed twice with JEV replicons, resulting in viruses with unnatural gene arrangements and reduced growth properties compared to those of wild-type JEV. No infectious recombinants of any kind were obtained when TBEV or WNV replicons were used. Interestingly, we never detected a fully infectious wild-type genome arising by homologous recombination in any of these systems. The results of this study show that the propensity of flavivirus genomes to recombine in the region coding for the structural proteins appears to be quite low, suggesting that recombination does not represent a major risk in the use of live, attenuated flavivirus vaccines.
BHK-21 cells were grown in Eagle's minimal essential medium (Sigma) supplemented with 5% fetal calf serum (FCS), 1% glutamine, and 0.5% neomycin (growth medium) and maintained in Eagle's minimal essential medium supplemented with 1% FCS, 1% glutamine, 0.5% neomycin, and 15 mM HEPES, pH 7.4 (maintenance medium). Vero cells (ATCC CCL-81) were grown in Eagle's minimal essential medium supplemented with 10% FCS, 30 mM l-glutamine, 100 units of penicillin, and 1 μg/ml streptomycin. Plaque assay infections were done in medium containing 1% FCS. Aedes albopictus C6/36 cells were grown in Eagle's minimal essential medium (without NaHCO3) supplemented with 10% FCS, 20 mM l-glutamine, 100 units of penicillin, 1 μg/ml streptomycin, 13 mM sodium hydroxide, 19 mM HEPES (pH 7.4), and 0.2% 50× tryptose-phosphate. For growth curve analysis, the FCS concentration was reduced to 1%.
Plasmid pTNd/c, used for generating infectious TBEV, contains a full-length genomic cDNA insert of TBEV strain Neudoerfl (GenBank accession number U27495) cloned in plasmid pBR322 (34). RNA transcribed from this plasmid was used as the TBEV wild-type virus control. Plasmid pTNd/5′ contains a 5′ cDNA fragment of the same viral genome (34). Full-length DNA templates for in vitro transcription of TBEV plasmids were generated as described previously (26).
pTBEV-ΔC, a derivative of pTNd/c containing a large deletion (62 amino acids) of the sequence coding for the capsid protein, as well as individual mutations in the signal sequence of the capsid protein, was used as the DNA template for in vitro synthesis of replicon ΔC. It has been described in an earlier publication (25) in which it was called C (Δ28-89)-S. The plasmids for production of replicons ΔME (10) and ΔME-eGFP (11) were also described previously.
Plasmid pTBEV-ΔE, the template for generating TBEV replicon ΔE, was constructed by first performing PCR using the primers 5′-TTTTACCGGTTTACGCTGATGTTGGTTGCGCTGTGGA-3′ and 5′-TTCCATCGATAGTGTGACTAGCAGGCCATGAGCA-3′ with pTNd/5′ as a template and then cloning this PCR product into pTNd/c using the restriction enzymes AgeI and ClaI.
For the construction of plasmid pWNV-ΔC, the template for generating the WNV ΔC replicon, two DNA fragments were made by amplifying portions of the plasmid clone pWNV-K1 (46), which contains nucleotides 1 to 3339 of WNV strain NY99 (isolate Crow V76/1). The first fragment was made by using the primers 5′-AGGTGTTCCACAGGGTAGCCA-3′ and 5′-TTTTGGATCCTTTTAGCATATTGACAGCCC-3′ and then digesting the resulting product with PacI and BamHI. The second fragment was made using primers 5′-TCTCGGATCCTCAAAACAAAAGAAAAGAGG-3′ and 5′-AAATAGGGGTTCCGCGCACA-3′ and digestion with BamHI and NotI. Ligation of these fragments resulted in an intermediate construct containing nucleotides 1 to 3339 of the WNV genome but lacking nucleotides 151 to 393 in the C gene, which were replaced by a BamHI restriction site. An additional SalI restriction site was created by introducing silent nucleotide substitutions at positions 151 and 156 using a site-directed mutagenesis kit (Invitrogen) with the primers 5′-GTGTCTGGAGCAACATGGGTcGAcTTGGTTCTCG-3′ (bold lowercase letters indicate mutated nucleotides) and 5′-ACCCATGTTGCTCCAGACACTCCTTCCAAG-3′. The full-length DNA template for RNA synthesis was generated by digestion of pWNV-K1-ΔC and pWNV-K4 with BstEII, followed by in vitro ligation using T4 DNA ligase (Invitrogen) and linearization by digestion with NotI.
For construction of plasmid pWNV-ΔE, pWNV-K1 was used as a template for mutagenic PCR using the primers 5′-TTTTCACCCAGTGTCGCTGTAAGCTGGGGCCACCA-3′ and 5′-TTTTAGATCTCGATGTCTAAGAAACCAGGAGG-3′. Restriction digestion of the PCR product and WNV-K1 with BglII and AdeI generated fragments that were ligated to form pWNV-K1-ΔE. This partial clone was further digested with PacI and BstEII, and the resulting fragment was cloned into pWNV-K4, yielding full-length DNA plasmid pWNV-ΔE. This full-length plasmid was linearized by NotI digestion before use as a template for in vitro RNA transcription.
All constructs containing JEV sequences were based on JEV strain SA-14 (China), which was kindly provided by Peter Mason, Yale University. A 5′ clone, pJEV-K3, containing nucleotides 1 to 5616, a 3′clone, pJEV-K4, containing nucleotides 5595 to 10977, and a full-length clone, pJEV/c, containing the whole genome (nucleotides 1 to 10977) of the wild-type virus in pBR322, were generated using standard cloning techniques (unpublished data).
For construction of pJEV-ΔC, mutagenic PCR was first performed using the primers 5′-GGCATCGATTAGTGGGAATACGCGGGGTAG-3′ and 5′-AAATTAATTAATACGACTCACTATAGAGAA-3′ with plasmid pJEV-K3 as the template. This PCR product and plasmid pJEV-K3 were both digested with ClaI and PacI and ligated, forming the intermediate construct pJEV-K3-ΔC. pJEV-K3-ΔC and pJEV/c were then digested with PacI and AgeI, and the smaller fragment of pJEV/c was replaced by the corresponding fragment of pJEV-K3-ΔC, yielding the full-length cDNA clone pJEV-ΔC.
For construction of pJEV-ΔE, mutagenic PCR was performed using the primers 5′-AGGCAGCCCCTAGGACCAGAACCACGTTTTCTTGGTTCGT-3′ and 5′-TTTTACGCGTGGTATTTACCATCCTCCTGCTGTTGGTCGCTCCGGCTTACAGTGACACTTGGATGTGCCATTG-3′ and plasmid pJEV-K3 as the template. This PCR product and plasmid pJEV-K3 were both digested with AvrII and MluI and ligated together to make the intermediate construct pJEV-K3-ΔE. pJEV-K3-ΔE and pJEV/c were then digested with PacI and BamHI, and the smaller fragment of pJEV/c was exchanged with the corresponding pJEV-K3-ΔE fragment, yielding the full-length cDNA clone pJEV-ΔE.
Infectious clones of JEV recombinants 1 and 2 were made as follows. Cytoplasmic RNA was isolated from BHK-21 cells that contained recombinant genomes and was transcribed into cDNA as described below for reverse transcription-PCR (RT-PCR). Then, PCR with primers 5′-TCGAGAGATTAGTGCAGTTT-3′ and 5′-CAGTACGACAAGTCACTATGGAC-3′ was used to generate a fragment that was digested with ClaI and AgeI and exchanged with the corresponding ClaI/AgeI fragment of pJEV/c to form the full-length DNA templates for in vitro transcription.
In vitro transcription and transfection of BHK-21 cells by electroporation were performed as described previously (28, 34, 43). RNA was synthesized from full-length cDNA clones or in vitro-ligated full-length templates (see above) using reagents of the T7 Megascript kit (Ambion) according to the manufacturer's protocol. The template DNA was digested by incubation with DNase I, and the quality of the RNA was checked by electrophoresis in a 1% agarose gel containing 6% formalin. RNA was purified using an RNeasy minikit (Qiagen) and quantified spectrophotometrically. Equimolar amounts of RNA were then introduced into BHK-21 cells by electroporation using a Bio-Rad gene pulser (1.8 kV, 25 μF, 200 Ω). For cotransfections, equal amounts of RNA (~1.1 × 1012 copies) of each construct were mixed before electroporation.
Twenty-four hours after transfection or infection, BHK-21 cells were fixed for 20 min with 4% paraformaldehyde in phosphate-buffered saline (PBS) (pH 7.4) and then washed two times for 5 min with PBS (pH 7.4) and permeabilized with ice-cold methanol for 6 min at −20°C. After three washes with PBS (pH 7.4) and blocking with 3% bovine serum albumin in PBS (pH 7.4) for 30 min at room temperature, intracellular E protein was visualized by sequential incubation with a monoclonal mouse antibody against E (1E9) (12) and a Rhodamine Red-X-conjugated anti-mouse immunoglobulin G antibody (Jackson Immune Research Laboratory). Simultaneously, enhanced green fluorescent protein (eGFP) fluorescence was enhanced by incubation with a polyclonal rabbit antibody against GFP (ab6556; Abcam), followed by incubation with a fluorescein isothiocyanate (FITC)-conjugated anti-rabbit immunoglobulin G antibody (Jackson Immune Research Laboratory). The fraction of transfected cells was determined by counting GFP- and/or Rhodamine Red-X-positive cells in multiple defined areas of the microscopic field. Counts for different fields were averaged and used for determination of the transfection efficiency.
For single-focus immunofluorescence, BHK-21 cells were seeded into 24-well tissue culture plates containing microscope coverslips. The confluent monolayer was infected with 10-fold dilutions of supernatant harvested from BHK-21 cells 3 days after transfection of wild-type virus or cotransfection with replicons ΔC and ΔME-eGFP. Then, cells were covered with a 3% carboxymethyl cellulose overlay dissolved in maintenance medium. After 50 h, cells were washed three times with ice-cold PBS (pH 7.4) and processed for double immunofluorescence staining as described above. Microscopic observation and documentation were accomplished by using a LSM 510 scanning confocal microscope (Zeiss) and the included software.
Supernatants from BHK-21 cells were collected at passage 1 on day 3 postinfection. Serial 10-fold dilutions of supernatants were applied to confluent monolayers of BHK-21 cells. After a 3-h incubation, supernatants were removed and cells were covered with a 3% carboxymethyl cellulose overlay dissolved in maintenance medium. Fifty hours after infection, cells were fixed with acetone-methanol (1:1) for 10 min at −20°C and treated with polyclonal rabbit anti-TBEV serum. Antibody-labeled cells were detected using an immunoenzymatic reaction consisting of sequential incubation with goat anti-rabbit immunoglobulin G-alkaline phosphatase and the corresponding enzyme substrate (SigmaFast Red TR/Naphthol AS-MX tablets).
RNA replication was analyzed by quantitative real-time PCR (qPCR) as described previously (28). Briefly, total intracellular RNA was isolated from transfected cells at different time points and transcribed into cDNA. Viral RNA was then quantified using the following primers and probes: forward primer, GAAGCGGAGGCTGAACAACT; reverse primer, TTGTCACGTTCCGTCTCCAG; and probe, TGTGTACAGGCGCACCGGCA. A prM-cleavage-defective mutant that replicates to wild-type levels, ΔR88 (8), and a nonreplicating mutant, ΔNS-5 (28), were used as positive and negative controls, respectively.
Supernatants of infected cells were sequentially diluted, and 200 μl of each dilution was applied to fresh BHK-21 cells grown in 24-well plates (Nunc). After 3 days, half of the medium was replaced by fresh medium, and after 6 days, protein E released into the supernatant was detected using a four-layer enzyme-linked immunosorbent assay (16) in the case of TBEV and a hemagglutination assay (HA) (46) in the case of WNV and JEV. After 6 days, the highest dilution yielding a positive signal in enzyme-linked immunosorbent assay or HA was cleared of cell debris, sequentially diluted, and used for infection of fresh BHK-21 cells.
Template DNA for synthesis of RNA probes for Northern blotting was generated using the following primers and templates. For TBEV, template pTNd/c and primers AGAGAGCAGAAGGGATTGA and TACTTAATACGACTCACTATAGGTGTGCAAGACACCCTTG generated a probe binding to nucleotides 4093 to 4773 in wild-type virus. For WNV, template pWNV-K4 and primers AGCGGCTGTTGGTATGGTATG and TAATACGACTCACTATAAGCTGCACTCCTCTTCTCCCT generated a probe binding to nucleotides 3448 to 4107 in wild-type virus. For JEV, template pJEV-K4 and primers AATGGCTGCTGGTACGGAATGGA and TAATACGACTCACTATACATGGTCTTTTTCCTCTCGTG generated a probe binding to nucleotides 3456 to 4103 in wild-type virus. After phenol-chloroform purification of the template DNA, the T7 promoter sequence (underlined) fused to each reverse primer allowed RNA synthesis using a T7 MAXIscript in vitro transcription kit (Ambion). The probe RNA was labeled by using 0.4 μl of 10 mM Bio-11-UTP (Ambion) instead of UTP. Probe RNA was separated from free nucleotides using Micro Bio-Spin 30 columns (Bio-Rad).
Cytoplasmic RNA was extracted from BHK-21 cells 24 h after transfection or infection. Ten μg of total RNA was applied to a 1% agarose gel, and Northern blotting was carried out according to the instructions of the NorthernMax-Gly kit (Ambion). Blotted RNA was detected using a BrightStar BioDetect kit (Ambion).
RT-PCR was performed using the same cytoplasmic RNA used for the Northern blot analysis, with the primers for cDNA synthesis and PCR listed in Table Table1.1. cDNA fragments of the 5′-terminal part of each virus genome, including the entire sequence coding for the three structural proteins capsid, prM, and E, were synthesized using a reverse primer binding to the sequence coding for NS1 of each virus (Table (Table1).1). Subsequently, PCR was carried out using primer pairs designed to detect recombinants (Table (Table1).1). Each PCR was carried out using an Advantage HF2 PCR kit (Clontech), and amplicons were subjected to 1% agarose gel electrophoresis as well as automated sequence analysis as described before (28).
Vero cells were grown to 80% confluence in 12-well plates and incubated for 1 h with virus suspensions serially diluted in maintenance medium. The cells were subsequently overlaid with Eagle's minimal essential medium containing 5% FCS (PAA), 1.5% glutamine (200 mM; Cambrex), 1% penicillin/streptomycin (10,000 U/ml penicillin, 10 mg/ml streptomycin; Sigma), 15 mM HEPES, and 0.25% agarose (Sigma). Four days after infection, the cells were fixed and stained with a solution containing 4% formaldehyde and 0.1% crystal violet.
Stock preparations of JEV recombinants were generated by collecting supernatants of BHK-21 cells transfected with these mutants at day 3 posttransfection. For analysis of growth properties of recombinant JEV, C6/36 mosquito cells grown in six-well plates were infected with wild-type JEV or JEV recombinant stock preparations at a low multiplicity of infection (MOI), 0.01. Aliquots of supernatants (500 μl) were taken at different time points, and virus titers were determined by plaque assay.
As a tool for testing the propensity of flaviviruses for recombination, we established a reciprocal packaging system to allow the cocultivation of two different defective viruses for prolonged passages. This system consisted of two replicons, each of which contained a deletion in a region of the RNA genome that encodes structural proteins that are essential for virus assembly. Individually, each of these replicons was incapable of forming infectious virus particles due to the lack of an essential component for virion assembly, but when present together in the cytoplasm of the same cell, they were capable of complementing each other in trans, thus allowing each of the defective genomes to be packaged and propagated as single-round infectious (or pseudoinfectious) particles and allowing copassage of these replicons in cell culture. By conducting passages at limiting dilutions, functional revertant wild-type virus genomes resulting from recombination between the two replicons could potentially be selected and enriched.
To set up the two-component trans-complementation system, we took advantage of two previously described and characterized TBEV replicons called ΔC (27) and ΔME-eGFP (11). As shown in Fig. Fig.1,1, replicon ΔC consisted of a full-length TBEV genome containing a 186-bp deletion in the region encoding the capsid (C) protein, and ΔME-eGFP lacked the entire region encoding the envelope proteins prM and E. In addition, ΔME-eGFP contained an additional artificial cistron inserted in the 3′ noncoding region that allowed the expression of the eGFP under the control of an internal ribosome entry site element.
To test the ability of these replicons to complement each other and to be packaged into infectious particles, BHK-21 cells were first transfected with each of the in vitro-synthesized, capped replicon RNAs, either individually or in combination. Transfected cells containing replicon ΔC were identified using an immunofluorescence assay with a monoclonal antibody recognizing the envelope protein E, which was expressed in ΔC but not in ΔME-eGFP. Using a rhodamine-conjugated secondary antibody for detection, the cells containing ΔC RNA appeared red under the fluorescence microscope (Fig. (Fig.2,2, left), as did control cells infected with the full-length wild-type genomic RNA, which also expressed the E protein. Transfected cells containing replicon ΔME-eGFP appeared green under the fluorescence microscope. Similar transfection efficiencies of more than 90% were observed with each of the constructs.
When BHK-21 cells were transfected by electroporation with an equal mixture of the two replicons (Fig. (Fig.2,2, bottom left), it was observed 1 day after transfection that approximately equal numbers of cells were positive for eGFP (green cells) and for E (red cells), and about 50% were positive for both (yellow cells in the merged image), indicating that they contained both ΔC and ΔME-eGFP.
Supernatants from each of the transfected cell cultures were then applied to fresh BHK-21 cells, and the cells were tested 24 h later for the presence of E and eGFP (Fig. (Fig.2,2, right). After one passage, the cells treated with the supernatant from the cotransfected culture (ΔC+ΔME-eGFP) again showed a mixed pattern of green, red, and yellow cells, indicating that each of the replicons had been packaged into infectious particles and that a subset of the cells had been coinfected with both replicons. As expected, the supernatants from cells transfected with defective ΔC or ΔME-eGFP RNA alone apparently did not produce infectious particles, and these replicons could not be propagated further. Also as expected, the supernatants from the positive control cells that had been transfected with full-length infectious genomic RNA (wild type) were able to infect fresh cells, as indicated by positive red staining for the viral E protein (Fig. (Fig.2,2, top right).
In this passage experiment, the presence of cells containing both ΔC and ΔME-eGFP suggested that the frequency of coinfection was high enough to enable a further round of trans-complementation and packaging of replicon RNAs. To test this, the supernatants from passage 1 were again transferred to fresh cells, and, as shown in Fig. Fig.22 (bottom right, Passage 2), a similar pattern was observed, with cells containing ΔC, ΔME-eGFP, or both. These experiments demonstrate that replicons ΔC and ΔME-eGFP, when used together, constitute a reciprocal packaging system that can be propagated in cell culture by serial passage.
Next, we infected fresh BHK-21 cells with various dilutions of supernatants containing the mixture of packaged ΔC and ΔME-eGFP replicons, added a methylcellulose overlay, and fixed and stained the cells 50 h later using an anti-TBEV polyclonal antibody and an alkaline phophatase-conjugated secondary antibody to detect the formation of infectious foci. As shown in Fig. Fig.3A,3A, the mixture of packaged replicons formed foci that were smaller than those formed by wild-type TBE virus, demonstrating that the ability of infectious material to spread from replicon-containing cells was impaired compared to that of the virus control. Furthermore, unlike wild-type virus, which showed a linearly proportional decrease in the number of plaques formed with increasing dilution, the number of foci formed with the packaged replicon mixture decreased very sharply with increasing dilutions (data not shown). Foci were observed only up to a dilution of 10−3, whereas infected cells could be detected by immunofluorescence staining up to a dilution of 10−5. This sensitivity to dilution is consistent with what would be expected if two particles, one containing ΔC and the other containing ΔME-eGFP, were required to make one infectious unit (i.e., one focus-forming unit).
To examine the foci more closely, cells were again infected with different dilutions of the ΔC and ΔME-eGFP mixture and overlaid with nitrocellulose as above, but instead of staining with polyclonal serum and alkaline phosphatase, the cells were fixed and prepared for detection of E and eGFP by fluorescence microscopy as described in the previous section. An example of a typical focus formed by infection with a mixture of packaged ΔC and ΔME-eGFP replicons at a dilution of 10−2 is shown in Fig. Fig.3B.3B. In the merged images, each of these foci could be seen to consist of a mixed population of cells, some of which were green (ΔME-eGFP), some red (ΔC), and some yellow (ΔC+ΔME-eGFP), similar to what was observed earlier in confluent monolayers shown in Fig. Fig.2.2. This means that each of the foci was formed by infection of neighboring cells by particles containing ΔC or ΔME-eGFP RNA or by both types of particles simultaneously. Importantly, examination of more than 100 individual foci in this manner did not reveal any that consisted of only red-stained E-protein-producing cells, which would have been expected if a wild-type revertant virus had arisen in the population due to recombination between the replicons.
In contrast to the defective replicon-containing particles, a fully infectious virion carrying a full-length genome would not depend on multiple infection of the same cell for its propagation, and if it were present in the population, it could be enriched by subsequent cell culture passages at high dilution. To test this principle experimentally, we carried out an experiment similar to the ones described above, except this time we spiked the mixture of packaged replicons with 10 focus-forming units of wild-type TBE virus to see whether the full-length genome would eventually become dominant after serial passage at limiting dilutions. For this experiment, however, instead of ΔME-eGFP, we used a replicon called ΔME (10), which was identical to ΔME-eGFP except that it had a normal viral 3′ end and lacked the artificial cistron encoding eGFP (Fig. (Fig.1).1). Lysates of infected BHK-21 cells were analyzed after the initial infection and after three end point dilution passages by Northern blotting using a probe recognizing the nonstructural region of the genome. As shown in Fig. Fig.4A,4A, the replicons were initially detected after infection, but after the three passages, only a single band was observed. This RNA was confirmed by a set of RT-PCRs (Fig. (Fig.4B)4B) and sequencing to be the genome of the wild-type virus (not shown). RT-PCR 1 was specific for the wild-type sequence, whereas RT-PCRs 2 and 3 yielded specific bands for replicons ΔC and ΔME, respectively, as well as a more slowly migrating band for the wild-type RNA. After three passages, only the products specific for the wild-type sequence were obtained by all three RT-PCR assays (Fig. (Fig.4B).4B). This experiment thus showed that the full-length viral genome had displaced the replicons during passage in cell culture and demonstrates that this system can be used as a “recombination trap” for selecting potential wild-type revertants within the population of replicating RNAs.
The replicon constructs used in the previous set of experiments for establishing the reciprocal packaging system, although identical in the region encoding the nonstructural genes, contained only a small stretch of 27 identical nucleotides between the C and prM genes where a single homologous recombination event could result in the creation of a full-length genome by restoring the missing capsid gene in ΔC (Fig. (Fig.1).1). Therefore, in order to increase the likelihood of such an event, we constructed a new replicon to use in place of ΔME that increased the region of identical sequence between the deleted regions. This construct, called ΔE, had all of the E gene deleted (nucleotides 973 to 2460) but retained the entire prM gene as a 572-nucleotide region of common sequence identity where homologous recombination with ΔC could potentially take place (Fig. (Fig.1).1). To confirm the ability of this new construct to replicate, a quantitative PCR analysis was performed. As shown in Fig. Fig.4C,4C, all TBEV constructs, including replicon ΔE, replicated to levels similar to that of the wild-type control replicon ΔR88, whereas the input RNA of a nonreplicating mutant (ΔNS-5) was quickly degraded after transfection. In addition to the TBEV constructs, we also made analogous ΔC and ΔE replicons using genomic clones of the mosquito-borne flaviviruses WNV and JEV (Fig. (Fig.11).
The recombination trap was then put to use by cotransfecting BHK-21 cells with the TBEV, WNV, and JEV ΔC and ΔE replicon pairs and carrying out multiple end point dilution passages to favor the selection and enrichment of any viruses with full-length genomes that might arise by recombination. Full-length infectious clones of each virus were used as controls. Typically, end point dilutions of complementing replicons ranged from 10−3 to 10−4, while wild-type virus controls could be diluted 10−6 to 10−7. Interestingly, a significant increase in the titer to between 10−5 and 10−6 was observed at passage 3 or passage 5 in two independent experiments with complementing replicons ΔC and ΔE of JEV but not with the TBEV and WNV constructs.
After 10 end point passages, intracellular RNA was analyzed by gel electrophoresis and Northern blotting with virus-specific probes (Fig. (Fig.5A).5A). As expected, each of the full-length, wild-type controls yielded a single major band. In the case of TBEV, transfection with ΔC together with ΔME or ΔE resulted in two major bands (Fig. (Fig.5A,5A, left panel, lanes 3 and 4) that corresponded in size to the original replicons (lanes 5, 6, and 7), suggesting that the two original replicons had been copassaged and maintained throughout the entire 10 passages. Likewise, cotransfection of ΔC and ΔE from WNV also yielded both replicon bands (Fig. (Fig.5A,5A, middle panel, lane 3).
In contrast to these results, however, in the case of JEV, Northern blotting revealed only a single band after passage of the ΔC+ΔE pair, suggesting that the replicons were not propagated as a pair and that a recombination event might have occurred. This phenomenon was observed in two completely separate experiments. Northern blot analysis of earlier passages indicated that replacement of the original pair of replicons with the new band had already occurred in the third and fifth passages of experiments 1 and 2, respectively (data not shown). The relative positions of the bands in the gel suggested that the final RNA recombination products were not identical in size in the two experiments (Fig. (Fig.5A,5A, right panel, lanes 3 and 4).
As a more sensitive and precise means of assessing whether recombinant RNA molecules were present in the population, we devised RT-PCR assays for detecting the restoration of deleted regions of the TBEV, WNV, and JEV genomes (see Materials and Methods). In these assays, neither of the original replicons could be amplified due to the lack of one of the primer-binding sites, but recombinants containing both of these sites would yield a PCR product. As shown in Fig. Fig.5B,5B, in the case of TBEV and WNV (left and middle panels, respectively), the RT-PCR assay yielded a product in the case of the wild-type control but not with the passaged replicon pairs, indicating that no detectable reversion to wild-type TBEV or WNV had occurred via interreplicon recombination. With JEV, however, a clearly different situation was observed (Fig. (Fig.5B,5B, right panel). Positive RT-PCR results were obtained not only with the wild-type control (lane 1) but also with the combination of ΔC and ΔE in the two separate experiments (lanes 3 and 4). The first of these PCR products (lane 3) migrated more slowly than the one obtained using the wild-type genome as template, whereas the second (lane 4) was approximately the same size. These results provided positive evidence that recombination events had indeed occurred with the JEV system. Other RT-PCR assays that were designed to detect the original replicons (see Materials and Methods) yielded the expected products for the original TBEV and WNV replicons but failed to yield detectable amplification products from the passaged JEV samples, indicating that both replicons had been completely displaced by the new recombinant in both of these experiments (data not shown).
These newly selected recombinant genomes were then reverse transcribed and sequenced. Each of them was found to contain all of the components of the JEV genome, but neither corresponded in its arrangement to a wild-type genome. Recombinant 1 (Fig. (Fig.6A,6A, top), which was obtained in experiment 1, was composed of the replicon ΔE sequence from its 5′ end up to nucleotide 2761 and still included the original deletion of nucleotides 978 to 2477 in the E gene. The remainder of the genome was derived from replicon ΔC and included the entire region extending from the beginning of prM to the 3′ terminus, thereby resulting in an aberrant genome containing a partial duplication of the NS1 region and a complete duplication of the prM region between prM and E. Recombinant 2 (from experiment 2) resembled the wild-type JEV except that it contained a tandem duplication of nucleotides 393 to 455, corresponding to the NS2B/3 protease cleavage site between C and prM (Fig. (Fig.6A,6A, bottom).
To characterize their biological properties, infectious clones of the recombinant 1 and recombinant 2 genomes were made as described in Materials and Methods. Full-length RNA was synthesized from these templates and used to transfect BHK-21 cells, which in turn released infectious particles into the cell supernatant. These viruses were then characterized by testing them in a standard plaque assay using Vero cells and by carrying out a multistep growth curve experiment using both BHK-21 cells and the mosquito cell line C6/36.
The plaque morphology produced by these viruses in Vero cells is shown in Fig. Fig.6B.6B. It was observed that the recombinant 1 virus produced large plaques that looked similar to those produced by wild-type JEV. The recombinant 2 virus, on the other hand, produced small plaques, suggesting that this virus was impaired in its ability to spread to other cells.
The growth curves shown in Fig. Fig.6C,6C, determined using an MOI of 0.01, show that both recombinants had impaired growth properties compared to wild-type JEV. Recombinant 2 showed strongly reduced growth kinetics in BHK-21 cells, whereas the growth rate of recombinant 1 was only moderately reduced compared to that of wild-type JEV. However, the titer of recombinant 1 fell sharply after 2 days due to a strong cytopathic effect. In C6/36 cells, both of the recombinants grew about an order of magnitude more slowly than the wild type, and recombinant 1 did not produce an unusually strong cytopathic effect in these cells.
These results suggest that although aberrant recombination events did occur in the JEV system, a normal wild-type revertant was not produced by a simple homologous crossover in any of the experiments with JEV, WNV, or TBEV.
Using a “recombination trap,” an experimental system designed to detect even rare recombination events between two self-replicating RNA molecules, we analyzed the propensity of three flaviviruses, TBEV, WNV, and JEV, for intermolecular recombination in the region encoding the structural proteins. The inclusion of long overlapping regions between the two trans-complementing replicons in our recombination trap provided ample opportunity for homologous crossover events to generate recombinants with wild-type genomes. Surprisingly, however, no such event was detected in any of the three flaviviral systems, even after prolonged copassaging, indicating that these viruses have a low propensity for exact homologous recombination in the region encoding the structural proteins. However, we did detect two aberrant recombination events in the JEV system, each of which gave rise to infectious virus progeny with an unnatural genome organization. Although passage at limiting dilutions apparently gave the two aberrant full-length recombinant forms of JEV a growth advantage over the original replicon pair, their growth properties would probably not have allowed them to prevail over a wild-type virus had one been present in the population.
The potential for unwanted recombination has been a topic of considerable controversy in the context of live flavivirus vaccine and vector development. Until now, the only evidence for flavivirus recombination has been inferred from sequence data from certain natural isolates (19, 20, 52-54). However, there has been a lack of general agreement about the significance of these observations, and due to the lack of appropriate experimental systems, it has not yet been verified under laboratory conditions that flavivirus genomes actually recombine. The recombination trap described here provides for the first time a sensitive system to address this issue experimentally, and the results obtained using this system represent the first direct observation of recombination between replicating flavivirus-derived RNA molecules in the laboratory. From this study, we can draw three major conclusions that are relevant for flavivirus vaccine and vector development.
The sensitivity of the recombination trap was demonstrated by the spiking experiment, which showed that as little as 10 infectious units of wild-type virus resulted in the complete displacement of the two replicons by the infectious wild-type virus within three passages. It is reasonable to assume that a single recombination event that gives rise to an infectious genome would produce at least 10 infectious particles and would thus outgrow the parental replicons within three passages or fewer. Flavivirus replicons typically multiply to 104 copies per cell (28, 48), and because the experimental conditions used allowed multiple rounds of infection at each passage, we estimate that at least 104 cells per passage would have been coinfected with complementary replicons, providing approximately 108 molecules that could potentially participate in a recombination event. Thus, a homologous crossover event resulting in the generation of wild-type virus—which could have happened at any position within the extended stretch of overlapping sequence identity between the two replicons—would have been selected at a frequency as low as 1 in 108 per passage. Although the possibility cannot be strictly excluded that the frequency of recombination is higher elsewhere in the genome, our data support the view that flaviviruses have a very low rate of recombination. Notably, the generation of wild-type virus was never observed for any of the three flaviviruses although they were tested over 10 passages.
Despite the fact that extensive overlaps of identical sequence were present in our replicon pairs to favor crossover events at homologous positions to regenerate wild-type genomes, the only recombination events observed were aberrant ones that generated unnatural, rearranged forms of the JEV genome. As with homologous recombination, longer sequence overlaps between the recombination partners provide more potential nonidentical crossover sites and would thus be expected to increase not only the likelihood of aberrant recombination within the overlap region but also the statistical chances of such an event producing a viable genome. Our results also raise the question of whether “wobbling” the codons of the overlap region, as has been done by some researchers to avoid homologous recombination, would actually reduce the recombination frequency if it is indeed confirmed that most crossovers are of the aberrant type. Furthermore, it remains unclear why no aberrant recombination was observed in the TBEV and WNV systems in spite of equivalent experimental conditions and sequence design.
It is possible that this kind of recombination is favored by as-yet-undefined RNA structures that were present in our JEV sequence but not in those of the other two flaviviruses. These questions can be addressed in future experiments with the recombination trap.
Aberrant recombination, which occurs by a crossover event at nonidentical genome positions, generates an unnatural genome organization with sequence duplications. In the case of the JEV recombinants generated and analyzed in this study, this unnatural genome structure apparently caused growth defects, as demonstrated by a reduced-plaque-size phenotype and/or growth kinetics in different host cells. This suggests that under natural conditions, progeny virus derived from such aberrant recombination events would, in most cases, have little, if any, chance to compete against the parental wild-type virus.
Overall, these conclusions suggest that recombination may turn out not to be a major concern with regard to live flavivirus vaccines and self-replicating flavivirus vectors. Our observations are in good agreement with the fact that, in spite of decades of use of live flavivirus vaccines and the cocirculation of several flaviviruses in several areas of the world in which infection is endemic, there has not been a single report of a naturally occurring recombinant involving a vaccine strain or of recombination between members of different flavivirus species. Furthermore, although numerous experiments with chimeric flaviviruses have clearly demonstrated that artificially constructed interspecies recombinants can be viable (13, 17, 22, 38, 40), such recombinants have never been observed to arise under natural conditions.
The low rate of recombination, even under conditions strongly favoring the selection of recombinants, suggests that recombination probably plays, at most, a minor role in the biology and evolution of flaviviruses, and this highlights a striking difference between flaviviruses and many other RNA viruses (31). Coronaviruses, for example, frequently undergo precise homologous recombination (33), and vaccine strains of poliovirus (a picornavirus) have been observed to recombine with each other and with other enteroviruses (6, 37). Recombination is also a major driver of genetic diversity and viral evolution of retroviruses such as human immunodeficiency virus (21, 51), and homologous recombination has also been reported among plant RNA viruses. For other viruses, such as alphaviruses and pestiviruses, aberrant homologous recombination and even nonhomologous recombination appear to be more common than homologous recombination, but the frequency of homologous recombination is still significant (3, 4, 14, 35, 36, 44). In fact, first-generation packaging systems for alphavirus vectors suffered from the risk of frequently observed recombination events between helper and vector replicons, regenerating infectious virus (55)—a problem that has been addressed by the development of tripartite packaging systems and other genetic modifications to prevent successful recombination. In the case of the pestiviruses, which belong to the same viral family as the flaviviruses, sequences of cellular origin can be acquired by nonhomologous recombination, driving alterations in viral pathogenicity (4, 5). Homologous recombination in these viruses was reduced in a pestivirus study when the sequence identity between recombination partners was reduced (9).
How may the low propensity for recombination between flaviviruses be explained at the molecular level? RNA recombination is thought to occur through a process in which the RNA replication complex falls off its template and continues RNA synthesis on a different template. A similar process is probably involved in the generation of deletion or duplication mutations during RNA replication (7, 32). Here, the replication complex also falls off and reinitiates, but it does so on the same template molecule upstream or downstream from its previous position, thus generating, respectively, duplications and deletions. Spontaneous deletion and duplication mutations, in contrast to intermolecular recombination events, are frequently observed in flaviviruses (29, 46). If, however, the flavivirus replication complex, similar to that of other RNA viruses, is capable of releasing its template and continuing RNA synthesis at another location, why would it not be able to freely switch templates in the course of this process and thus generate a recombined RNA product? One possible explanation would be that replication takes place in individual, secluded compartments within the infected cell that predominantly contain RNA derived from a single template molecule (30, 56). An alternative explanation would be that the RNA template somehow remains associated with the replication complex at a site other than the active site, which would favor reinitiation on the same RNA molecule after a temporary release of the template from the active site of the synthetase.
The use of trans-complementation systems such as the ones described here as recombination traps will allow more detailed investigation of the types of recombination events that flavivirus genomes are capable of undergoing as well as their relative frequencies. In addition, the low propensity for recombination of flavivirus replicons with deletions in the structural region makes it possible for them to be used as reliable tools to study the complementation of individual genes derived from different viral strains or species.
This project was funded by the Austrian Fonds zur Förderung der wissenschaftlichen Forschung (FWF project no. P20533-B03).
We are grateful to Steven L. Allison for his invaluable assistance during data evaluation and preparation of the manuscript. We thank Regina Kofler for her participation in mutant construction and Franz X. Heinz for providing materials and reagents. Finally, we are grateful to the staff of the microbial ecology department of the University of Vienna for their help with the use of the LSM 510 scanning confocal microscope.
Published ahead of print on 28 October 2009.