J. Mol. Biol. (1998) 276, 7±17

The Natural 6 S RNA Found in Qbbb-Infected Cells isDerived from Host and Phage RNA

Elita Avota1, Valdis Berzins1, Elmars Grens1, Yuri Vishnevsky1

RuÈ diger Luce2 and Christof K. Biebricher2*

1Biomedical Research andStudy Center, Ratsupites 1LV-1067 Riga, Latvia2Max-Planck-Institute forBiophysical ChemistryD-37070 GoÈttingen, Germany

0022±2836/98/060007±11 $25.00/0/mb

The RNA of Escherichia coli infected with RNA bacteriophage Qb was iso-lated and screened for replicable short-chained RNA. In contrast to ear-lier assumptions we show that (i) short-chained replicable RNA is a veryminor part of the RNA synthesized in the infection cycle, and (ii) that thereplicable RNA isolated from infected cells is derived from cellular RNA,in particular 23 S rRNA and 10 Sa RNA, and from Qb RNA itself. Noneof the many RNA species known from in vitro experiments was found.The RNA species isolated were all inef®cient templates. No replicableRNA could be isolated from non-infected cells. Even in cells expressinghigh amounts of Qb replicase very few RNA species could be isolated.RNA generated in vitro in template-free synthesis is therefore not derivedfrom RNA species found in vivo, and replicable RNA found in vitro isgenerated by a mechanism fundamentally different from the one operat-ing in vivo.

# 1998 Academic Press Limited

Keywords: Qb replicase; 6 S RNA; template activity; bacteriophage Qb

*Corresponding author

Introduction

RNA bacteriophages replicate their genome by asingle enzyme, the RNA replicase. The replicases,e.g. that of bacteriophage Qb, were found to behighly speci®c for their viral RNA while ``ignoringthe RNA of the host cell'' (Spiegelman, 1971).Despite its high discrimination power, Qb replicasedoes accept a number of other RNA molecules astemplate. Poly(C) or C-rich nucleotide copolymersare accepted as template, but the reaction endsafter synthesis of the complement to the template.If the replica strand separates from the templateand can itself serve as a template, autocatalyticampli®cation by RNA replication can take place.

Beside Qb RNA itself, several short-chainedRNA species ef®ciently replicated by Qb replicasehave been isolated in vitro and their sequencesdetermined (Mills et al., 1973, 1975; Schaffner et al.,1977; Biebricher, 1987; Biebricher & Luce, 1992)and their replication mechanism has been studiedin detail (Dobkin et al., 1979; Biebricher, 1986,1987). Much less is known about non-viral RNAspecies formed in vivo by Qb replicase. Banerjeeet al. (1969) isolated a heterogeneous RNA fractionwith an average sedimentation coef®cient of 6 Sfrom Qb-infected cells; it was absent in non-infected cells and replicated by Qb replicase in vitro.While to our knowledge no further attempts were

971496

taken to characterize this 6 S RNA, it was generallyassumed that short-chained RNA formed in vitroand in vivo would resemble one another.

A seemingly unlimited number of different RNAspecies replicated by Qb replicase are generated bytemplate-free synthesis, i.e. in vitro incubation inabsence of extraneously added template with NTPsubstrate for long times (Sumper & Luce, 1975;Biebricher et al., 1981a,b). The emergence of theseRNA species could not be accounted to residualRNA impurities left in the highly puri®ed enzymepreparations used for these studies (Biebricher et al.,1986, 1993). With two exceptions (Munishkin et al.,1988; Moody et al., 1994), signi®cant sequencerelationship of the replicable RNA species to thegenomes of the virus and its host could not befound. (A third possible exception, MDV-1 (Millset al., 1973), has some sequence relationship to QbRNA, but to a similar degree also to other RNA ofphylogenetically distant sources.)

Analysis of replicating RNA generated in vivo,however, is only feasible if sample contaminationwith replicating RNA from extraneous sources canbe excluded, requiring extremely careful workunder special conditions (Biebricher et al., 1993).

In this paper we describe the cloning andsequencing of RNA species isolated from Qb-infected Escherichia coli cells using a novel cloning

# 1998 Academic Press Limited

8 6 S RNA in Q�-Infected Cells

procedure of replicating species (Biebricher &Luce, 1993; Rohde et al., 1995).

Results

Cloning and sequencing of RNA isolated frominfected cells

At the time of infection, no RNA that can beampli®ed exponentially at a measurable rate by Qbreplicase should be found in the cell. When hostcells are infected with Qb phages, only the phageRNA enters the cell, because experiments withcarefully puri®ed Qb phages have shown that thephage particles are free of other replicable RNA(Biebricher et al., 1993). The process leading to gen-eration of short-chained replicable RNA beginsindependently in each infected cell. If it is a rare,non-instructed process, formation of a large varietyof different replicable species would be expected.In each cell, only one replicable species wouldeventually survive, because less ef®cient templateswill be displaced by competition. Within the popu-lation of infected host cells, however, less ef®cienttemplates are protected from competition by thecompartmentation and can grow out. Once thewalls of infected cells are broken and their contentscombined, subsequent ampli®cation of the RNAin vitro by Qb replicase leads to the selection ofonly the most ef®cient templates. Thus, ampli®ca-tion of the isolated RNA by Qb replicase had to beavoided after a representative RNA populationwas collected.

RNA from three sources was investigated: (i)from infected cells late in the infection cycle(40 minutes), (ii) from non-infected cells, and (iii)from recombinant cells overproducing Qb repli-case. It is expected that the contents of recombinantcells overproducing replicase are the same as ininfected cells, except that other parts of the Qb gen-ome are missing in the former and the replicaseconcentration there is higher by more than an

Table 1. Sources of RNA templates

E. coli strain Treatment

AB259(F�) InfectedAB259(F�) InfectedAB259(F�) InfectedAB259(F�) InfectedAB259(F�) InfectedAB259(F�) InfectedHfr H InfectedDHa5(pREP) InducedDHa5(pREP) InducedDHa5(pREP) InducedAB259(F�) Non-infectedAB259(F�) Non-infectedAB259(F�) Non-infected

The conditions for growth, infection andand Methods. RNA isolation: (a) by protonate±phenol extraction. C: RNA complexeby nitrocellulose ®ltration before cloning; Rcloning.

order of magnitude. The sources of the RNA prep-arations and the RNA extraction methods are listedin Table 1.

Care was taken to isolate the RNA unbrokenfrom the bacterial cells and to avoid proceduresthat could lead to preferential loss of certain RNAfractions. Strong denaturants were used to preventRNases, in particular RNaseI, from degrading theRNA. The isolation avoided high salt concen-trations to prevent precipitation of RNA. In a fewcases enrichment of the short-chained RNA byfractional precipitation was used to get morematerial. Electrophoretic separation of RNAextracted from infected and non-infected cellsshowed predominantly 23 S and 16 S rRNA andtRNA (Figure 1). The patterns of RNA extractedfrom the three different sources are identical exceptfor the Qb RNA band seen in RNA isolated frominfected cells. We found that with the guanidiniumisothiocyanate±acid phenol method (Eurogentec)the recovery of Qb RNA was poor; most was lostin the procedure. However, single- and double-stranded variant RNA preparations were quantitat-ively recovered by this procedure (not shown).

Replicable short-chained RNA is thus only avery small fraction of the total RNA found. There-fore, the RNA population had to be screened forreplicating species. The selective cloning methodchosen (Biebricher & Luce, 1993) was based on thefact that replicating species, both plus and minusstrands, have invariant termini: At their 50 ends thesequences (G)GG, at their 30ends CC(A). Since thiscondition is necessary but not suf®cient for replica-tion, the ability of the isolated sequences to repli-cate had to be con®rmed later. The cloningprocedure allowed production of RNA of the samesequence and polarity as the template RNA byin vitro transcription from their cDNA clones.

Preliminary cloning experiments, however,showed no clones at all from cDNA isolated fromnon-infected cells and only a small number ofclones from infected cells. Because the procedure is

Isolation method Designation

a I1-a I1C-a I1R-a I4-b I2-b I5-b I3-a T1-b T2-a T1C-a N1-b N2-a N1C-

RNA isolation are described in Materialsplast lysis; (b) by guanidine±isothiocya-d in vitro to Qb replicase and isolated: RNA ampli®ed by Qb replicase before

Figure 1. Gel electropherogram of total RNA prep-arations used for cloning. 5 � 108 cells were collected bycentrifugation, solubilized for 15 minutes in 50 ml 1%sodium dodecylsulfate/1% b-mercaptoethanol andloaded on a polyacrylamide gel containing in the upperpart 3.5%, in the lower part 12.6% acrylamide. Runningbuffer was 0.1 M Tris borate, 0.1% sodium dodecylsul-fate. A,B: non-infected; C,D: infected with Qb phage;A,C: E. coli HfrH; B,D: E. coli Q13; Q: Qb phages with1% sodium dodecyl sulfate, 1% mercaptoethanol. 5 mland 10 ml of each sample were loaded. R: tRNA fromyeast (Sigma). Note that Qb RNA has a higher mobilitythan 16 S rRNA. When the RNA was heated prior toelectrophoresis, the mobilities of 23 S and 16 S rRNAremained the same, while Qb RNA moved as a doubleband more slowly than 23 S rRNA. After standing, thecompact structure of Qb RNA was restored.

Figure 2. Gel electropherogram of PCR products priorto ligation into the vector. The cloning procedure was asdescribed previously (Biebricher & Luce, 1993; Rohdeet al., 1995): RNA was polyadenylated at its 30-terminusheated to 115�C with the ®rst strand primer in a sealedplastic capillary and retrotranscribed with AMV reversetranscriptase. The template RNA was destroyed by heat-ing in 30% piperidine and the cDNA isolated. Secondstrand synthesis was with a primer containing thedouble-stranded T7 promoter sequence and a 30-terminalGGG overhang and T7 DNA replicase lacking 30-exonu-clease (``Sequenase''). Products of preparation I3(Table 1) were ampli®ed by PCR as described in theMethods section and analysed by gel electrophoresis ona 12.6% polyacrylamide gel (lane A). Lane L are chainlength markers. The strongest bands correspond tocDNA with chain lengths of 58, 130 and 330.

6 S RNA in Q�-Infected Cells 9

known to work well for replicating RNA(Biebricher & Luce, 1993; Zamora et al., 1995;Rohde et al., 1995), the amount of short-chainedreplicable material must have been small; speci®campli®cation was therefore necessary. Because thepolymerase chain reaction at the DNA level is syn-chronized and shows little selection, it was used toamplify the cDNA before cloning. Experimentsshowed that PCR worked well on cDNA clones ofMNV-11 (86 nt) and SV-5 (187 nt) and SV-7(135 nt) at a MgCl2 concentration of 6 mM in theampli®cation buffer.

Electrophoretic analysis of the ampli®cationproducts derived from infected cells indicated abroad distribution of DNA, over a range from 50to 700 nt, with several well-de®ned bands(Figure 2). cDNA ampli®ed from different RNApreparations showed different patterns of cDNAproducts, con®rming the stochastic nature of gen-erating the RNA species in vivo. After ampli®ca-tion, a satisfactory number of clones for

characterization was obtained. Clones with insertswere used for sequencing; however, care wastaken to avoid arti®cial selection, i.e. positiveclones were sequenced at random. A fewsequences had garbled termini or contained ®rststrand primers on both ends; those sequenceswere considered as cloning artefacts and werenot further investigated and are omitted from thesequence tables.

Sequences of the cloned material

As expected, a broad spectrum of sequences wasfound, ranging in chain length from 25 to 200. Atthe outset we anticipated ®nding RNA species thathave been found in in vitro experiments. However,these were extremely rare: Only one RNA prep-aration isolated from infected cells showed speciesclosely resembling the sequence of MNV-11 orfragments thereof (Biebricher, 1987). Since otherRNA was usually found in all preparations fromthe source that gave MNV-11-like RNA, it is poss-ible that MNV-11 (which is regularly investigatedin our laboratory) had been introduced accidentlyas an impurity into that preparation. If so, it is notclear why only MNV-11 mutants not found beforein the mutant distribution (Rohde et al., 1995)appeared, why all clones were in minus strand

Figure 3. Sequences of cDNA clones related to the Qb genome or RNA ``variants'' found in vitro to replicate with Qbreplicase. The RNA preparations are listed in Table 1. An r after the name of the clone means that the clone wasfound in the minus strand polarity. Abbreviations: �: nucleotide corresponding to the sequence; ±: nucleotide wasdeleted; NÃ : nucleotide N was inserted before this position. Clones designated with a ? were assayed for the replica-tion with Qb replicase. Bold and underlined sequences were found to replicate.

10 6 S RNA in Q�-Infected Cells

polarity, and what is the origin of the short replic-able clones corresponding to a part of the MNV-11.

The second surprise was that cDNA clonesderived from Qb RNA were rare, and entirelyabsent from the 30-terminus where replicationstarts and the internal sites where the replicasebinds (Meyer et al., 1981; Barrera et al., 1993). Wefound only cDNA clones corresponding to a sec-tion in the Qb replicase gene. Several clones of thesame region were found from different RNA prep-arations, both from infected cells and cells overpro-ducing replicase, all of them in minus strandpolarity (Figure 3) and all replicable. The presenceof this species in cells overproducing Qb replicaseis remarkable: While replicase mRNA is certainlypresent in substantial amount in induced cells, theRNA species isolated were of antisense polarity.Neither the primary sequence nor the publishedsecondary structure (Beekwilder et al., 1995) gaveany indication for this peculiar selection of thisRNA.

Aside from the one example cited above, not asingle RNA species of the many that have beensequenced and characterized in various in vitroexperiments appeared. Since our cloning methodworks well with these species, preferential loss isunlikely. On the other hand, a large variety of newreplicable RNA species were found, all of themstrongly related to host RNA, in particular to thestable RNA species sedimenting with 4.5 S, 5 S,6 S, 10 S, 16 S, 23 S and to tRNA.

The largest group of these clones is related to23 S rRNA (Figure 4), but showed varying degreesof mutational sequence deviations, indicating areplication process. In all cases the numbers of

Figure 4. Map of E. coli 23 S rRNA showing the regionsthat were found to be cloned. Fat bars correspond toRNA species replicated by Qb replicase, thin lines tospecies that did not replicate or were not tested.

mutations were suf®ciently small to allow anunequivocal relationship to be derived (Figure 5).

Of 42 cDNA clones from the total RNA prep-aration of Qb-infected AB259(F�) cells, 28 hadsequences derived from 23 S rRNA sequences. Of15 cDNA clones from recombinant E. coli overpro-ducing Qb replicase, only three sequences wererelated to 23 S rRNA. Six cDNA clones withsequences related to 23 S rRNA were found innon-infected host cells, but none of them could byreplicated in vitro by Qb replicase. The sequencefound in phage-infected cells was complementaryto 23 S rRNA in only a few cases, a strong indi-cation that this sequence was formed by RNAreplication. It is likely that the cloning procedureselected against antisense strands, which readilyform partial double strands with the excess of 23 SRNA under the high-salt conditions of the enzy-matic polyadenylation step, and while polyadenyl-ation works well with both single and double-stranded RNA it is dif®cult to elongate RNAhybridized to a longer complement.

Figure 4 shows that the cDNA clones werederived from certain regions of the rRNA. A largenumber appeared at the 30 terminus of the 23 SrRNA. It is not only the most likely locus for start-ing copying by Qb replicase, but it is also readilypolyadenylated and triggers most of the ®rststrand synthesis in the cloning procedure. Reversetranscriptase is not very processive and fragmentsending with a C cluster are cloned easily. The near-est G cluster is 68 nucleotides upstream, andindeed eight sequences of this length were foundin phage-infected, in cells overproducing replicaseand even in non-infected cells. Despite the ratheref®cient binding of replicase to RNA of thatsequence (see below), none of them had detectabletemplate activity, even though some cDNA clonescontained several mutations. The same mutationsappeared often in different cDNA clones. It seemslikely that the mutations were introduced by repli-cation, which may have occurred in vivo with thehelp of host factors, but all sequences found wereof plus strand polarity. Three longer sequenceswith chain length 180 to 190 were found that couldbe replicated by Qb replicase, one of them in

Figure 5. Sequences of cDNA clones related to sequences of 23 S rRNA. n: position in the nucleotide sequence wasambiguous; otherwise the nomenclature is as indicated for Figure 3.

6 S RNA in Q�-Infected Cells 11

minus strand polarity. Correct termini for the repli-case were found, obviously created in the ampli®-cation process.

Four longer cDNA clones of replicable RNA(chain lengths 180 to 190), all of sense polarity,were found in a stretch between positions 1400and 1600 of the 23 S rRNA. Several other clones,with two exceptions all from infected cells, werefound, the replication rates of which were too lowto detect, one of them in minus polarity.

Five cDNA clones, none of them replicating,were found in a stretch between positions 800 and1000 of the 23 S rRNA. Three were found in non-

infected cells, one in cells overproducing replicaseand one in infected cells. Most end at a structurallyexposed CCCAAA in the position 1005 to 1010 ofthe 23 S rRNA. The 50-termini of these clones wereincorrect, and the sequences showed no further sig-ni®cant deviations, suggesting that these sequencesmay be cloning artefacts.

A single replicating species, 220 nt long, corre-sponding to the position 266 to 487 of the 23 SrRNA, was found in infected cells.

In contrast to 23 S, only two derivatives from16 S rRNA were found, both from non-infectedcells, with corrupted termini and without detect-

Figure 6. Sequences of cDNA clones related to the sequences of other stable RNA in the host cell.

12 6 S RNA in Q�-Infected Cells

able template activity, suggesting cloning artefacts(Figure 6). However, none of it was derived fromthe 30-terminus, which is most likely to give riseto artefacts. Clones derived from tRNAs, on theother hand, were relatively frequent. This wasexpected, because some tRNAs have the correctends for cloning, suggesting selection by the clon-ing method. In accord with this interpretation, allthese species had plus strand polarity and wereinactive in replication. There was one exception:

A very short (26 nt) RNA derived from tRNA2Glu

was found to have template activity. It is notclear, however, why cDNA clones derived fromtRNA were isolated predominantly from infectedcells while few were found in non-infected cells,and why their sequences contained mutations toofrequent to be due to errors in the PCR ampli®-cation step. It is possible that the modi®ed basesin the tRNA contribute to the mutations. Twospecies isolated previously in vitro (Munishkin

Figure 7. Sequences of cDNA clones related to the sequences of mRNA from the host cell.

Table 2. Replication rates of short-chained RNAsisolated from infected cells

Clone Origina % 10ÿ3 sÿ1b

I1-53 23 S RNA (267±486) 0.38I1-45 23 S RNA (1414±1606) 0.15I1-13 23 S RNA (1421±1608) 0.35I2-13 23 S RNA (1451±1608) 0.35I1-3 23 S RNA (1494±1608) 0.35I1-32r 23 S RNA (2652±2842) 0.40I1-1 23 S RNA (2726±2904) 0.80I1-24 23 S RNA (2717±2904) 0.88T1C-3 23 S-RNA (2842±2904) NDI1-4r Qb RNA (3543±3597) 0.70I1R-30* E. coli 6 S RNA (150±184) 0.20I1R-12* E. coli tRNAglu

2 (50±77) 0.20I1R-27* ? (MNV-11-like) 1.4MNV-11 Template-free synthesis in vitro 3.80

The replication ef®ciencies of in vitro transcripts from thecDNA clones by Qb replicase were determined by using themas templates in Qb replicase reactions in vitro. RNA transcriptswith asterix(*) are cDNA clones of total RNA extracted fromphage-infected cells that has been ampli®ed in vitro by Qbreplicase for ten minutes.

a Homology relation of cloned RNAs to native cell RNAsequences (nucleotide coordinates of homology regions on hostRNA).

b The replication rate % was measured from the slope of thelinear part of the nucleotide incorporation pro®les as RNAstrands synthesized per second and replicase molecule.

6 S RNA in Q�-Infected Cells 13

et al., 1988; Moody et al., 1994) that were relatedto tRNAVal or tRNA1

Asp were not found in ourcollection.

Several other RNA species corresponding tostable RNA species, e.g. 4.5 S RNA, 5 S rRNA and10 Sa RNA were found (Figure 6). A particularpiece of 10 Sa RNA found quite frequently hadtemplate activity. A few clones had sequences offragments of mRNA; they were not investigated indetail (Figure 7).

Characterization of isolated RNA

Banerjee et al. (1969) characterized the 6 S RNAafter amplifying the RNA with Qb replicase.RNA preparation I1 was also ampli®ed andcloned (designated as I1R). The diversity of theproducts was strongly reduced, only a few typeswere found. Half of the clones were mutants ofMNV-11, the rest of the clones were derivedfrom tRNAGlu

2 or from stable 6 S RNA. Thisexperiment shows that due to the strong selectionduring the ampli®cation, it is impossible toobtain satisfactory information about the originalreplicable population.

A prerequisite of RNA replication is binding toreplicase. Replication and binding are poorly corre-lated, however, for there are RNA species thatbind well, but do not replicate, while others repli-cate fairly well but bind only weakly to replicase.To enrich the RNA that has the strongest bindingto replicase, RNA preparations (I1, N1 and T1)from non-infected, infected and overproducingcells were incubated with replicase in the absenceof nucleoside triphosphates and the complexes iso-lated by ®ltration on nitrocellulose (Meyer et al.,1981; Werner, 1991). The RNA was isolated fromthe complexes and cloned as described. Again thediversity was found to be small. From all threeRNA preparations, only one band was seen in elec-trophoresis of all three samples (not shown) andthis material turned out to be the 30-terminal 62 ntpiece of 23 S RNA (T1C-3). We conclude that thispiece of RNA is bound particularly strongly to Qbreplicase.

The replication ef®ciency of the individual cloneswith Qb replicase was measured by reproducingthe original RNA strands by in vitro transcriptionfrom their cDNA clones by run-off synthesis(Milligan et al., 1987; Biebricher & Luce, 1993) anddetermining their replication kinetics. It wouldhave been too time-consuming to assay all RNA

species; representative clones (marked with ? inFigures 3 to 7), particularly those found severaltimes, were assayed instead (Table 2).

None of the species isolated had a replicationrate in the range measured for in vitro isolatedRNA species, e.g. MNV-11, MDV-1, MNV-12, SV-5or SV-7; they were much lower. This made themeasurements dif®cult, because a single strand ofan optimized variant grows to macroscopicappearance (1010 strands) in just 20 minutes. Repli-cation rates were therefore measured at approxi-mately equimolar template/replicase ratios and forrestricted time periods of 10 to 20 minutes. Thereplication rates of the RNA clones shown inTable 2 refer to replication in the linear growthphase, where replicase is saturated with template.For some species, the ten minute period did noteven suf®ce to double the input RNA.

The danger of amplifying replicable impuritiesintroduced into the replication mixture made care-ful analysis of the replication products necessary.Transcripts and replication products were investi-gated by electrophoresis: If replication occurs, aband corresponding to the transcript should be

14 6 S RNA in Q�-Infected Cells

accompanied by another, stronger band corre-sponding to double-stranded RNA. During furtherincubation, both bands should increase in intensityconcurrently. As mentioned, the allowed time spanwas too low in many cases to observe this beha-viour. In such cases the mere appearance of adouble-strand band was taken as proof that repli-cation was taking place, while in cases where theinput RNA band was unaltered and no nucleotideswere incorporated the species were considered tobe non-replicating. With replicating species, onlythe double strand band was usually found(Figure 8). Single strands were not observed evenafter melting of the RNA prior to electrophoresis,because the separated strands rapidly re-combinedto double strands. It was proved by glyoxylationof the RNA prior to electrophoresis (McMaster &Carmichael, 1977) that the double strands con-tained full-length complementary strands: Oneband with the same mobility as the transcripts

Figure 8. Electropherogram of in vitro replication pro-ducts with Qb replicase. RNA was synthesized by run-off transcription from plasmid I3-10 cut at the 30 termi-nus of the cDNA with BstXI, using T7 RNA polymerase.An aliquot of the transcription mixture was used with-out puri®cation to inoculate a Qb replicase mixture(RNA:replicase ratio �1). Upper part: incorporation of[32P]ATP. Aliquots of the replication mixture wereloaded on a 12.6% polyacrylamide gel after the indi-cated incubation times. Upper left: transcription mixtureunlabelled, replication mixture labelled; Upper right:transcription mixture labelled, replication mixtureunlabelled; R1: labelled transcript; R2: a mixture oflabelled transcript and labelled replicate. Lower left:Electropherogram of I3-10 RNA denatured by glyoxyla-tion, stained by acridine orange. Conditions accordingto McMaster & Carmichael (1977). A: transcript; B: repli-cate; C: mixture of transcript and replicate. Lower right:Electropherogram of native I3-10 RNA on a 12.6% poly-acrylamide gel, stained with ethidium bromide. D: tran-script; E: replicate; R: SV5 replicate (187 nt, Biebricher &Luce, 1992).

used as template was found. The kinetics of con-version of labelled transcript template RNA todouble strand (Figure 8) shows that initiation ofsynthesis, presumably binding to RNA, is rate-determining. Once synthesis of a complementarystrand had been initiated, it rapidly was convertedto the full-length strand.

The low amount of synthesized strands in somecases made it dif®cult to distinguish between tran-scription into complementary strand and full repli-cation, because double strand formation is a deadend for replication (Biebricher et al., 1982, 1984).True replication only takes place if the negativestrands produced are also used as template. Thefact that the RNA was cloned predominantly inplus-strand polarity suggests that net RNA syn-thesis has taken place. However, it was found thatthe fraction of newly synthesized plus strandamong the in vitro replication products using tran-scripts as templates was consistently quite small(Figure 8). It is possible that in vivo replication isaided by host factors that bind to single-strandedRNA as in the case with Qb RNA itself.

Discussion

The majority of sequences found were nearlyidentical to fragments of the most abundant RNAin the cell. The question arises whether the clonesfound are really replicating RNA or merely arte-facts. Most of the sequences were not artefacts forthe following reasons: (i) It has been shown inmany experiments with a large variety of replicat-ing RNA species produced in vitro that replicatingspecies are ef®ciently cloned by the method; noselection bias has been observed (Biebricher &Luce, 1992; Rohde et al., 1995; Zamora et al., 1995).(ii) The majority of sequences, in particular thosewith template activity, were found repeatedly. (iii)The sequences had mutations which could not beattributed to the ampli®cation method. (iv) Thecomplementary sequence was occasionally found.

Why then were most sequences of plus strandpolarity? A likely interpretation is as follows. Wehave shown that most of the RNA species formdouble strands at a much higher rate than the opti-mized ``RNA variants'' formed in vitro, which areselected for slow double strand formation(Biebricher et al., 1982, 1984). Even though theRNA was heated to melt double strands, most ofthe replicating species are still lost to rapid doublestrand formation, except for the plus strands thatremain single-stranded because their complementshave formed partial double strands with the excessof cellular RNA. The 30-ends of the minus strandsin the partial double strands cannot be polyadeny-lated because they are recessed and therefore getlost in the cloning procedure.

How do the experiments presented here ®t tothe ®ndings of Banerjee et al. (1969)? The estimatedaverage chain length of 110 to 130 nt ®ts quitewell. Also, the material is heterogeneous as

6 S RNA in Q�-Infected Cells 15

described earlier, but the heterogeneity we foundafter PCR ampli®cation was higher than wasobserved previously. The main reason for this dis-crepancy is obviously the ampli®cation of the pro-ducts by Qb replicase prior to the analysis ofBanerjee et al. (1969) which leads to selection of themost ef®cient templates. The template activity ofthe 6 S was reported to be a third of Qb RNAitself, which is relatively low, because Qb RNA isreplicating more slowly than in vitro RNA variants.The only possible discrepancy is the visible amountof 6 S RNA after pulse-labelling infected cells,while the only difference of the total RNA ofinfected and non-infected cells we have seen is theadditional band of viral RNA. Banerjee et al. (1969)cited previous work where RNA sedimentingaround 6 S has been attributed to abortive syn-thesis and/or RNase degradation. Probably only afraction of the RNA sedimenting with 6 S observedby Banerjee et al. (1969) was replicable. Further-more, sedimentation through a glycerol gradient isa relatively crude fractionation method which pro-vides much less resolution than gel electrophoresis.

The results of this work are clearly not in agree-ment with two previous not experimentally veri-®ed assumptions:

First, the amounts of 6 S RNA that accumulatein infected cells late in the infection cycle (Banerjeeet al., 1969; Chetverin & Spirin, 1995) have appar-ently been overestimated. We found only smallamounts of replicable RNA species and most ofthem replicated at a low rate. Probably only asmall fraction of the infected cells contained anyshort-chained replicable RNA at all.

Second, the RNA isolated from infected cellsdoes not resemble the RNA in template-free in vitrosynthesis. A few clones resembling MNV-11obtained with the ®rst RNA preparation, but notin any others, were probably introduced as animpurity. Since the cloning method was originallydeveloped for in vitro products and works quitewell with them, this result cannot be due to theirpreferential loss in the cloning procedure. The PCRampli®cation procedure likewise worked withoutproblems for cDNA clones of the optimized var-iants MNV-11, SV-5 and SV-7.

If the products of template-free incorporationin vitro are not found in vivo, however, then thehypothesis that they originate from RNA impuri-ties in the replicase carried over from infected cellscannot be correct. We have shown in this workthat the short-chained replicable RNA species pro-duced late in the infection cycle of bacteriophageQb are derived from host RNA or from Qb RNAand replicate inef®ciently. The RNA bound moststrongly to replicase was the non-replicating 30-terminal fragment from 23 S rRNA. In contrast,template-free incorporation in vitro produces alarge variety of different RNA species, none ofwhich has signi®cant relationship to hostsequences (Biebricher & Luce, 1993). While the ®rstRNA products generated both in vitro and in vivoreplicate inef®ciently, they differ in chain length:

The ®rst detectable in vitro products have chainlengths in the range of 30 to 40 nt, while the pro-ducts isolated from infected cells have usuallychain lengths of 100 to 150 nucleotides or evenmore.

From the amounts of RNA found it seems thatthe appearance of this material is an infrequentevent. In other words: only a few infected cellscontain such RNA. In non-infected cells, replicableRNA was not found. It might seem plausible toassume that replicase occasionally accepts a hostRNA or a fragment of it and gradually optimizes itto a replicating RNA. If this were true, however,one should expect a high incidence of such RNA incells which express replicase in high amounts forseveral hours without the replicase productionbeing shut off by translation repression or lysis,but this was not observed. While a few replicableRNA species were found in overproducing cells,notably derived from the large amount of mRNAfor the replicase itself, the number of cDNA clonesobtained from such cells was smaller than thenumber of cDNA clones from infected cells. Sev-eral explanations for this are possible. It was foundpreviously that Qb replicase is unable to convert5.8 S RNA from yeast into a replicable molecule;only some nucleotides were added at the 30 termi-nus (Biebricher & Luce, 1992). However, RNArecombination has been observed, as a rare event,both in vivo (Munishkin et al., 1988; Palasingam &Shaklee, 1992) and in vitro (Biebricher & Luce,1992), probably by a copy choice mechanism. It isconceivable that replication started at the regulartemplate, plus or minus strand Qb RNA, and thenswitched rarely to a host RNA template, continu-ing synthesis there.

It was shown previously that Qb replicase aswell as the DNA-dependent RNA polymerase ofE. coli and the bacteriophages T7 and T3 condensenucleoside triphosphates more or less at random inthe absence of template, although at rates severalorders of magnitude more slowly than template-instructed RNA synthesis (Biebricher et al., 1986;Biebricher & Luce, 1996). Uninstructed conden-sation is suppressed by addition of an RNA orDNA that binds to the enzyme, whether it sup-ports template-instructed synthesis or not(Biebricher & Orgel, 1973; Biebricher & Luce, 1996).Accordingly, non-instructed synthesis by T7 RNApolymerase has been shown to be suppressedin vivo. This would account for the differencebetween the short-chained replicable RNA speciesfound in vitro to the species isolated from infectedcells.

As with most enzymatic processes, template dis-crimination is not an all-or-none process; RNAtemplate activities vary from optimal to negligible.As long as the synthesis products do not lead toautocatalytic ampli®cation, the limited templatespeci®city is innocuous. The emergence of replicat-ing RNA species from non-templates indicates thatability to replicate is very rare; neighbours insequence space may have a limited template

16 6 S RNA in Q�-Infected Cells

activity, and eventually an interfering replicatingRNA is formed. On the other hand, the probabilityof ®nding a template with good replication ef®-ciency is small: Indeed we did not succeed in opti-mizing any RNA species from cells to a replicationef®ciency comparable to in vitro RNA products.The replication ef®ciencies in vitro were always sosmall that signi®cant ampli®cation could not beachieved. The experiments described here showthat initiation of synthesis and separation of tem-plate and replica are the inef®cient steps; replicaelongation seems to proceed rapidly even if thestrands fail to separate permanently. It is likelythat RNA replication in vivo is enhanced by hostfactors; the replication ef®ciency of Qb plus strandwith our replicase preparation, which lacks thehost factor HFI, was also found to be quite small.

Materials and Methods

Growth of bacteria and isolation of RNA

Strain E. coli AB259(F�) or HfrH was grown in 2TYmedium at 37�C under aeration into the mid-log phase(absorbance A590 � 0.5) and infected with Qb phage at amultiplicity of 20. Aeration was continued for 40 min-utes and the cells harvested by centrifugation. DHa5cells harbouring plasmid pREP, overproducing the virus-encoded subunit of Qb replicase (Shaklee et al., 1988,kindly obtained by Dr P. Kaesberg) were grown in LBmedium at 28�C to an A590 of 0.5, the pL promoter wasinduced by raising the temperature to 42�C for 15 min-utes and aeration continued for one hour at 37�C. TheRNA was isolated from the cells by the protoplast lysismethod (Ausubel et al., 1992) or by the guanidinium iso-thiocyanate/phenol method (RNA Insta-pure kit fromEurogentec). Gel electrophoresis of the total RNA prep-aration showed predominantly the 16 S and 23 S rRNAand tRNA. Some RNA preparations were enriched inshort RNA by fractional sedimentation of templates withisopropanol according to Zubay (1966).

RNA cloning

Cloning was performed as described previously(Biebricher & Luce, 1993; Rohde et al., 1995), except thatthe cDNA was ampli®ed by the polymerase chainreaction with Taq DNA polymerase (Fermentas) for20 rounds (18 seconds 95�C, 18 seconds 55�C and42 seconds 72�C) using the primers pCGGAAGCTTC-CATTTTTTTGGG (®rst strand) and pGGGAATTCTAA-TACGACTCACTATAGGG followed by digestion withEcoRI and HindIII or by using the primers GAATTC-TAGGGATCCATTTTTTTGGG and GGGGGAAGCT-TAATACGACTCACTATAGGG followed by digestionwith BamHI and HindIII.

Sequencing was done by the dideoxy terminatormethod (Sanger et al., 1977) using the T7 sequencing kitfrom Pharmacia according to the protocol of the supplieror by a ABI-373A sequencer.

Determination of template activity

Recombinant plasmid DNA samples were preparedby the alkaline extraction method using QIAgen columnsaccording to the minipreparation protocol of the suppli-

er, cut with BstXI (Fermentas) at 55�C, extracted withphenol and precipitated. An aliquot of RNA prepared bytranscription with T7 RNA polymerase was used with-out further puri®cation for the Qb template assay asdescribed previously (Zamora et al., 1995). Care wastaken to avoid introduction of replicable impurities intothe assay mixture; controls without addition of templatewere included in each experiment. If no incorporationwas observed after incubating holo Qb replicase withsaturating amounts of template for 20 minutes, the tem-plate was considered to be non-replicable. Replicationrates in the linear growth phase were calculated fromincorporation pro®les of radioactively labelled ribonu-cleoside triphosphates.

Acknowledgements

The excellent technical assistance of Livija Gintnere isacknowledged. We are indebted to Dr Maija Bundule forhelp in the preparation of Qb replicase and for discus-sions and Dr William C. Gardiner (Austin) for criticalremarks of the manuscript. The support of this projectby the VW foundation (grant I/70 008) is gratefullyacknowledged.

References

Ausubel, F. M., Brent, R., Kingston, R. E., Moore, D. D.,Seidman, J. G., Smith, J. A. & Struhl, K. (1992).Short Protocols in Molecular Biology, 2nd edn., GreenePublishing Associates and John Wiley & Sons, NewYork, NY.

Banerjee, A. K., Eoyang, L., Hori, K. & August, J. T.(1969). Replication of RNA viruses. Replication of anatural 6 S RNA by the Qb RNA polymerase. J. Mol.Biol. 45, 181±193.

Barrera, I., Schuppli, D., Sogo, J. M. & Weber, H. (1993).Different mechanisms of recognition of bacterio-phage Qb plus and minus-strand RNAs by Qbreplicase. J. Mol. Biol. 232, 512±521.

Beekwilder, M. J., Nieuwenhuizen, R. & van Duin, J.(1995). Secondary structure model for the last 2domains of single-stranded RNA phage Qb. J. Mol.Biol. 247, 903±917.

Biebricher, C. K. (1986). Darwinian evolution of self-replicating RNA. Chemica Scripta, 26B, 51±57.

Biebricher, C. K. (1987). Replication and evolution ofshort-chained RNA species replicated by Qbreplicase. Cold Spring Harbor Symp. Quant. Biol. 52,299±306.

Biebricher, C. K. & Luce, R. (1992). In vitro recombina-tion and terminal elongation of RNA by Qbreplicase. EMBO J. 11, 5129±5135.

Biebricher, C. K. & Luce, R. (1993). Sequence analysis ofRNA species synthesized by Qb replicase withouttemplate. Biochemistry, 32, 4848±4854.

Biebricher, C. K. & Luce, R. (1996). Template-free syn-thesis of RNA species replicating with T7 RNApolymerase. EMBO J. 15, 3458±3465.

Biebricher, C. K. & Orgel, L. E. (1973). An RNA thatmultiplies inde®nitely with DNA-dependent RNApolymerase: Selection from a random copolymer.Proc. Natl Acad. Sci. USA, 70, 934±938.

Biebricher, C. K., Eigen, M. & Luce, R. (1981a). Productanalysis of RNA generated de novo by Qb replicase.J. Mol. Biol. 148, 369±390.

6 S RNA in Q�-Infected Cells 17

Biebricher, C. K., Eigen, M. & Luce, R. (1981b). Kineticanalysis of template-instructed and de novo RNAsynthesis by Qb replicase. J. Mol. Biol. 148, 391±410.

Biebricher, C. K., Diekmann, S. & Luce, R. (1982).Structural analysis of self-replicating RNA syn-thesized by Qb replicase. J. Mol. Biol. 154, 629±648.

Biebricher, C. K., Eigen, M. & Gardiner, W. C. (1984).Kinetics of RNA replication: plus-minus asymmetryand double-strand formation. Biochemistry, 23,3186±3194.

Biebricher, C. K., Eigen, M. & Luce, R. (1986). Template-free RNA synthesis by Qb replicase. Nature, 321,89±91.

Biebricher, C. K., Eigen, M. & McCaskill, J. S. (1993).Template-directed and template-free RNA synthesisby Qb replicase. J. Mol. Biol. 231, 175±179.

Chetverin, A. B. & Spirin, A. S. (1995). Replicable RNAvectors: prospects for cell-free gene ampli®cation,expression and cloning. In Progress in Nucleic AcidResearch and Molecular Biology (Cohn, W. E. &Moldave, K., eds), vol. 51, pp. 225±270, AcademicPress, San Diego, CA.

Dobkin, C., Mills, D. R., Kramer, F. R. & Spiegelman, S.(1979). RNA replication: Required intermediatesand the dissociation of template, product, and Qbreplicase. Biochemistry, 18, 2038±2044.

McMaster, G. K. & Carmichael, G. G. (1977). Analysis ofsingle- and double-stranded nucleic acids on poly-acrylamide and agarose gels by using glyoxal andacridine orange. Proc. Natl Acad. Sci. USA, 74,4835±4838.

Meyer, F., Weber, H. & Weissmann, C. (1981).Interactions of Qb replicase with Qb RNA. J. Mol.Biol. 153, 631±660.

Milligan, J. F., Groebe, D. R., Witherell, G. W. &Uhlenbeck, O. C. (1987). Oligoribonucleotide syn-thesis using T7 RNA polymerase and syntheticDNA templates. Nucl. Acid Res. 15, 8783±8798.

Mills, D. R., Kramer, F. R. & Spiegelman, S. (1973).Complete nucleotide sequence of a replicating RNAmolecule. Science, 180, 916±927.

Mills, D. R., Kramer, F. R., Dobkin, C., Nishihara, T. &Spiegelman, S. (1975). Nucleotide sequence ofmicrovariant RNA: another small replicatingmolecule. Proc. Natl Acad. Sci. USA, 72, 4252±4256.

Moody, M. D., Burg, J. L., Difrancesco, R., Lovern, D.,Stanick, W., Lingoerke, J., Mahdavi, K., Wu, Y. &Farrell, M. P. (1994). Evolution of host-cell RNAinto ef®cient template RNA by Qb replicase ± theorigin of RNA in untemplated reactions. Biochemis-try, 33, 13836±13847.

Munishkin, A. V., Voronin, L. A. & Chetverin, A. B.(1988). An in vivo recombinant RNA capable ofautocatalytic synthesis by Qb replicase. Nature, 333,473±475.

Palasingam, K. & Shaklee, P. N. (1992). Reversion of QbRNA phage mutants by homologous RNArecombination. J. Virol. 66, 2435±2442.

Rohde, N., Daum, H. & Biebricher, C. K. (1995). Themutant distribution of an RNA species replicatedby Qb replicase. J. Mol. Biol. 249, 754±762.

Sanger, F., Nicklen, S. & Coulsen, A. R. (1977). DNAsequencing with chain-terminating inhibitors. Proc.Natl Acad. Sci. USA, 74, 5463±5467.

Schaffner, W., RuÈ egg, K. J. & Weissmann, C. (1977).Nanovariant RNAs: Nucleotide sequence and inter-action with bacteriophage Qb replicase. J. Mol. Biol.117, 877±907.

Shaklee, P. N., Miglietta, J. J., Palmenberg, A. C. &Kaesberg, P. (1988). Infectious positive- and nega-tive-strand transcript RNAs from bacteriophage QbcDNA clones. Virology, 163, 209±213.

Spiegelman, S. (1971). An approach to the experimentalanalysis of precellular evolution. Quart. Rev. Bio-phys. 4, 213±253.

Sumper, M. & Luce, R. (1975). Evidence for de novo pro-duction of self-replicating and environmentallyadapted RNA structures by bacteriophage Qbreplicase. Proc. Natl Acad. Sci. USA, 72, 162±166.

Werner, M. (1991). Kinetic and thermodynamic charac-terization of the interaction between Qb replicaseand template RNA molecules. Biochemistry, 30,5832±5838.

Zamora, H., Luce, R. & Biebricher, C. K. (1995). Designof arti®cial short-chained RNA species that arereplicated by Qb replicase. Biochemistry, 34, 1261±1266.

Zubay, G. (1966). Isolation of Transfer RNA. In Pro-cedures in Nucleic Acids Research (Cantoni, G. L. &Davies, D. R., eds), pp. 455±460, Harper & Row,New York, NY.

Edited by K. Nagai

(Received 1 August 1997; received in revised form 17 October 1997; accepted 27 October 1997)