Nonreplicative RNA Recombination in Poliovirus

Abstract

Current models of recombination between viral RNAs are based on replicative template-switch mechanisms. The existence of nonreplicative RNA recombination in poliovirus is demonstrated in the present study by the rescue of viable viruses after cotransfections with different pairs of genomic RNA fragments with suppressed translatable and replicating capacities. Approximately 100 distinct recombinant genomes have been identified. The majority of crossovers occurred between nonhomologous segments of the partners and might have resulted from transesterification reactions, not necessarily involving an enzymatic activity. Some of the crossover loci are clustered. The origin of some of these “hot spots” could be explained by invoking structures similar to known ribozymes. A significant proportion of recombinant RNAs contained the entire 5′ partner, if its 3′ end was oxidized or phosphorylated prior to being mixed with the 3′ partner. All of these observations are consistent with a mechanism that involves intermediary formation of the 2′,3′-cyclic phosphate and 5′-hydroxyl termini. It is proposed that nonreplicative RNA recombination may contribute to evolutionarily significant RNA rearrangements.

Recombination between viral RNA genomes, first discovered in poliovirus (16, 23), is now known to be widespread among animal, plant, and bacterial viruses (reviewed in references 2, 7, 18, 22, and 39). It is generally believed that recombination and other covalent rearrangements in viral RNA genomes, such as deletions and insertions, occur during RNA replication as a result of template switching (17, 19, 26, 36). In the framework of this view, the elongation of a nascent RNA strand may slow down and prematurely terminate, for example, due to stable secondary structure elements (43, 47) or nucleotide misincorporations (32). Then, the dissociated 3′ terminus anneals to another template or to another site of the same template, wherein the strand elongation resumes to produce a recombinant molecule. Recent studies of the conditions for the template-switch recombination between viral RNA genomes in cell-free systems (11, 27–29, 41) should greatly facilitate the elucidation of its mechanism(s).

The first indication of the existence of a nonreplicative transesterification mechanism for RNA recombination was recently obtained in the in vitro Qβ phage system which employed Qβ phage replicase to detect replicable RNA species generated from nonreplicable RNA fragments (7, 8). The goal of the present study was to assess whether viable recombinant viruses could be generated from nonreplicating and nontranslatable parts of a viral RNA genome. To this end, several pairs of the poliovirus RNA fragments have been designed. In each pair, one of the putative recombination partners lacked a segment encoding the polyprotein and the 3′-untranslated region (3UTR), whereas the other partner possessed lethal modifications in essential translational (and in one case also in replicative) cis elements of the 5′-untranslated region (5UTR). Numerous infectious clones with a variety of crossover points have been recovered after transfections of susceptible cells with mixtures of the noninfectious partners. The results suggest that a nonreplicative mechanism (as opposed to the replicative template-switch mode) might be involved in the generation of the recombinant RNAs in our system.

MATERIALS AND METHODS

Construction of the 5′ partners.

Plasmid pT7PV1 (34) carrying the full-length poliovirus genome was linearized by EcoRI and, after blunting of the termini with the Klenow fragment, was fused to BglII linker(s) (Pharmacia). A segment containing the entire vector sequence, together with the T7 RNA polymerase promoter and 66 5′-terminal poliovirus bases, was excised from this construct by treatment with KpnI and BglII. This segment was fused to one of the three modified versions of the remaining part of the viral 5UTR. To obtain these versions, the pBM1 vector (34) was modified by insertion of two marker mutations (C₄₅₁→AG and G₅₅₂→AC). Appropriate fragments of mutated pBM1 were prepared by PCR by using sense primer (positions 59 to 79 of the poliovirus RNA) and one of the BglII site-containing antisense primers (GGAGCAGATCTAGCAAACAG [BG], TGAGATCTCACTTTCACCGG [BN], or TAGATCTCTAATGTCTCACTTTCAC [BY]), followed by treatment of the products with KpnI and BglII.

Construction of the 3′ partners.

Plasmid pBM1 with an additional BalI site (39) was digested with BalI followed by religation. A plasmid with an inverted BalI segment (nucleotides [nt] 635 to 727) was selected, and this inversion was introduced into the full-length viral genome. The inversion resulted in a decrease in the reproductive potential of the virus due to the appearance of two AUGs in the IRES/AUG₇₄₅ spacer (at positions 682′ and 653′); a third AUG at position 729 was a result of the introduction of the BalI site (see text above). A large-plaque virus revertant was selected. It contained mutations A_682′→G and G_651′→U (changing two of the AUGs), as well as A_693′→G. cDNA corresponding to the inverted portion of the revertant genome was obtained by reverse transcriptase PCR (RT-PCR), followed by the BalI treatment, and it was used to replace the corresponding segment in the pBM1 modified as described above. TTCCTTTT₅₆₇ was replaced by an oligo(dA) block of identical length to give the PA2 construct. PA2 was treated with BamHI, and fragment with coordinates 226 to 670′ was deleted to generate construct ΔBB. The modified 5UTRs were introduced into the full-length viral genomes. The ΔL construct was prepared by ligation of the following three fragments: (i) a fragment containing the T7 promoter and the poliovirus segment with coordinates 101 to 987, generated by PCR with pT7PV1/PA2 (see above) with the sense primer TAATACGACTCACACTATAGGTAACTTAGACGC (T7 promoter and poliovirus sequence 101 to 113) and the antisense primer corresponding to positions 1058 to 1078, followed by digestion with ScaI; (ii) a fragment corresponding to poliovirus positions 988 to 5601, prepared by digestion of pT7PV1 by ScaI and BglII; and (iii) a fragment harboring the 3′-terminal portion of the poliovirus genome and vector sequences, prepared by digestion of pT7PV1 by BamHI, isolation of the appropriate segment, treatment with the Klenow enzyme, and digestion with BglII. The primary structures of all of the modified portions of the 5′ and 3′ partners were checked by sequencing.

Preparation of transcripts and transfection.

For the generation of RNA transcripts, the plasmids were linearized with BglII (5′ partners) or EcoRI (3′ partners), and the transcription by T7 RNA polymerase was carried out as described earlier (34). The transcripts were purified by sucrose gradient centrifugation (5 to 20%, 4 h, and 40,000 rpm [Beckman SW41 rotor]). Ethanol-precipitated RNA was dissolved in water, and its concentrations were determined spectrophotometrically. DEAE-dextran-mediated transfection of AGMK cells was carried out as described previously (34) by using a mixture of the 5′ and 3′ fragments (1 μg of each; molar ratio of ∼11:1). In some experiments, the fragments were premixed in 40 μl of 100 mM NaCl–10 mM EDTA–0.1% sodium dodecyl sulfate (SDS)–80 mM Tris-HCl (pH 7.5) and heated at 60°C for 10 min, followed by a slow cooling to 37°C and incubation at this temperature for 24 h. The RNA fragments were then twice precipitated with ethanol, redissolved in the same buffer without SDS, and used for transfections as previously described. In some experiments, transfection was carried out by using Lipofectin (Gibco) as recommended by the manufacturer. Partners preannealing and different transfection protocols did not result in any appreciable differences in either frequencies or localization of the crossovers.

Modification of the 5′-transcript termini.

For the periodate oxidation, 22 μl of 40 mM NaIO₄ was added to 2 μg of RNA in 18 μl of a buffer (222 mM sodium acetate, pH 5.3; 22 mM EDTA). The mixture was incubated in the dark at room temperature for 30 min. The RNA was ethanol precipitated, washed with 70% ethanol, dried, dissolved in 50 μl of water, and again precipitated with ethanol. The treatment with aniline was performed as follows. First, 100 μl of 0.5 M aniline containing 10 mM sodium acetate (pH 5.0) was added to the RNA preparation dissolved in 100 μl of water. Then, after incubation at room temperature for 120 min in the dark, the RNA was twice precipitated with ethanol. The 3′-phosphorylated 5′ partners were prepared as follows. First, 10 μl of a mix containing 4 μg of BN transcript, 0.5 mM (final concentration) pCp, 10 U of RNasin (Pharmacia), 1 mM ATP, and 2.5 U of T4 RNA ligase (Pharmacia) was incubated at 37°C for 60 min. The phosphorylated RNA was then thrice extracted with a phenol-chloroform mixture and twice precipitated with ethanol. The alkaline phosphatase treatment was carried out as described earlier (8), but the incubation was carried out at 37°C in the presence of 1 U of RNasin per μl.

Analysis of recombinant genomes.

The material from a plaque was suspended in 1 ml of Earle’s saline and subjected to RT-PCR (44) by using oligonucleotide primers corresponding to positions 1058 to 1078 (antisense) and 59 to 79 (sense) of the poliovirus RNA. After purification by agarose electrophoresis, the DNA product was sequenced by using Sequenase v.2.0 (U.S. Biochemicals) according to the manufacturer’s protocol but in the presence of 10% dimethyl sulfoxide (33). If the PCR product exhibited heterogeneity, it was treated with KpnI and ScaI, and the 68-to-987 fragment was cloned into pBSM13(−) pretreated with KpnI and Ecl136II. At least three clones of each fragment were sequenced. The viral RNA was isolated and sequenced as described previously (44) by using avian myeloblastosis virus RT and primers corresponding to positions 744 to 759 and 877 to 894.

RESULTS

Design of the recombination partners.

Since the nonreplicative recombination should not necessarily be homologous, the putative partners were designed in such a way that their joining could generate viable genomes even if the recombination was nonhomologous. The crossovers were targeted to a nonessential region of the poliovirus RNA, namely, to the spacer separating the internal ribosome entry site (IRES) from the initiator AUG₇₄₅. The spacer is known to tolerate diverse and profound modifications without significant phenotypic changes (15, 20, 34, 39). The 5′ partners donating the leftward part of the putative recombinants were designed to lack the viral protein-coding capacity, whereas the 3′ partners, serving as donors of the rightward part, harbored lethal lesions in the essential cis-acting elements of the 5UTR.

The recombination partners used in this study are schematically depicted in Fig. 1. All the three 5′ partners were composed of a truncated portion of the 5UTR retaining all of its essential replicative and translational cis elements. They had a trimmed spacer (somewhat varying in length) that separates the IRES and the initiator AUG₇₄₅. The constructs ended with different short nonviral oligonucleotides. Finally, the leftward partners contained two marker mutations at positions 451 and 552, which did not affect the viral phenotype (not shown).

FIG. 1.

Schematic representation of the recombination partners. Solid lines correspond to segments of the poliovirus genome; the black bar denotes the inverted segment of the viral RNA (its coordinates are shown as n′). Cryptic (position 586) and initiator (position 743) AUG triplets are marked by open and solid stars, respectively. The intact and mutated (i.e., containing an oligoadenylate replacement) oligopyrimidine moieties of the essential OAT element are shown as solid and open diamonds, respectively. The silent marker mutations introduced into the 5′ partners at positions 451 and 552 are indicated by black dots. The borders of the essential replicative (oriL) and translational (IRES) elements are given on the scheme of the 5′ partners. The coordinates of the 3′-terminal nucleotides and the sequences of nonviral oligonucleotides fused to the 3′ ends of the 5′ partners, as well as the segments deleted in different 3′ partners, are given in the tables. For other details, see the text.

On the other hand, the polyprotein reading frame and the 3′-terminal replicative element oriR (35), as well as the poly(A) stretch, were preserved in the 3′ partners, but one or more of the essential cis-acting elements of their 5UTR were either deleted or mutated. The ΔBB construct lacked the entire IRES, as well as the adjoining oligopyrimidine-AUG tandem (OAT), an element essential for the cap-independent internal initiation of poliovirus RNA translation (30, 31, 34). In the PA2 RNA, the oligopyrimidine moiety of OAT was inactivated by converting UUCCUUUU₅₆₇ into an (A)₈ stretch, a change resulting in a lethal phenotype and an apparently complete suppression of the in vitro translational template activity of the poliovirus RNA (not shown). The same OAT-inactivating mutation was introduced also into ΔL RNA, but this construct also lacked 100 5′-terminal nucleotides corresponding to the essential replicative “clover-leaf” element oriL (3). Furthermore, portions of the IRES/AUG₇₄₅ spacer (positions 635 to 669 in ΔBB or positions 635 to 727 in PA2 and ΔL) in the 3′ partners were inverted to facilitate generation of heteroduplexes with the 5′ partners.

None of the six RNA constructs shown in Fig. 1 could by themselves generate any detectable infectious progeny when introduced into primary African green monkey kidney (AGMK) cell cultures by using the DEAE-dextran or Lipofectin transfection techniques. When the poliovirus 5UTR with the OAT modified as in constructs PA2 and ΔL was fused to the luciferase gene and the HeLa cells were transfected with the resulting construct, no luminescence above the background level was generated (not shown).

Generation of recombinants and characterization of crossovers.

When the transcripts corresponding to the 5′ and 3′ partners were mixed in pairs and transfected into AGMK cells, plaques were reproducibly developed at days 3 to 6, that is, 1 to 2 days slower than upon transfections with the wild-type transcripts. The yield of recovered viruses varied in different experiments but generally comprised several clones per microgram of the 3′ partner (which was present in a 1:11 molar ratio to the 5′ partner).

The relevant regions of the 5UTR (∼250 nt upstream of the initiator AUG₇₄₅) of the viral RNA isolated from the plaques were sequenced. The overwhelming majority of the sequenced genomes corresponded to distinct recombinants (Fig. 2). The recombinant nature of the RNAs was unambiguously demonstrated by the simultaneous presence of the sequences uniquely donated by a 3′ partner and those originating from a 5′ partner. All the recombinants but one retained the two silent mutations that tag positions 452 and 552 of the 5′ partners. The only exception concerned recombinant o8 retaining the latter but not the former tags. Most likely, this was a result of more than one recombination event.

FIG. 2.

The location of crossover sites on the 5′ partners (A) and 3′ partners (B). The composite crossover maps showing the results obtained with different combinations of the partners are presented, but the crossovers downstream of position 648 are given separately for each of the three 5′ partners. The crossover sites are denoted either by vertical bars (when the mapping was possible with a single-nucleotide accuracy) or triangles (when there was a short oligonucleotide identity in the recombining regions of the two partners). The nonviral (oligo)nucleotides found between the bodies of the 5′ and 3′ partners are shown to the right of the vertical bars corresponding to the terminal nucleotides in panel A. The position of the crossover of recombinant o22 on the 3′ partner (nt 203 to 204) lies outside the genomic region shown. The UGAAA sequence, a component of the putative cryptic hammerhead ribozyme, is shadowed. The regions of identity in the 5′ and 3′ partners (positions 568 to 634) are in boldface. The nucleotides in the inverted regions of the 3′ partners (positions 635 to 669 in ΔBB and positions 635 to 727 in the PA2 and ΔL partners) are denoted as n′. The regions of the 3′ partners supposed to form heteroduplexes with the 5′ partners are overlined (the longer overline corresponds to BG, whereas the shorter one corresponds to BN and BY partners). The recombinants obtained with oxidized (as well as oxidized and aniline-treated) or ligated with pCp 5′ partners are marked by the prefixes o and p, respectively. The crosses between individual partners yielded the following recombinants: BG × PA2, 1, 3, 5, 8, 15, 21, 25, 31, 32, 35, 36, o3, o4, and o17 to o19; BG × ΔBB, 26, 30, 33, 34, 37, and 38; BG × ΔL, o16; BN × PA2, 2, 4, 6, 10, 11, 17, 22, 23, 27, 28, o1, o2, o5, o6, o8, o10, o13, o24, and o27; BN × ΔBB, 7, 14, 18, 19, 29, and o22; BN × ΔL, 9, 12, 13, o7, o9, o11, o12, o14, o15, o23, o25, o26, o28 to o33, and p1 to p19; BY × PA2, 24 and o20; BY × ΔBB, 20, 39, and 40; BY × ΔL, 16, and o21.

The crossover points were distributed over ∼110- and ∼220-nt segments in the 5′ and 3′ partners, respectively, all but a few mapping to the IRES/AUG₇₄₅ spacer. There appeared to be several clusters (“hot spots”) of crossovers in both partners (Fig. 2). When several recombinants had identical crossover sites on the 5′ partner, their crossover sites on the 3′ partner often differed and vice versa. In most cases, the location of crossovers could be determined with a high precision, since the length of the identical oligonucleotide stretches in the crossover regions of two recombining partners varied from 0 to 7 nt, being predominantly equal to 0 to 2 nt. In a small subset of recombinants (altogether, 14 of 105 sequenced), the crossovers occurred within a 67-nt region shared by the both partners (positions 568 to 634). The precise mapping of crossovers in these cases was impossible, and such recombinants are not shown in Fig. 2.

Effects of modification of the 3′ end of 5′ partners.

Chetverin et al. (8) reported that periodate oxidation of the 3′ terminus of the 5′ partner dramatically suppressed recombination in the Qβ system. Surprisingly, a similar treatment of the poliovirus RNA-derived 5′ partner was accompanied by a severalfold increase rather than decrease in the number of plaques generated and also in some shortening of the time of their appearance (Table 1). Furthermore, the recombinant genomes thus obtained exhibited a remarkable peculiarity. The entire 5′ partner was fully incorporated into a significant proportion (more than one-third) of the recombinant genomes (Fig. 2; Table 2). Most of these recombinants contained also a short oligonucleotide insertion between the 3′ end of the 5′ partner and an internal nucleotide of the 3′ partner. Aniline treatment (expected to remove the modified nucleoside and generate a 3′-phosphate terminus), did not markedly affect the recombinogenic properties of the oxidized 5′ partners, whereas subsequent dephosphorylation by alkaline phosphatase resulted in an apparent return of the recombination frequency to the values typical of the nonoxidized fragment (Table 1). The generation of recombinants containing the entire 5′ partner was also severely suppressed by the latter treatment (Table 2). These data suggested that the “activation” was probably due to generation of a 3′-phosphorylated form of the 5′ partner (after elimination of the oxidized nucleoside). In line with this suggestion, essentially similar results (frequent incorporation of the full-length 5′ partner, short “foreign” oligonucleotide insertions, and an apparent increase in the recombinogenic potential) were observed when the 3′-phosphate was introduced into the 5′ partner by ligating it with pCp (Tables 1 and 2; Fig. 2).

TABLE 1.

Effects of modifications of the 5′ partner’s 3′ end on recombination^a

Expt no.	Treatment of the 5′ partner	No. of plaques/μg of RNA at days after transfection
		3	4
1	None	0	4.3 ± 4.7
	NaIO4	11.2 ± 4.1	Confluent
	NaIO4 + aniline	9.2 ± 5.8	Confluent
	NaIO4 + aniline + phosphatase	0.3 ± 0.5	0.5 ± 1.0
2	None	0	2.0 ± 1.7
	pCp ligation	8.5 ± 2.6	10.5 ± 3.1

^aExperiments with the BN/ΔL pair of partners are represented. The modifications of the 5′ partner’s 3′ end were performed as described in Materials and Methods. The average numbers of plaques ± the standard deviations are given. 

TABLE 2.

Effects of modifications of the 5′ partner’s 3′ end on incorporation of the full-length 5′ partners into recombinant genomes^a

Treatment of the 5′ partner	No. of recombinants
	Total	With truncated 5′ partner	With full-length 5′ partner
None	40	40	0
NaIO4	25	15	10
NaIO4 + aniline	18	11	7
NaIO4 + aniline + phosphatase	4	4	0
pCp ligation	18	10	8

^aThe modifications of the 5′ partner’s 3′ end were performed as described in Materials and Methods. The table includes the results obtained with all of the sequenced recombinant genomes. 

Genetic stability of the recombinants.

The recombinant nature of the rescued genomes was unambiguously substantiated in the preceding sections. Most of the crossover sites were identified by sequencing of the viral RNA species isolated from individual plaques. Eighteen recombinant viruses isolated from the plaques were subjected to two additional passages in vitro. Sequencing of the relevant regions of the viral RNAs isolated from 15 viruses demonstrated their complete identity to the plaque-derived material. The genomes of three viruses showed some heterogeneity. They were subjected to two further passages and plaque cloning. The heterogeneity was traced to the variability of the length of the oligoadenylate tract introduced into the PA2 and ΔL 3′ partners. The number of A residues varied in the range of 9 to 15, but no other changes, compared to the material from the primary plaques, could be detected. Thus, the recombinants proved to be genetically stable, at least under the conditions tested.

DISCUSSION

Nonreplicative recombination versus template switch.

Numerous viable recombinants between noninfectious fragments of the poliovirus RNA were identified here. The recombination partners were designed to be unable to direct the synthesis of any viral proteins (RNA polymerase included) unless they are fused to each other. One may argue that the 3′ partners with the intact open reading frame could have been translated, though inefficiently, thus producing the viral RNA polymerase. Hence, the generation of infectious progeny could have been a result of replicative recombination between the newly synthesized minus strand and the 5′ partner. Were it so, the PA2 RNA, being full length, should be quasi-infectious. Indeed, it could have been readily converted into a viable form by a single extended deletion similar to those that occurred upon the pseudoreversion of pPV1/Δ8 (15, 34). The PA2 transcript, however, proved to be completely dead in numerous appropriate assays, suggesting that this RNA was practically devoid of translation template activity. We consider this a strong argument for the notion that the recombinants have been generated by a mechanism other than the replicative template switch. There are also several other lines of indirect evidence against the template switch as the main cause of the recombination in our system. (i) The recombination “hot spots” did not preferentially map to regions of identity in the partners. (ii) The inverted segment of the 3′ partners presumed to form heteroduplexes with the 5′ partners contained very few crossover sites, whereas the heteroduplex-mediated template-switch mechanism predicted that such regions should be enriched in crossovers (25). It may be added that the crossovers within region 669′ to 649′ of the 3′ partners (e.g., 20, o13, p3, etc.) were only observed with the shorter (BN or BY) 5′ partner, which could not hybridize to that region. (iii) The entire 5′ partner was preserved in a significant proportion of the recombinants. Having no complementarity to the 3′ ends of the putative nascent negative strands (given the unlikely possibility that such strands could have been transcribed from the 3′ partner prior to the recombination event), the 5′ partner could hardly serve as an accepting template. (iv) The preservation (and even a significant enhancement) of the recombinogenic potential of the 5′ partners upon the oxidation or phosphorylation of their 3′ termini argues strongly against the possible use of these termini as primers for the synthesis of the positive strands (again under the unlikely assumption that the negative strands could somehow have already been generated). Moreover, this fact is difficult to reconcile with any template-switch mechanism.

The existing data do not allow us to discriminate between the extracellular and intracellular origin of the nonreplicative recombinants. Nor can it be ruled out that a certain proportion of recombinants (e.g., those with identical or similar sequences in the crossover regions of the both partners) could have arisen through secondary intermolecular or intramolecular template-switching involving primary nonreplicative recombinants. This could have occurred when the primary recombination had generated viruses with a low level of fitness. Generation of unfit viruses seemed unlikely for the recombinants with the primary crossover sites within the IRES/AUG₇₄₅ spacer because this spacer is known to be nonessential and highly promiscuous with regard to the primary structure (15, 20, 34, 39). However, primary recombination events could have also involved more upstream sites. In such cases, the recombination might result in viruses with a reproductive potential lowered for a variety of reasons (e.g., an AUG between the IRES and the initiator codon, excessive secondary structure, or genome length, etc.). As a consequence, the restoration of viral fitness could have resulted from a secondary template-switch recombination accompanied by the deletion of a deleterious genomic segment.

Thus, we conclude that at least the majority of the recombinants described here were generated by reactions other than template switching.

Possible nonreplicative mechanisms of RNA recombination.

A nonreplicative mode of recombination implies that recombining RNAs are cleaved at some points and the exposed termini are cross-ligated. There are two known enzyme-catalyzed chemical mechanisms, differing by the nature of the cleavage products, that can result in such RNA rearrangements. According to one mechanism, phosphodiester bonds are attacked by an external nucleophile (e.g., a water molecule) which exposes the 3′-hydroxyl and 5′-phosphate termini. Such termini can then be cross-ligated in a way that requires an activation (e.g., adenylylation) of the 5′-phosphate group (1). Alternatively, a direct transesterification reaction can occur in which the 3′-hydroxyl attacks a phosphodiester bond within an uncleaved partner molecule (6). The second mechanism of cleavage of putative partners of nonreplicative recombination includes an attack of a phosphodiester bond by the adjacent 2′-hydroxyl which plays the role of an internal nucleophile; this results in the 2′,3′-cyclic phosphate and 5′-hydroxyl termini. Such termini can be cross-ligated in a transesterification reaction which is chemically equivalent to the reversal of the RNA cleavage (24, 46).

Chetverin et al. (8) reported that, in the presence of Qβ replicase and no other enzyme, the 5′ partner was entirely incorporated into a recombinant molecule if it possessed hydroxyls at the 3′ end. This finding suggested that RNA recombination occurred by a direct attack of the 3′-hydroxyl at internucleotide bonds within the 3′ partner. The presence of Qβ replicase was essential for this reaction to occur (9). In the experiments reported here, the transcriptionally generated 5′ partners possessed free 3′-hydroxyls, but the entire 5′ partner was never observed to be incorporated into the recombinants unless its 3′ terminus was modified. Thus, the first mechanism apparently was not operating in our system.

The second mechanism, involving the 2′,3′-cyclic phosphate and 5′-hydroxyl termini, might operate provided that appropriate ribonucleases and ligases were available in the transfected cells. It is also possible that the second mechanism could operate without any enzyme. Of course, recombination requires that the generated 2′,3′-cyclic phosphate and 5′-hydroxyl termini belonging to different partners should be ligated and, hence, that they should be brought into a close proximity to each other in one of several possible RNA foldings.

Possible contribution of cryptic ribozyme activities.

Sequencing of the recombinant genomes has revealed several clusters of crossover sites (Fig. 2), suggesting that these sites are more reactive than others. The elevated reactivity might be due to a higher rate of RNA cleavage, a higher rate of ligation of the exposed termini, or both. Interestingly, some of the corresponding RNA segments can be folded into secondary structures similar to known ribozyme motifs.

One of these clusters is located on the 5′ partners ca. 12 nt downstream of a UGAAA sequence (Fig. 2). It is possible to draw structures corresponding to the consensus hammerhead ribozyme (Fig. 3A; reference 5) by base-pairing of a segment surrounding this sequence to another region of the same or an identical molecule (Fig. 3B). These structures, if formed, might well exhibit enzymatic properties characteristic of hammerhead ribozymes, generating 2′,3′-cyclic phosphate termini corresponding to some of the observed crossover sites on the 5′ partner. Recombination requires, of course, one more cleavage generating a counterpart 5′-hydroxyl terminus at the other partner. Although the existing data do not allow us to draw ribozyme-like structures for the counterparts, it is clear that an increased production of either of the termini to be ligated would result in the overall recombination enhancement.

FIG. 3.

A model invoking the involvement of a hammerhead-like ribozyme activity in the origin of a specific hot spot of crossover sites on the 5′ partners. (A) The conserved nucleotides (boxed) and the consensus structure of hammerhead ribozymes. R, Y, and H represent a purine, pyrimidine, or any nucleotide except G, respectively. The cleavage site is shown by the arrow. (B) A hypothetical folding of the 5′ partner, generating a consensus hammerhead ribozyme annealed to a segment of the 3′ partner.

In many of the above cases, more or less stable heteroduplexes could be drawn in which the putative 5′-hydroxyl of the 3′ partner and the 2′,3′-cyclic phosphate on the 5′ partner are juxtaposed by base-pairing the corresponding RNA segments to a complementary “guiding” sequence (not shown). Formation of such structures is a prerequisite for efficient RNA ligation by a number of protein or ribozyme catalysts (4, 10, 21, 37, 42).

It is obvious that active ribozyme structures able to cleave and ligate RNA molecules, if they do not serve specific purposes, as in small plant pathogenic RNAs (40) or hepatitis δ virus (42), should be selected against during viral evolution. However, if the activity could only be expressed in a negligible proportion of the viral genomes, it may persist unnoticed by the negative selection mechanisms. In this sense, the cryptic ribozyme activities proposed above appear to be biologically neutral.

Incorporation of the entire 5′ partner.

The mechanism of generation of recombinants containing the full 5′ partner in the present system appeared to be fundamentally different from that operating in the experiments of Chetverin et al. (8), since the oxidation (phosphorylation) of the 5′ partner resulted in the activation or suppression of its recombinogenic potential in these two systems, respectively. In the present study, the incorporation of the entire 5′ partner was usually accompanied by the appearance of nonviral oligonucleotides between the bodies of the 5′ and 3′ segments. These “foreign inserts” were most likely generated during the preparation of transcripts with T7 RNA polymerase. The enzyme is known to produce, during runoff transcription in vitro, variable extensions showing complementarity to regions near the 3′ end of the correct transcript (resulting from self-priming) (45). Nearly all of the insertions observed in our recombinants did exhibit such a complementarity (Fig. 4).

FIG. 4.

A model explaining incorporation of the full-length 5′ partner into the recombinant genomes. Foreign insertions, assumed to be added by T7 RNA polymerase, are shaded. Putative reacting nucleotides are boxed. The 3′-terminal nucleotide of the 5′ partner is assumed to possess the 2′,3′-cyclic phosphate group.

Since the consequences of phosphorylation and oxidation of the 5′ partners were the same (a significant increase in the recombinogenic potential and generation of recombinant RNAs containing the entire 5′ fragment), we assumed that the chemical natures of the molecules involved were also identical, being represented by a 3′-phosphorylated 5′ partner (for the oxidized molecule, this could be a result of elimination of the oxidized nucleoside due to its reaction with cellular amines).

It is difficult to imaging how the 3′ phosphate group could by itself accelerate the joining of the 5′ partner to the 3′ partner. However, the cellular RNA 3′-terminal phosphate cyclase (14) could convert this group into the 2′,3′-cyclic phosphate, which could then react with the exposed 5′-hydroxyl of a cleaved 3′ partner, provided that the reacting groups happen to be next to each other. Importantly, in many cases these groups could have been brought into proximity to one another in putative heteroduplex intermediates involving the 3′ and 5′ partners (Fig. 4). It may be added that 3′-phosphorylated and 5′-hydroxyl-terminated viral RNA fragments can certainly be generated within the infected cell as a result of nucleolytic degradation. Therefore, the mechanism described above might well be operating upon natural viral infections.

Biological significance.

The existence of nonreplicative mechanisms of RNA recombination should not be interpreted to mean the negation of the template switch mode. Moreover, it is likely that the overwhelming majority of cases of homologous RNA recombination have been due to the replicative mechanisms, as suggested by the character and distribution of crossover sites and other evidence (see references 2, 12, 13, 19, 26, 29, and 47).

However, under certain conditions, nonreplicative joining of RNA fragments may occur. Such events, though very rare, may be of extreme evolutionary importance because they may result in the generation of novel genomes from otherwise incompatible parents. Moreover, such nonreplicative rearrangements may occur in cellular RNA as well and could be fixed by reverse transcription.