Demidenko and Victor I. UgarovHelena V. Chetverina, Alexander A. Alexander B. Chetverin, Damir S. Kopein,  Recombination between RNA MoleculesDiverse Mechanisms to Promote Viral RNA-directed RNA Polymerases UseCatalysis:RNA: Structure, Metabolism, and

doi: 10.1074/jbc.M412684200 originally published online December 17, 20042005, 280:8748-8755.J. Biol. Chem. 

  10.1074/jbc.M412684200Access the most updated version of this article at doi:

  .JBC Affinity SitesFind articles, minireviews, Reflections and Classics on similar topics on the

 Alerts:

  When a correction for this article is posted• 

When this article is cited• 

to choose from all of JBC's e-mail alertsClick here

  http://www.jbc.org/content/280/10/8748.full.html#ref-list-1

This article cites 51 references, 22 of which can be accessed free at

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from

at R

oyal Melbourne Inst of T

echnology (CA

UL

) on May 30, 2014

http://ww

w.jbc.org/

Dow

nloaded from

Viral RNA-directed RNA Polymerases Use Diverse Mechanisms toPromote Recombination between RNA Molecules*

Received for publication, November 9, 2004, and in revised form, December 17, 2004Published, JBC Papers in Press, December 17, 2004, DOI 10.1074/jbc.M412684200

Alexander B. Chetverin‡, Damir S. Kopein, Helena V. Chetverina, Alexander A. Demidenko§,and Victor I. Ugarov

From the Institute of Protein Research of the Russian Academy of Sciences, Pushchino, Moscow Region, 142290 Russia

An earlier developed purified cell-free system wasused to explore the potential of two RNA-directed RNApolymerases (RdRps), Q� phage replicase and the polio-virus 3Dpol protein, to promote RNA recombinationthrough a primer extension mechanism. The substratesof recombination were fragments of complementarystrands of a Q� phage-derived RNA, such that if alignedat complementary 3�-termini and extended using oneanother as a template, they would produce replicablemolecules detectable as RNA colonies grown in a Q�replicase-containing agarose. The results show thatwhile 3Dpol efficiently extends the aligned fragments toproduce the expected homologous recombinant se-quences, only nonhomologous recombinants are gener-ated by Q� replicase at a much lower yield and througha mechanism not involving the extension of RNA prim-ers. It follows that the mechanisms of RNA recombina-tion by poliovirus and Q� RdRps are quite different. Thedata favor an RNA transesterification reaction cata-lyzed by a conformation acquired by Q� replicase dur-ing RNA synthesis and provide a likely explanation forthe very low frequency of homologous recombination inQ� phage.

Recombinations (sequence exchange and rearrangements) be-tween and within RNA molecules are rare but biologically im-portant events contributing to the evolution and diversity of RNAviruses (1, 2) and generating defective interfering RNAs thatattenuate viral infections (3). In contrast to splicing and othertypes of regular RNA rearrangements, recombinations occurwithout apparent sequence or structure specificity (1, 2). Thereare indications that recombination may occur between cellularRNAs (4–6), eventually resulting, by means of reverse transcrip-tion and integration, in alterations in the chromosomal DNA.Spontaneous Mg2�-catalyzed rearrangements in RNA sequences(7) might have been a mechanism for evolution in the prebioticRNA world and might have evolved into contemporary sequence-specific ribozyme-catalyzed reactions (8, 9).

RNA recombination was discovered more than 40 years agoas an exchange of genetic markers between polioviruses (10,11), and since then similar approaches were used to demon-strate that genomes of RNA viruses of animals, plants, andbacteria are all capable of recombination (2, 4, 12, 13). How-ever, such in vivo experiments utilizing living cells, as well asin vitro studies that used crude cell lysates could not uncoverthe underlying molecular mechanisms or even definitely an-swer the question if recombining entities were RNA moleculesthemselves or their cDNA copies, because every living cellcontained enzymes capable of reverse transcription and appro-priate dNTP substrates. It became evident that further pro-gress in this field depended on the availability of adequate invitro systems whose composition and other parameters can bestrictly controlled by the experimenter (2, 14).

The first example of such a sort has been the cell-free systememploying purified Q� replicase, RNA-directed RNA polymer-ase (RdRp)1 of bacteriophage Q� (15). The system also includestwo RNA molecules, “5� fragment” and “3� fragment,” whosesequences supplement each other to the entire sequence ofRQ135 RNA, an efficient Q� replicase template (16), and arederived from the 5� and 3� segments of that RNA, respectively.None of the fragments alone can be amplified by Q� replicase;however, fusion of their sequences in a manner as they arearranged in the original RQ RNA results in the appearance ofreplicable molecules (15), which are detected and counted byusing the Q� replicase version of the molecular colony tech-nique (17, 18). To this end, a mixture of the fragments is seededon a Q� replicase-containing agarose layer, which is then cov-ered with a nylon membrane impregnated with replicase sub-strate ribonucleoside triphosphates (rNTPs) to initiate replica-tion. As the reaction takes place in agarose, copies of replicableRNAs concentrate around the progenitor templates in the formof RNA colonies. The colonies are detected by hybridizing themembrane with a labeled probe, and their number reflects therecombination frequency. Experiments in this system provedthat recombination can occur between RNA molecules directly,without involving DNA intermediates (15). However, manyfeatures of the cell-free RNA recombination appeared to bedifferent from those usually observed in the in vivo studies.

Studies on recombination in RNA viruses mainly revealedhomologous recombination (1, 2), in which sequences surround-ing the crossover site in the recombination substrates (recom-bining RNAs or segments of an RNA molecule) and in theproduct molecule are entirely or essentially identical to each

* This work was supported by the program “Molecular and CellBiology” of the Russian Academy of Sciences, Russian Foundation forBasic Research Grant 02-04-48320, International Association for thepromotion of co-operation with scientists from the New IndependentStates of the former Soviet Union Grant 01-2012, a grant from theMinistry of Industry, Science and Technology of the Russian Federa-tion, and an International Research Scholar’s award from the HowardHughes Medical Institute (to A. B. C.). The costs of publication of thisarticle were defrayed in part by the payment of page charges. Thisarticle must therefore be hereby marked “advertisement” in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

‡ To whom correspondence should be addressed: Institute of ProteinResearch, Pushchino, Moscow Region, Russia 142290. Tel.: 7-0967-73-2524; Fax: 7-095-924-0493; E-mail: alexch@vega.protres.ru.

§ Present address: University of Chicago, Chicago, IL 60637.

1 The abbreviations used are: RdRp, RNA-directed RNA polymerase;rNTPs, ribonucleoside triphosphates; PAAE mechanism, primer align-ment and extension mechanism; 3Dpol, poliovirus RdRp; RQ RNA,replicable by Q� replicase, a non-genomic RNA capable of exponentialamplification by Q� replicase; 3�C fragment, the complementary copy ofthe 3� fragment; nt, nucleotide(s).

THE JOURNAL OF BIOLOGICAL CHEMISTRY Vol. 280, No. 10, Issue of March 11, pp. 8748–8755, 2005© 2005 by The American Society for Biochemistry and Molecular Biology, Inc. Printed in U.S.A.

This paper is available on line at http://www.jbc.org8748

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from

other. To explain homologous recombination, a template switch(copy choice) mechanism was proposed (19), according to whichthe recombinant molecule is generated as a by-product of RNAsynthesis by viral RdRp, which after copying a portion of thefirst (“donor”) template, occasionally switches to another (“ac-ceptor”) template containing a sequence complementary to thegrowing end of the nascent strand. If that mechanism operatedin the cell-free Q� system, the 3� and 5� fragments would serveas donor and acceptor templates, respectively (Fig. 1A). How-ever, no homologous recombinants were generated in this sys-tem, even though the fragments were provided with homolo-gous foreign sequences to facilitate template switching. Mostrecombinant molecules contained the entire sequence of the 5�fragment fused with sequence of the 3� fragment, either intactor 5�-truncated to a various extent (Fig. 1B; Ref. 15). Mostimportantly, recombination was totally prevented when the 3�hydroxyl at the 5� fragment terminus was either removed byperiodate oxidation or blocked by a phosphate group. The keyrole of the 3� hydroxyl of the acceptor template could not beaccounted for by the template switch mechanism, but wellconformed to a hypothesis that RNA recombination occurs as atransesterification reaction in which the 3� hydroxyl of the 5�fragment attacks phosphate groups within the 3� fragment. Inthe absence of any indication of the ability of RNA polymerasesto promote such reactions, it was suggested that recombinantsarose because of a site-nonspecific self-splicing activity of RNAmolecules (15).

Such an RNA activity was indeed detected with the use of amodified experimental scheme, in which, before applying to theQ� replicase-containing agarose, a reaction mixture was oxi-dized with periodate to suppress further recombination byeliminating any free 3� hydroxyls at the recombining RNAmolecules (7). However, the spontaneous recombinations be-tween RNA molecules appeared to be several orders of magni-tude less frequent than in the presence of Q� replicase. More-over, they did not require free 3� hydroxyl groups, indicatingthat a quite different reaction chemistry was employed, mostprobably, a Mg2�-catalyzed RNA cleavage generating frag-ments with 2�,3�-cyclic phosphate and 5� hydroxyl termini,which are then cross-ligated. These observations suggestedthat the 3� hydroxyl-dependent RNA recombinations are some-how promoted by Q� replicase (7, 20).

The results obtained in the cell-free Q� system stimulatedattempts to detect similar phenomena in vivo, by transfectingsusceptible cells with two supplementing fragments of the ge-nome of an RNA virus, or with inactivated genome and acomplementing fragment. Experiments with derivatives of po-liovirus (5, 21) and bovine viral diarrhea virus (6) RNAs dem-onstrated that viable viruses could be rescued even if none ofthe recombination substrates could be translated or if transla-tion could not result in the active viral RdRp. Similarly to thespontaneous rearrangements observed in the Q� system, thisnonreplicative recombination was not affected by elimination ofthe 3� hydroxyls. Moreover, its frequency increased when the 5�and 3� fragments bore 3� phosphoryl and 5� hydroxyl groups,respectively, suggesting an involvement of the 2�,3�-cyclic phos-phate intermediate (5, 21). Crossing poliovirus RNA fragmentsoverlapping within a nonessential segment of the 5�-untrans-lated region resulted in the rescue of only nonhomologous re-combinants (5). At the same time, only homologous and aber-rant homologous recombinants that retained the translationreading frame were rescued when RNA fragments overlappedwithin the sequence coding for protein 3Dpol (poliovirus RdRp)(21), indicating that natural selection can considerably distortthe results of in vivo experiments.

Quite different results were obtained in another type of in

vivo experiments, in which an expressible cDNA clone of polio-virus capable of producing active RdRp, but incapable of repli-cation because of two point mutations in the 5�-untranslatedregion, was rescued by recombination with a mutation-free5�-untranslated region fragment (22). In these experiments, noviruses were rescued if the 5� fragment had been modified at

FIG. 1. Substrates of RNA recombination. The sequences origi-nated from RQ135 RNA (16) are shown only partially by white letters onblack background; homologous (or complementary) stretches of foreignsequences are shown on gray background. A, homologous recombinantRNA that would be generated from the 5�(BamHI) and 3� fragments bythe template switch mechanism. Such a recombinant was produced byreverse transcriptase from avian myeloblastosis virus (15). B, the non-homologous recombinant RNA most frequently produced from the samefragments by Q� replicase in the purified cell-free system (15). C,homologous recombinant RNA that would be generated from the samefragments by the PAAE mechanism (22) provided that the 3� fragmentis copied along its entire length and the copy (3�C fragment) dissociatesfrom the template. D, a nonhomologous recombinant RNA that mightbe generated by the PAAE mechanism when the 3�C fragment remainsbase-paired to the 3� fragment and priming occurs on occasional meltingof the duplex terminus. E, the pairs of opposite polarity fragments usedin this work to probe the PAAE mechanism.

Mechanisms of Replicative RNA Recombination 8749

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from

the 3� terminus with cordycepin (3�-deoxyadenosine). Thus, aswith Q� replicase, RNA recombination in the presence of po-liovirus RdRp appeared to be entirely dependent on the avail-ability of the free 3� hydroxyl of the 5� fragment. However, incontrast to the Q� replicase-promoted reaction (15) and RdRp-independent poliovirus RNA recombination in vivo (5), onlyhomologous recombinants were produced, even though codingsequences were not involved. To explain their observations, theauthors proposed a “primer alignment and extension” (PAAE)mechanism, in which no templates are switched. Instead, apre-existing fragment of poliovirus RNA (in nature, this can bea fragment generated by abortive synthesis or by an enzymaticdegradation exposing the 3�-OH termini) is extended by RdRpusing the strand of opposite polarity as a template, to which ithybridizes by a complementary 3� terminal segment.

The striking similarity of the responses of Q� and poliovirussystems to the elimination of the 3� hydroxyl of the 5� fragment,together with the fact that Q� replicase is involved in the 3�hydroxyl-dependent recombination, raised the possibility thatsimilar mechanisms operate in both systems. However, theabsence of homologous recombination in the Q� system re-mained unexplained. Whether the latter fact reflects a funda-mental mechanistic difference between Q� replicase and polio-virus RdRp, or the mere limitations of the respective in vitroand in vivo systems and/or sequence or structure dissimilari-ties of the recombination substrates remains to be answered.

To resolve these alternatives, we compared effects of the twoRdRps under similar conditions and with the same RNA sub-strates, using the cell-free Q� system to monitor RNA recom-bination. The results of these studies demonstrate that al-though ongoing RNA synthesis is required for the generation ofrecombinant RNAs by both Q� replicase and 3Dpol, the mech-anisms employed by the two enzymes are entirely different. Inparticular, whereas the PAAE mechanism is efficiently used by3Dpol to generate recombinant molecules from RNA fragmentsof opposite polarities, it is totally rejected by Q� replicase.

EXPERIMENTAL PROCEDURES

Enzymes—Q� replicase was isolated from Escherichia coli HB101cells transformed with plasmid pRep (23) as described (24, 25). Thepoliovirus RdRp (3Dpol) gene was PCR-amplified using plasmidpKKT7E-3D (26) as a template and primers that introduced an up-stream NcoI site and a downstream His6-coding sequence followed bythe BamHI site, cloned in pET15b between sites NcoI and BamHI andexpressed in E. coli BL21(DE3) cells (27). A highly purified 3Dpolprotein containing the hexahistidine tag peptide at the C terminus wasisolated by chromatography on Zn2�-iminodiacetate-Sepharose CL-4B(28) from a cell lysate prepared in a buffer of 50 mM Tris-HCl, pH 8.0,10% glycerol, 100 mM NaCl, 0.1% Nonidet P-40 (29), dialyzed againstthe same buffer containing 2 mM dithiothreitol, and after addition ofglycerol to 50%, was stored at �20 °C. The resulting preparation wasactive in the poly(A)-directed synthesis of poly(U) (30) and contained nodetectable activities of E. coli RNA polymerase, polynucleotide phos-phorylase, or ribonucleases.

RNA Fragments—The 5� and 3� fragments were synthesized by run-off transcription with T7 polymerase using corresponding plasmidsdigested with appropriate restriction endonucleases, and gel-purified asdescribed (15). Plasmid for the synthesis of the 3�C fragment wasprepared by a PCR templated with a pUC18-derived plasmid, in whicha T7 promoter/RQ135�1(�) cDNA construct was inserted between sitesHindIII and SmaI (31), using primers 5�-CTGCAGGCATGCAAGCTTA-ATACGACT-3�, partially overlapping the sequence of the T7 promoter(bold) and containing the HindIII site (underlined), and 3�-AGTT-TAGGGAGCATCTAGGAGATCTCAGCTGGACGTCCTTAAG-5�, par-tially overlapping the sequence of the 3� fragment (bold) and containingand an identical sequence to the foreign sequence of the 5�(BamHI)fragment (italic), including the PstI site (underlined). The PCR productwas digested at the HindIII site, blunt-ended by filling in the recessed3� terminus with Klenow enzyme, digested at the PstI site, and ligatedinto plasmid pUC18 that had been digested at PstI and SmaI anddephosphorylated. The primary structure of the resulting plasmid waschecked by sequencing.

RNA Recombination—To separate the recombination and replicationsteps (as in experiments of Figs. 2 and 5A), the earlier devised proce-dure (7) was employed. Unless indicated otherwise, the recombiningfragments (3 � 1011 molecules each) were annealed in a 2 times incu-bation buffer (see figure legends), not including Mg2� and rNTPs, byincubating during 2 min in a boiling bath followed by cooling to �30 °Cduring 1 h. After incubation under specified conditions followed by theaddition of EDTA to chelate all Mg2�, the reaction mixture was ex-tracted with phenol/chloroform (32), oxidized with sodium periodate(33), desalted by passing through a Sephadex G-25 spun column, andmelted (7).

Detection and Sequencing of Recombinant RNAs—In the experi-ments of Fig. 2, RNA colonies were grown in agarose slabs (18 � 18 �0.37 mm) as reported (15, 18). A 10-�l sample containing the specifiedamount of RNA was distributed over Q� replicase-containing agaroseand covered with a nylon membrane (Hybond N, Amersham Bio-sciences) containing rNTPs. In the experiments of Figs. 3 and 5A, RNAcolonies were grown in round (14 mm diameter, 0.4 mm thick) poly-acrylamide gels that were earlier used for growing DNA colonies (34).Pre-cast and dried gels were reconstituted by soaking in 70 �l of asolution containing the RNA sample and all the replication reactioncomponents but rNTPs that were introduced with a nylon membranecovering the gel. As compared with agarose gels, the use of polyacryl-amide gels resulted in a 5–10 times higher recovery of recombinantRNAs. In either case, the final concentrations of reaction componentswere: 80 mM Tris-HCl, pH 7.8, 8 mM MgCl2, 1 mM EDTA, 20% glycerol,1 mg/ml acetylated bovine serum albumin (Sigma), 35 �g/ml Q� repli-case, and 1 mM each of rNTPs. After incubation during 1 h at 22 °C, themembranes were fixed (18) and hybridized with a 32P-labeled5�(BamHI) fragment to reveal the colonies by autoradiography (15).Then RNAs were extracted from gels, cloned, and sequenced (15).

RESULTS

RNA Synthesis Is Required for the Generation of RecombinantRNAs by Q� Replicase—To investigate requirements of the 3�hydroxyl-dependent recombination, we used the experimentalapproach earlier established for exploring the ability of RNAmolecules to self-recombine (7). The same polarity 5� and 3�fragments of RQ135 RNA were incubated under chosen condi-tions with or without the addition of the components of thecell-free replication system (Q� replicase, rNTPs, and Mg2�), andbefore assaying for the presence of RNAs capable of replicating ina Q� replicase-containing gel, the incubation mixture wastreated with sodium periodate and melted. This separated therecombination and replication events, and RNA colonies onlygrew if recombination had occurred before the oxidation step.Control samples, in which a pre-oxidized 5� fragment substitutedfor the normal one, were run in parallel to ascertain if the ob-served recombination was 3� hydroxyl-dependent.

We observed no recombination between the RNA fragmentsabove the level of spontaneous reaction (7) unless the incuba-tion mixture contained all the reagents needed for RNA syn-thesis, including each of the four rNTPs (not shown). Fig. 2Ashows that recombination requires the same concentration ofthe initiating nucleotide GTP as does copying of the 3� frag-ment (monitored by the generation of a double-stranded prod-uct) that has inherited the initiation site of RQ135 RNA. Re-quirements of the 5� fragment copying are saturated at a lowerGTP concentration, as reported earlier (35).

Thus, recombination between the 5� and 3� RQ135 RNAfragments is observed under conditions that provide for theircopying by Q� replicase. Whereas recombination was alwaysaccompanied by copying of the 3� fragment (the donor templatein terms of the template switch mechanism), it was not firmlylinked with copying of the 5� fragment (the acceptor template).For example, elimination of the 3� hydroxyl group of the 5�fragment by periodate oxidation prevents recombination (Fig.2A and Ref. 15), but only slightly affects synthesis of the com-plementary copy (35). A reverse example is provided by the PstIand �EcoRI variants of the 5� fragment, which are hardlycopied by Q� replicase (Fig. 1 in Ref. 35), but are excellent

Mechanisms of Replicative RNA Recombination8750

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from

recombination substrates (not shown).Fig. 2B shows that chain terminator 3�-deoxy-ATP (cordyce-

pin 5�-triphosphate) inhibits recombination (as well as the syn-thesis of full-sized complementary copies of the recombiningfragments, not shown) if added to the reaction mixture at a 20:1ratio to ATP, at which it is expected to be used by Q� replicaseinstead of ATP at about a 20% probability (36). It follows thatRNA synthesis is required for recombination, rather thanmerely accompanying it.

Q� Replicase Does Not Utilize the Primer Alignment andExtension Mechanisms—The fact that 3� hydroxyl-dependentrecombination between the 5� and 3� fragments requires RNAsynthesis, in particular, copying of the 3� fragment raises apossibility that it is driven by the PAAE mechanism. In thiscase, the genuine substrates of recombination would be frag-ments of opposite polarities, i.e. the 5� fragment and the 3�Cfragment (the complementary copy of the 3� fragment).

To explore the PAAE mechanism directly, we replaced the 3�fragment by its complement synthesized by run-off transcrip-tion with T7 RNA polymerase, and checked if the 5�- and 3�Cfragments can serve as primers for their own extension by Q�replicase using each other as a template. Furthermore, bymanipulating foreign sequences, we prepared several pairs ofthe 5�- and 3�C fragments whose 3�-terminal sequences werecapable of complementary overlapping one another to variouslengths (Fig. 1E); the fragments were designated after therestriction endonucleases used to cleave the plasmid DNAs

before transcription. If the PAAE mechanism operates, oneshould expect that: 1) pairs with longer complementary over-laps will recombine at a higher frequency; 2) mostly homolo-gous recombinants will be generated at longer overlaps; and3) recombination between the fragments of opposite polarities(5� and 3�C) will be more efficient than between the samepolarity fragments (5� and 3�), because the 3�C fragment doesnot need to be synthesized by Q� replicase and is not base-paired with the 3� fragment.

It turned out that recombination between fragments5�(BamHI) and 3�C(PstI) overlapping by 18 nt (nucleotides)does occur (Fig. 3A) and is promoted by preliminary annealingof the fragments (Fig. 3B). Fragment pairs with shorter com-plementary overlaps, 5�(SalI) � 3�C(PstI) and 5�(PstI) � 3�C(P-stI) (cf. Fig. 1E), recombine at a lower frequency (Fig. 3A).Oxidation of both the 5�- and 3�C fragments is required tosuppress recombination, indicating that the 3� hydroxyl groupof each of them is important (Fig. 3C). These observations arein apparent agreement with the PAAE mechanism.

However, other findings do not support this mechanism. In-stead of the expected increase in recombination frequency, re-combination between the fragments of opposite polarities turnedout to be some 3 orders of magnitude less efficient than betweenthe same polarity fragments (Fig. 3B). Also, fragments5�(BamHI) and 3�C(SphI), capable of a longer complementaryoverlap (24 nt, cf. Fig. 1E) recombine at a 10 times lower fre-quency than do fragments 5�(BamHI) and 3�C(PstI) (Fig. 3, A andC). Finally, sequencing has shown that homologous recombi-nants are not generated (Fig. 4). Almost every recombinant mol-ecule contains sequences originated from the full size 5�- and 3�Cfragments, separated by an insert of variable length. The insertscontain nucleotide stretches complementary to the foreign se-quences of the fragments (shown in bold and underlined). Theirpossible origin is discussed below. Thus, the above data demon-strate that, although Q� replicase is capable of promoting recom-bination between the fragments of opposite polarities, it does itwithout using the PAAE mechanism.

FIG. 2. Requirements of the Q� replicase-promoted recombi-nation. The annealed 5�(BamHI) and 3� fragments were incubatedduring 1 h at 22 °C in the presence of 10 mM Tris-HCl, pH 7.8, 100 mM

NaCl, 10 mM MgCl2, 1 mM EDTA, 35 �g/ml Q� replicase, and unlessspecified otherwise, 1 mM each of rNTPs. Then the reaction mixture wasextracted with phenol/chloroform, oxidized with periodate, and de-salted. A, during incubation, the GTP concentration was varied asindicated. Top panel, the reaction mixture was assayed for the presenceof replicable RNAs. Each sample contained 109 molecules of each of the3� fragments and either the native (5�) or a periodate-oxidized (5�oxi) 5�fragment. Bottom panel, analysis of the reaction mixtures prior tophenol/chloroform extraction by electrophoresis in a polyacrylamide gelunder non-denaturing conditions (32) followed by silver staining (52).Lane M contained a mixture of the 5�(BamHI) and 3� fragments; ss,single-stranded; ds, double-stranded RNAs (cf. Ref. 35). Because of astrong binding to replicase (35), most of the single-stranded 5� fragmentmigrated at top of the gel (not shown). B, effects of 3�-deoxy-ATP(cordycepin triphosphate) on recombination between the 5�(BamHI)and 3� fragments (109 molecules each). The reaction mixture contained0.2 mM ATP and, at a specified incubation time, 3�-deoxy-ATP wasadded to the final concentration of 4 mM; the concentration of Mg2� wasadjusted to compensate for the increased concentration of nucleotides.Before phenol/chloroform extraction, 3�-deoxy-ATP was dephosphoryl-ated (together with other NTPs) by additionally incubating the reactionmixture during 20 min with 1 unit of calf intestine alkaline phospha-tase (molecular biology grade, Roche Molecular Biochemicals).

FIG. 3. Recombinations between RNA fragments of oppositepolarities in the presence of Q� replicase. After annealing, theindicated 5�- and 3�C fragments (108 molecules each, unless otherwiseindicated) were introduced into the RNA amplification gels withoutpreliminary incubation or any other treatment. A, correlation betweenthe recombination frequency and the length of a complementary over-lap between foreign sequences of the fragments, which was, from left toright, 24, 18, and 6 nt, and null; cf. Fig. 1E). B, effect of annealing andcomparison to recombination between the same polarity fragments. C,effect of periodate oxidation of the recombination substrates.

Mechanisms of Replicative RNA Recombination 8751

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from

Poliovirus RdRp Uses the Primer Alignment and ExtensionMechanism to Produce Homologous Recombinants from Q�-specific RNAs—Fig. 5A demonstrates generation of replicableRNAs from the 5�- and 3�C fragments by poliovirus RdRp. Tokeep the effects of 3Dpol undistorted by Q� replicase, we incu-bated the RNA fragments in the presence of rNTPs and 3Dpoland, before applying to the Q� replicase-containing agarose,extracted the reaction mixture with phenol/chloroform to re-move 3Dpol and oxidized with sodium periodate to eliminatethe free 3� hydroxyls at the RNA molecules. It is seen thatfragment pairs 5�(BamHI) � 3�C(PstI) and 5�(BamHI) �3�C(SphI), overlapping by 18 nt and 24 nt complementarysequences, respectively, recombine at the same frequency thatincreases upon annealing of the fragments and is 4–5 orders ofmagnitude higher than in the presence of Q� replicase (cf. Fig.3). The frequency falls considerably when the length of a com-plementary overlap decreases to 6 nt (pair 5�(SalI) � 3�C(PstI))and is not detectable with fragments 5�(PstI) and 3�C(PstI)whose termini do not show a potential for significant comple-mentary interactions.

Sequencing of the replicable RNAs obtained showed thatonly homologous recombinants were produced. Moreover, be-cause there was a G:U opposition at the 3� end of the 3�C(SphI)aligned with the 5�(BamHI) fragment (Fig. 1E), it was possibleto distinguish between recombinants produced by extendingthe 5�(BamHI) fragment and those produced by extending the3�C(SphI) fragment. It is seen that both the fragments wereextended, with the 5�(BamHI) fragment producing a canonical3� terminal opposition (C:G) being extended at a higher fre-quency (Fig. 5B). Thus, 3Dpol enzyme manifests a primer-de-pendent template-directed RNA polymerization activity withQ�-specific RNAs, in accord with observations made on polio-virus-derived sequences (37–39).

DISCUSSION

Mechanistic Differences between Q� Replicase and PoliovirusRNA Polymerase—This article presents results of the first com-

parative study of intermolecular RNA recombination promotedby RdRps of two different RNA viruses under similar physico-chemical conditions, using the same recombination substratesand the same amplification system. Therefore, any effects thatmight influence the results, such as effects of the primary or ahigher RNA structure, or of a selective amplification of somesort of recombinant molecules, are eliminated. The resultsclearly demonstrate that the fact that only nonhomologousrecombinants are generated in the cell-free Q� system,whereas mainly homologous recombination is observed in po-liovirus, reflects a fundamental difference in the mechanisms

FIG. 5. Recombinations between RNA fragments of oppositepolarities in the presence of poliovirus RdRp. A, fragments thathad been annealed (bottom panel) or not (top panel) were incubatedduring 30 min at 30 °C under conditions optimal for the primer-depend-ent 3Dpol activity (39): 50 mM Hepes-KOH, pH 7.0, 0.8 mM MgCl2, 5 mM

dithiotreitol, 0.1 mM each of rNTPs, and 30 �g/ml 3Dpol. The reactionmixtures additionally contained the following reagents introduced to-gether with the enzyme preparation: 5 mM Tris-HCl, pH 8.0, 10 mM

NaCl, 5% glycerol, and 0.01% Nonidet P-40. Before assaying replicableRNAs in aliquots containing the specified number of molecules of eachof the indicated fragments, the reaction mixtures were extracted withphenol/chloroform, oxidized with periodate, and desalted. B, primarystructures of RNAs generated by recombination between fragments5�(BamHI) and 3�C(SphI). A value in parentheses indicates the numberof clones sharing that sequence.

FIG. 4. Primary structures of re-combinants generated by Q� repli-case from RNA fragments of oppositepolarities. Square brackets indicateRNA termini that were deprived of 3� hy-droxyls by periodate oxidation. Values inparentheses indicate the number of cloneswhose sequences were identical. For otherexplanations, see “Results” and the leg-end to Fig. 1.

Mechanisms of Replicative RNA Recombination8752

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from

employed by Q� and poliovirus 3Dpol, rather than artifacts ofeither the in vivo or in vitro systems.

The puzzling observation that recombination between the 5�and 3� fragments of an RQ RNA was entirely suppressed byeliminating the free 3� hydroxyl of the 5� fragment constitutedthe main argument against the template switch mechanismand in favor of a transesterification mechanism in the cell-freeQ� system (15). This conclusion was later challenged byPierangeli et al. (22) who made a similar observation in apoliovirus system in vivo and argued that the 3� hydroxylrequirement might also indicate that the 5� fragment served asa primer for its own extension using the 3�C fragment as atemplate. Our finding that the Q� replicase-promoted recom-bination requires an ongoing RNA synthesis (Fig. 2) apparentlysupported their PAAE mechanism. However, it should be notedthat this mechanism had problems in explaining why the newlysynthesized 3�C fragment was not used as a primer with the 5�fragment as a template, as evidenced by the total suppressionof recombination in the cell-free Q� system upon elimination ofthe free 3� hydroxyl at the 5� fragment (15). One might arguethat, as far as the very 5� terminus of the 3� fragment was nothomologous to the 5� fragment, an end-to-end copying of the 3�fragment would result in a 3�C fragment whose 3� end wouldlack complementarity to the 5� fragment, making its extensionon the 5� fragment impossible (Fig. 1C). However, such anargument would only be valid if the Q� replicase-promotedrecombination was homologous, which is opposite to what isobserved in the experiment. In principle, generation of nonho-mologous recombinants by the PAAE mechanism could be con-ceived by taking into account the facts that the newly synthe-sized 3�C fragment remains base-paired to the 3� fragment (35),and that Q� replicase cannot unwind the RNA duplex (40, 41).Under these circumstances, priming might occasionally occurat sites of local complementarity when the duplex spontane-ously unwinds a short distance away from its terminus as aresult of thermal motion (Fig. 1D). Of course, such a processwould be very inefficient, but the observed recombination is notefficient either. It should be, however, noted that in this caseextension of the 5� fragment would have no obvious preferenceover extension of the 3�C fragment, and the 100% inhibition ofrecombination by the oxidation of the 5� fragment would re-main unexplained.

The results of this study show that, even when experimentalconditions are most favorable for the PAAE mechanism, Q�

replicase denies using it. The reluctance of Q� replicase toextend RNA primers is unexpected in view of an earlier reportthat it employs short oligoribonucleotides to by-pass the nor-mal GTP-dependent initiation on homopolymeric templates(42). In contrast, under similar conditions and with the sameRNA substrates this mechanism is readily used by the poliovi-rus RdRp, even though the extended primers and the templatesare heterologous to poliovirus. Although unexpected, these re-sults are in accord with the in vivo observations that homolo-gous recombination in the Q� phage is a million times lessfrequent than in poliovirus (20, 43, 44).

Possible Mechanism of RNA Recombination by Q� Repli-case—Two important features are seen in most sequences re-sulting from Q� replicase-promoted recombination between the5�- and 3�C fragments: 1) both the fragments donate theirentire sequences, and 2) an extra sequence is inserted betweenthem (Fig. 4). The insert contains a stretch of nucleotidescomplementary to the foreign sequences of the fragments, sug-gesting that it has been generated by partial copying of afragment. The fact that the primary structure of the insertdepends on which of the fragments was pre-oxidized with pe-riodate suggests that such a copying occurred after the synthe-

sis of fragments by T7 RNA polymerase, i.e. it was performedby Q� replicase.

Recently, we have shown that Q� replicase can copy deriva-tives of a 3�-truncated RQ RNA in a GTP-independent mode(35). There can be either de novo initiation of complementarycopies, with a template being copied along the entire lengthirrespective of its 3� terminal sequence, or a 3�-terminal elon-gation of the template, including a snapback RNA synthesisproducing a hairpin in which the template and the complemen-tary copy make up opposite sides of the stem. Formation ofsimilar hairpins as a result of the 3�-terminal elongation ofreplicable RNAs by Q� replicase was reported earlier (45).Taking into account that the 5�- and 3�C fragments are 3�-truncated derivatives of the (�) and (�) strands of the RQ135RNA, respectively (16), one can imagine the following scenariosleading to the generation of a replicable RNA, i.e. one in whichproper polarity and order of the fragments are observed. Gen-eration of recombinant RNA with the longest insert (the up-permost sequence of Fig. 4A) is considered as an example(Fig. 6A).

In Mechanism 1, recombinant RNA is generated by a trans-esterification reaction in which the 5� fragment attacks aninternucleotide phosphate in the 3�C fragment that has been 3�terminally extended producing a hairpin twice the size of thetemplate. Because this scenario requires that each of the frag-ments possesses the free 3� hydroxyl group, it should be re-jected because this particular recombinant was generated in anexperiment employing the periodate-oxidized 5� fragment,which lacks such a group.

In Mechanism 2, also of a transesterification type, the 3�terminal hydroxyl of the 3�C fragment, which has been par-tially extended beyond its 3� end in a snapback manner,attacks the �-phosphate at the 5� terminus of the full-lengthcomplementary copy of the 5� fragment. Finally, in Mecha-

FIG. 6. Possible mechanism of the Q� replicase-promoted RNArecombination. Thin arrows indicate directions of strand extension;thick arrows indicate directions of nucleophilic attacks of 3� hydroxylson phosphorus atoms. For other explanations, see “Discussion” and thelegend to Fig. 1. A, conceivable scenarios of the generation from thefragments of opposite polarities, 5�(BamHI) and 3�C(SphI), of the up-permost recombinant sequence as depicted in Fig. 4A. B, the main siteof the proposed attack of the 3� fragment by the 3� hydroxyl of the samepolarity 5� fragment (15).

Mechanisms of Replicative RNA Recombination 8753

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from

nism 3, the 3�C fragment, which has been 3� terminallyextended as in Mechanism 2, is further extended using the 5�fragment as a template. In other words, Mechanism 3 em-ploys extension of a primer not aligned with a template; asimilar mechanism was proposed to explain the synthesis ofRNA longer than the template by RdRps of bovine viraldiarrhea virus and some plant viruses (46). Both Mechanisms2 and 3 can operate with the oxidized 5� fragment, becauseoxidation does not prevent fragment copying (35). However,the transesterification reaction (Mechanism 2) seems to bepreferable for the following reasons.

In Mechanism 2, the 5� fragment is provided in a double-stranded form that protects internucleotide phosphates of itscomplementary copy from attack by the 3� terminal hydroxyl ofthe extended 3�C fragment, and this could explain why the 5�fragment almost always donates its entire sequence to theresulting recombinant (Fig. 4). In Mechanism 3, the 5� frag-ment is provided in a single-stranded form, and therefore theabsence of internal priming, which would result in recombi-nants with a 3� terminally truncated 5� fragment sequence,remains unexplained. This consideration is further strength-ened by the sequences of recombinants generated from thesame polarity fragments, 5� and 3�, which include the entiresequence of the 5� fragment and a variably truncated 3� frag-ment (15). In that case, the situation is reversed: the 3� frag-ment would be provided single-stranded in a transesterifica-tion mechanism (Fig. 6B) and double-stranded in a primerextension mechanism (Fig. 1D) and, again, the data conform tothe transesterification mechanism.

The same conclusion can be drawn from comparison of re-combination frequencies observed with fragments of the same(5� and 3�) and opposite (5�- and 3�C) polarities, on assumptionthat the same mechanism operates in both cases. For a primerextension mechanism, the immediate substrates are fragmentsof opposite polarities, whereas for a transesterification mecha-nism the immediate substrates are the same polarity frag-ments. Hence, for any type of a primer extension mechanism,the fragments of opposite polarities should recombine at ahigher frequency than the same polarity fragments, but inreality they recombine at about a 1000-fold lower frequency, inagreement with the transesterification mechanism.

If the transesterification mechanism operates, what is thenthe role of the ongoing RNA synthesis? The answer is notobvious for the recombination between the same polarity frag-ments, in which case both the recombination substrates areready for use from the very beginning. One possibility might bethat Q� replicase occasionally catalyzes transesterification re-actions between RNA molecules while being in a special con-formation that the enzyme only acquires when it synthesizesRNA (cf. Ref. 35). In this regard, it should be noted that theproposed transesterification reaction, comprising an attack ofthe 3� terminal hydroxyl of one RNA molecule on an internucle-otide phosphate of another, is chemically analogous to theattack of the leading 3� hydroxyl of a nascent strand on the�-phosphate of a nucleotide to be added next.

Diversity of Replicative Mechanisms for RNA Recombina-tion—Until recently, it was generally accepted that viral RNArecombination involves viral RdRp that eventually switchesbetween templates during RNA synthesis (1, 2, 13, 14, 19, 47),with “template switch” and “replicative mechanism” being usedas synonymous terms, as opposed to recently discovered “non-replicative” transesterification mechanisms (5–7, 21). In theclassical template switch (copy choice) model, the followingsteps can be distinguished (e.g. Ref. 2): 1) pausing of RdRp (e.g.at sites of secondary structure); 2) dissociation of RdRp carry-ing the nascent strand from the first (donor) template; 3) bind-

ing of the RdRp-nascent strand complex to the second (accep-tor) template; 4) elongation of the nascent strand on the secondtemplate. In a modified “processive” model (48), RdRp switchesto the second template without leaving the first template. ThePAAE mechanism (22) differs from the classical templateswitch in that the first two steps are omitted; yet, as in thetemplate switch model, the recombinant RNA is generatedthrough a template-directed elongation of an RNA primer bystepwise addition of monomer nucleotides to its 3� terminus.Our data obtained with poliovirus RdRp perfectly agree withthe PAAE model and further experiments are needed to ascer-tain whether 3Dpol can play the complete template switchscenario with natural heteropolymeric RNAs, as it was sug-gested from the results of in vitro studies employing ho-mopolyribonucleotides (49).

A quite different result has been obtained with Q� replicase.As for the template switch mechanism, ongoing RNA synthesisis also needed in this case and, for that reason, the recombina-tion should be regarded as a replicative one. However, thisrecombination does not seem to result from elongation of anRNA primer. The available data indicate that the recombinantmolecule might be generated by an RdRp-catalyzed transes-terification reaction, i.e. by adding to the 3� terminus of an RNA(a piece of) another RNA, rather than a mononucleotide. Be-sides Q� and related phages, such a mechanism might alsooperate in other viral systems manifesting a very low rate ofhomologous recombination, e.g. in alphaviruses (50, 51).

Whatever is the precise molecular mechanism used by eachof these enzymes, the very fact that the two viral RdRps stud-ied here behave so differently when confronted with the sameRNA substrates under similar conditions suggests that thereexist more than one type of replicative mechanism for RNArecombination.

Acknowledgments—We thank Dr. Steve Schultz (University of Colo-rado, Boulder) for plasmid pKKT7E-3D; Alexander Simonenko, DmitryLesnyak, Zakir Tnimov, Nadezhda Androsova, Tatiana Popkova, andLarissa Shutova for technical assistance; and Dr. Vadim Agol for crit-ical reading of the manuscript.

REFERENCES

1. King, A. M. Q. (1988) in RNA Genetics (Domingo, E., Holland, J. J., andAhlquist, P., eds) Vol. II, pp. 149–165, Chemical Rubber Company Press,Boca Raton, FL

2. Lai, M. M. (1992) Microbiol. Rev. 56, 61–793. Schlesinger, S. (1988) in RNA Genetics (Domingo, E., Holland, J. J, and

Ahlquist, P., eds) Vol. II, pp. 167–185, Chemical Rubber Company Press,Boca Raton, FL

4. Chetverin, A. B. (1997) Semin. Virol. 8, 121–1295. Gmyl, A. P., Belousov, E. V., Maslova, S. V., Khitrina, E. V., Chetverin, A. B.,

and Agol, V. I. (1999) J. Virol. 73, 8958–89656. Gallei, A., Pankraz, A., Thiel, H-J., and Becher, P. (2004) J. Virol. 78,

6271–62817. Chetverina, H. V., Demidenko, A. A., Ugarov, V. I., and Chetverin, A. B. (1999)

FEBS Lett. 450, 89–948. Spirin, A. S. (2002) FEBS Lett. 530, 4–89. Chetverin, A. B. (2004) FEBS Lett. 567, 35–41

10. Hirst, G. K. (1962) Cold Spring Harbor Symp. Quant. Biol. 27, 303–30911. Ledinko, N. (1963) Virology 20, 107–11912. Bujarski, J. (1996) Semin. Virol. 7, 361–36213. Nagy, P. D., and Simon, A. (1997) Virology 235, 1–914. Agol, V. I. (1997) Semin. Virol. 8, 77–8415. Chetverin, A. B., Chetverina, H. V., Demidenko, A. A., and Ugarov, V. I. (1997)

Cell 88, 503–51316. Munishkin, A. V., Voronin, L. A., Ugarov, V. I., Bondareva, L. A., Chetverina,

H. V., and Chetverin, A. B. (1991) J. Mol. Biol. 221, 463–47217. Chetverin, A. B., Chetverina, H. V., and Munishkin, A. V. (1991) J. Mol. Biol.

222, 3–918. Chetverina, H. V., and Chetverin, A. B. (1993) Nucleic Acids Res. 21,

2349–235319. Cooper, P. D., Steiner-Pryor, S., Scotti, P. D., and Delong, D. (1974) J. Gen.

Virol. 23, 41–4920. Chetverin, A. B. (1999) FEBS Lett. 460, 1–521. Gmyl, A. P., Korshenko, S. A., Belousov, E. V., Khitrina, E. V., and Agol, V. I.

(2003) RNA (N. Y.) 9, 1221–122322. Pierangeli, A., Bucci, M., Forzan, M., Pagnotti, P., Equestre, M., and Perez

Bercoff, R. (1999) J. Gen. Virol. 80, 1889–189723. Shaklee, P. N., Miglietta, J. J., Palmenberg, A. C., and Kaesberg, P. (1988)

Virology 163, 209–213

Mechanisms of Replicative RNA Recombination8754

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from

24. Blumenthal, T. (1979) Methods Enzymol. 60, 628–63825. Berestowskaya, N. H., Vasiliev, V. D., Volkov, A. A., and Chetverin, A. B.

(1988) FEBS Lett. 228, 263–26726. Hansen, J. L., Long, A. M., and Schultz, S. C. (1997) Structure 5, 1109–112227. Studier, F. W., Rosenberg, A. H., Dunn, J. J., and Dubendorf, J. W. (1990)

Methods Enzymol. 185, 60–8928. Lindner, P., Guth, B., Wulfing, C., Krebber, C., Steipe, B., Muller, F., and

Plukthun, A. (1992) Methods 4, 41–5629. Neufeld, K. L., Richards, O. C., and Ehrenfeld, E. (1991) J. Biol. Chem. 266,

24212–2421930. Flanegan, J. B., and van Dyke, T. A. (1979) J. Virol. 32, 155–16131. Morozov, I. Yu., Ugarov, V. I., Chetverin, A. B., and Spirin, A. S. (1993) Proc.

Natl. Acad. Sci. U. S. A. 90, 9325–932932. Sambrook, J., and Russell, D. W. (2001) Molecular Cloning, 3rd Ed., Cold

Spring Harbor Laboratory Press, Cold Spring Harbor, NY33. Steinschneider, A., and Fraenkel-Conrat, H. (1966) Biochemistry 5, 2729–273434. Chetverina, H. V., Samatov, T. R., Ugarov, V. I., and Chetverin, A. B. (2002)

BioTechniques 33, 150–15635. Ugarov, V. I., Demidenko, A. A., and Chetverin, A. B. (2003) J. Biol. Chem.

278, 44139–4414636. Kramer, F. R., and Mills, D. R. (1978) Proc. Natl. Acad. Sci. U. S. A. 75,

5334–533837. Richards, O. C., and Ehrenfeld, E. (1998) J. Biol. Chem. 273, 12832–1284038. Arnold, J. J., Ghosh, S. K. B., and Cameron, C. E. (1999) J. Biol. Chem. 274,

37060–3706939. Rodriguez-Wells, V., Plotch, S. J., and DeStefano, J. J. (2001) Nucleic Acids

Res. 29, 2715–272440. Weissmann, C., Feix, G., Slor, H., and Pollet, R. (1967) Proc. Natl. Acad. Sci.

U. S. A. 57, 1870–187741. Biebricher, C. K., Diekmann, S., and Luce, R. (1982) J. Mol. Biol. 154,

629–64842. Feix, G., and Hake, H. (1975) Biochem. Biophys. Res. Commun. 65, 503–50943. Palasingam, K., and Shaklee, P. N. (1992) J. Virol. 66, 2435–244244. Olsthoorn, R. C., and van Duin, J. (1996) J. Virol. 70, 729–73645. Biebricher, C. K., and Luce, R. (1992) EMBO J. 11, 5129–513546. Kim, M. J., and Kao, C. (2001) Proc. Natl. Acad. Sci. U. S. A. 98, 4972–497747. Kirkegaard, K., and Baltimore, D. (1986) Cell 47, 433–44348. Jarvis, T. C., and Kirkegaard, K. (1991) Trends Genet. 7, 186–19149. Arnold, J. J., and Cameron, C. E. (1999) J. Biol. Chem. 274, 2706–271650. Weiss, B. G., and Schlesinger, S. (1991) J. Virol. 65, 4017–402551. Raju, R., Subramaniam, S. V., and Hajjou, M. (1995) J. Virol. 69, 7391–740152. Igloi, G. L. (1983) Anal. Biochem. 134, 184–188

Mechanisms of Replicative RNA Recombination 8755

at Royal M

elbourne Inst of Technology (C

AU

L) on M

ay 30, 2014http://w

ww

.jbc.org/D

ownloaded from