ABSTRACT
The secondary structure of the RNA from the single-stranded RNA bacteriophages, like MS2 and Qb, has evolved to serve a variety of functions such as controlling gene expression, exposing binding sites for the replicase and capsid proteins, allowing strand separation and so forth. On the other hand, all of these foldings have to perform in bacterial cells in which various RNA splitting enzymes are present. We therefore examined whether phage RNA structure is under selective pressure by host RNases. Here we show this to be true for RNase III. A fully double-stranded hairpin of 17 bp, which is an RNase III target, was inserted into a non-coding region of the MS2 RNA genome. In an RNase III- host these phages survived but in wild-type bacteria they did not. Here the stem underwent Darwinian evolution to a structure that was no longer a substrate for RNase III. This was achieved in three different ways: (i) the perfect stem was maintained but shortened by removing all or most of the insert; (ii) the stem acquired suppressor mutations that replaced Watson-Crick base pairs by mismatches; (iii) the stem acquired small deletions or insertions that created bulges. These insertions consist of short stretches of non-templated A or U residues. Their origin is ascribed to polyadenylation at the site of the RNase III cut (in the + or - strand) either by Escherichia coli poly(A) polymerase or by idling MS2 replicase.
The single-stranded (ss)RNA phages, like MS2, are plus strand viruses containing a genome of ~3500 nt (Fig. 1 ). Apart from encoding the four proteins necessary for phage multiplication and spreading, the sequence also prescribes a certain RNA secondary structure that substantially contributes to the reproductive success of the molecule. For example, genotypically neutral mutations that affect critical aspects of RNA folding can lead to a dramatic loss in phage titer (1 ,2 ). Here we ask whether phage RNA structures, in general, represent a compromise since folding may also be subject to selection pressure exerted by host RNases. We showed, for instance, that large single-stranded regions are selected against, probably because they are sensitive to RNases (3 ). In this paper we focus on the question whether RNA phage structure is adapted to RNase III and, if so, how RNase III targets in the genome are inactivated.
RNase III has been found in eubacteria and eukarya (4 ). In bacteria the enzyme is encoded by the rnc locus and is involved in processing of rRNA from the precursor 30S transcript (5 ), but it also regulates expression of several operons, including its own (6 -8 ). RNase III cleavage sites have been determined for mRNA species of the DNA phages T3, T7 and [lambda] (9 ) but RNA from the ssRNA phages was reported as resistant to cleavage (10 ).
Though it is not precisely defined what the exclusive features of an RNase III site are, it has been well documented that fully double-stranded (ds)RNA structures of sufficient length are cut on one or both sides of the stem (11 ,12 ). A 3' overhang of 2 nt results when scission occurs on both sides. Here we introduce such long dsRNA structures in an intercistronic region of the MS2 genome. These mutant phages grow normally in an rnc- host, but not in the wild-type bacterium. There the RNA falls prey to RNase III, but is rescued by the replication machinery as an RNase III-resistant structure.
Plasmids containing the full-length MS2 cDNA and derivatives thereof were grown in Escherichia coli F- strain M5219 at 28oC as previously described (3 ). Phage were amplified in E.coli F+ strain KA797 or RCL1 at 37oC. RCL1, which is an rnc105 derivative of KA797, was obtained by transduction with a P1 lysate generously provided by P.Regnier. Transductants were screened for tetracycline resistance and assayed by SDS-PAGE for overproduction of PNPase (90 kDa), which is regulated by RNase III (13 ). The rnc105 mutant has a Gly -> Asp substitution that inactivates the protein.
Plasmids pMS.A, pMS.C, pMS.D, pMS.R and pMS.S are derivatives of the infectious cDNA clone pMS2000 (14 ), which contain a 34 bp insert at the XbaI site (position 1303 of the MS2 genome). The construction of pMS.A, pMS.C and pMS.D has been described before (3 ). Inserts R and S are made in the same way. pMS.F and pMS.H contain oligonucleotides C (5'-CTAGATAGAACTGCTAGCTAACTAGCTGTTCTTT) followed by D (5'-CTAGAAAGAACAGCTAGTTAGCTAGCAGTTCTAT) in the coding strand, while in pMS.H the order is reversed (DC). Phage generated by cells harboring pMS.R, pMS.S, pMS.F or pMS.H are denoted R.n, S.n, F.n or H.n respectively, where n = 0 specifies the starting mutants. Their evolutionary offspring are indicated by R.1, R.2, etc. This numbering is arbitrary and does not necessarily imply an evolutionary route. The single letter code is sometimes used for simplicity to indicate the starting mutant.
Phage obtained after overnight growth of M5219 cells containing a mutant MS2 cDNA clone (cycle 1) were plated on lawns of KA797 cells (cycle 2) as described (2 ). Subsequently, plaques were amplified overnight at 37oC in liquid cultures of KA797 in 2.5 ml LC broth. This is defined as cycle 3. Thereafter, phages were either further passaged or purified by PEG precipitation as described before (2 ).
Phages from rnc- cells were obtained as follows. Infectious plasmids (~50 ng) were directly transformed into RCL1 cells (F+) by means of electroporation (12.5 kV/cm). After incubating the transformed cells for 1 h at 37oC, 100 µl fresh RCL1 cells and 2 ml LC containing 50 µg/ml tetracycline were added and the total mixture kept with shaking overnight at 37oC (cycle 1). Appropriate dilutions were then plated either on lawns of RCL1 or KA797 cells (cycle 2). Individual plaques were picked and amplified in liquid cultures of RCL1 or KA797 cells (cycle 3).
Phage RNA was purified by phenol/chloroform extraction and precipitation with ethanol. The pellet was dissolved in H2O and this solution was used for reverse transcription in a total volume of 20 µl using primer DUI59 as described before (2 ). The resulting cDNA was amplified by PCR according to standard procedures as recommended by the supplier (Eurogentec), using primers DUI59 (complementary to nt 1409-1422) and DUI715 (homologous to nt 1182-1205). Samples were analyzed on 2% agarose gels. PCR fragments were sequenced with DUI59 after separation of the strands with Dyna beads (Dynal). Revertants from pMS.F and pMS.H were sequenced at the RNA level.
A sample of 1 µg phage RNA was briefly pretreated at 70oC to equilibrate the structure. Subsequently the sample was incubated with 0.5 U RNase III for 10 min at the indicated temperature in buffer containing 10 mM Tris-HCl, pH 7.5, 100 mM KCl, 10 mM MgCl2, 1 mM [beta]-mercaptoethanol and 5% glycerol. After extracting the RNA with phenol, extension of primer DUI59 was used to determine the cleavage sites. (RNase III was prepared in the laboratory of D.Court and made available to us by P.Regnier.)
Some time ago we prepared an infectious MS2 cDNA clone. Bacteria harboring this sequence in a plasmid produce overnight ~1011 infectious phage/ml culture. These phage are released into the medium. To control the problem of reinfection we generally use bacteria that lack the fertility (F) factor and can therefore not be attacked by these phage from without.
Viruses are known for their economic use of genetic material, as witnessed by the presence of overlapping genes in their genomes. MS2 is no exception and, as can be seen in the genetic map of Figure 1 , there are only three small non-coding regions. Position 1303, close to the maturation-coat spacing, was found to tolerate small insertions. The local structure of the wild-type RNA is shown in Figure 1 and the XbaI site just ahead of the maturation stop codon can be conveniently used to insert extra material (3 ). Actually, some of these inserts extend the C-terminus of the maturation protein somewhat, but there was never an indication that this affected phage growth.
Two of the inserts, whose fate was tested previously, are depicted in Figure 2 as A and C. It was then shown that type A inserts containing a large unstructured loop could not be stably maintained in the phage genome; deletions in the loop took place that reduced its size to between 4 and 7 nt. In contrast, equally sized structured insertions of type C were stable for many cycles (3 ).
At the time insert C was designed with two mismatches because we had never seen long uninterrupted stems in phage RNA (18 -20 ). We reasoned that the absence of perfect stems might be related to the activity of RNase III in E.coli. To test this idea we prepared inserts R and S in which the two mismatches have been replaced by Watson-Crick pairs. R and S differ in the loop sequence and 2 bp. F- cells were then transformed with plasmids carrying the inserts in MS2 cDNA.
Replacement of the mismatches by base pairs has dramatic effects; the titer of the respective plasmids pMS.R and pMS.S drops 1000 times as compared with pMS.C. Furthermore, whereas insert C is stably integrated into the MS2 RNA genome, insertions R and S are not. Ten plaques derived from the MS2 cDNA construct with insert S were sequenced. Nine had suffered a deletion that yielded wild-type and one turned out to be one triplet shorter than wild-type (S.1 in Fig. 3 ). (Mutant D will be discussed later.)
More direct evidence for RNase III pressure was obtained by in vitro experiments. Phage S.0, carrying the full insert S, was obtained from rnc- cells and the RNA isolated and probed with RNase III. Cuts were detected by inhibition of primer extension. The results are shown in Figure 5 . In the absence of RNase III at 37oC two strong stops can be seen at positions indicated with open arrows (Figs 3 and 5 ). As these stops do not occur when the long stem is disrupted (see below), we interpret these bands as stops induced by the high stability of the stem, preventing progression of reverse transcriptase. In the presence of RNase III at 37oC we see three bands at further downstream positions marking the enzyme cleavage sites, as shown by solid arrows in Figures 3 and 5 .
Above we showed that perfectly base paired inserts are indeed cleaved by RNase III. This then leads to the question whether the revertants have acquired RNase III resistance. To test this we isolated RNA from revertants R.8 and R.9 and from the parental phage R.0. This last RNA was, as before, obtained from rnc- cells. All three RNAs were probed with RNase III and the results are shown in Figure 6 . The sensitivity pattern obtained from parental R.0 is virtually identical to that of S.0 (Fig. 5 ). This is no surprise, since the difference between R.0 and S.0 is no more than two A-U -> U-A base pair changes and an inverted loop sequence.
We showed above that inserts R and S are stably maintained in the phage genome in rnc- cells. This fact allows us to study the evolution of the insert when integrated in the phage RNA genome. In the previous experiments the insert was only encoded in the MS2 cDNA.
Accordingly, rnc+ cells were infected with mutant phage R.0. As expected, the titer on rnc+ is three to four orders of magnitude down as compared with rnc-. Ten plaques were picked, amplified and sequenced. The revertants fall into three classes. In one class, represented by R.10-R.13, mismatches in the stem are created, presumably leading to decreased RNase III sensitivity (Fig. 7 ). These revertants look very much like phage carrying insert C or D (Figs 2 and 3 ), which were found to be stable in rnc+ cells while showing only weak RNase III cuts (indicated in Fig. 3 ).
We wondered how much more RNA could be stowed away in the genome and in the virion. Accordingly, double sized fragments containing uninterrupted stems of 34 rather than 17 bp were inserted in the XbaI site. The hairpins predicted to result from the inserts are shown in Figure 8 as F and H. Surprisingly, both hairpins remained stably incorporated in infectious MS2 phage for several cycles when propagated in rnc- strains. However, when rnc+ cells were transformed with the plasmid the titer dropped 1000-fold. A small number of plaques was analyzed. The revertants had lost the top half or more of the hairpin, as indicated in Figure 8 by the dotted lines. One phage (H3) carried an extra A, not present in the insert.
The genome of the ssRNA phages assumes a specific secondary structure that maximizes the reproductive success of the molecule. It does so by optimizing strand separation (22 ,23 ), gene expression (24 -26 ), replication (27 ,28 ) and probably many other processes. Here we have tried to understand to what extent these RNA structures are restrained by the fact that they must function in an environment where RNases operate.
Of all the nucleases characterized in E.coli, RNase III seems best suited to study its impact on shaping phage RNA structure. Defective mutants are available and rnc- cells grow normally. In addition, an RNase III substrate structure can be designed. Accordingly, various synthetic DNA fragments encoding a fully dsRNA hairpin were inserted in MS2 cDNA. In vitro experiments showed these structures to be RNase III targets.
Introducing the information for a 17 bp perfect stem in the infectious MS2 cDNA clone produced phage in possession of the full insert in an rnc- host. However, when the same plasmid was used to generate phage in an rnc+ host the insert could no longer be found in the RNA genome. Nevertheless, if the same insert contained two mismatches it was stably maintained. These experiments show that RNase III exerts selection pressure on the structure of phage RNA.
Although the fully double-stranded hairpin cannot be maintained in rnc+ cells, this does not mean that phage carrying this feature are doomed to die in this host. Revertants arise that have acquired resistance to the enzyme. This can occur in three different ways.
One major route is to reduce the length of the stem. Many times, but not always (see for instance S.1, F.2 and R.16), this leads to the wild-type. Another possibility is creation of bulges in the stem. This can be achieved either by small deletions or by insertions. The increased RNase III resistance of these structures was experimentally shown for R.8 and R.9 (Fig. 6 ).
Finally, escape from RNase III cleavage is attained by the creation of mismatches in the stem. This solution is illustrated by R.10-R.13. These revertants look very much like phage with insert C or D, which were found to be stable (3 ) and showed only weak RNase III sensitivity (Fig. 3 ).
Not all of our revertants belong to one class exclusively. Sometimes there is a mixture of the different resistance pathways. For instance, R.9 and R.16 have acquired a mismatch in addition to insertions and deletions respectively. Similarly, R.15 seems to have started with a C.C mismatch, created by a G -> C substitution. Possibly, this conferred insufficient protection and the 5 nt deletion UAUCU appeared as a further improvement.
The conversion of an RNase III-sensitive to an RNase III-resistant structure is the apparent result of an evolutionary process that selects for the presence of imperfections in the long regular stem caused by deletions, insertions or substitutions.
The insertions are non-templated and consist of small rows of Us or As. It is known that Q[beta] replicase has a tendency to add A residues to the growing chain when the template strand is lost or otherwise absent (29 ). Thus, revertant R.3 (Fig. 4 ) could originate as follows. The MS2 replicase copying the plus strand will face a template cut by RNase III at one of the positions found by us. The idling replicase adds a few As before it manages to engage with the template again across the cut. In the plus strand the extra As now turn up as an oligo(U) bulge at the site of the RNase III cut. Probably, these events can also happen during copying of the cleaved minus strand, resulting in addition of As in the plus strand, as seen for R.6 and R.9. Similarly, insertions of Us on the 5'-side of the stem (e.g. R.5) could result from RNase III cleavage on the 5'-side of the plus strand.
Another interesting possibility to account for the insertions is to consider the action of host enzymes on RNase III-cleaved plus or minus strand phage RNA. It has been shown that E.coli poly(A) polymerase (PAP) can add A residues to 3'-ends created by RNase III or to the exonucleolytically resected products (30 -32 ).
These two explanations, though tentative, are consistent with the fact that the insertions cluster at the measured or predicted RNase III cutting sites in the plus and minus strands. How the replicase crosses the cleavage site is enigmatic. Perhaps base pairing of the lower part of the stem keeps the two halves together, favoring a restart across the cleavage site (see below).
In many of the revertants the U/A insertions are accompanied by small deletions, the origin of which we do not know, but one could suggest that they result from imprecise replicase restarts across the site of the cut. Alternatively, these deletions are the result of 3' exonucleolytic degradation by RNase II or PNPase at the site of cleavage (31 ).
Except for R.2, these scars on the RNA are found on only one side of the stem at a time, suggesting either that templates are cut on only one side by RNase III or that only those templates that are cut once can be rescued. We think the latter possibility is the more likely one, as non-templated Us are found on either side of the stem.
The large majority of our revertants are wild-type, arising by complete deletion of the insert. This event is probably the result of RNA recombination favored by the XbaI repeat sequence, but could, when MS2 cDNA is the infectious agent, also take place at the DNA level. On the other hand, deletions such as those occurring in R.15 and R.16 cannot easily be explained by any feature of the sequence and they may reflect the action of exonucleases on the cleaved RNA.
Finally, suppressor mutations are obtained by simple replication errors that create mismatches (R.10, R.11 and R.12). Here it is at first glance surprising that the mutations consist of the less likely transversions. We think transitions are not found because they can only lead to A.C mismatches or U.G pairs. U.G pairs probably do not sufficiently distort the helix to prevent RNase III cleavage. The A.C mismatches likewise will turn into U.G pairs in the minus strand.
Recently Chetverin et al. (33 ) reported evidence for RNA recombination by spontaneous covalent linkage between two base paired RNA half molecules. In their model the 3'-OH group of the 5' half molecule attacks a phosphodiester bond in the 3' half molecule, releasing the nucleotides 5' of the splice site. In our system the two parts of the MS2 RNA genome are also held together by base pairs, which should favor the transesterification reaction. Furthermore, RNase III, as well as potential subsequent exonuclease, cleavage produces the required 3'-OH group. Many of our revertants could have been created by such a mechanism or by enzymatic joining of the half molecules (34 ). In particular, F.2 can be thought of as having arisen by end-to-end ligation after staggered RNase III cleavage (Fig. 8 ).
In the present study we have used two types of infectious agents. The insert, encoding the 17 bp perfect stem, was present either in infectious MS2 cDNA or in the intact phage RNA genome. The revertants obtained from either agent have in common that they all have developed RNase III resistance in rnc+ cells, but they differ in the sort of solution reached. This, we believe, must reflect their different histories in the cell. One obvious dissimilarity is that the pathway leading from MS2 cDNA to live phage necessarily involves transcription of the DNA by E.coli RNA polymerase. It would seem that this stage in phage production is highly sensitive to RNase III. This we deduce from the finding that infection with MS2 cDNA with insert R does not lead to a solution with mismatches. Apparently, in these cells there is simply no intact MS2 RNA transcript present in which to introduce the mismatches. Instead, all of the revertants, except wild-type, bear the marks of RNase III cuts in the form of non-templated A or U residues.
Our results show that RNase III exerts selection pressure on the structure of the RNA phage genome. However, the phage adapts easily to this pressure. Upon introduction of an RNase III target there is an initial 1000-fold drop in titer, suggesting that during every replication round there is a 0.1% chance of accidently creating a resistant phage. Adaptation to RNase III coincides with a return of the titer to a level indistinguishable from wild-type.
Although we have proposed several ways in which the insertions/deletions in the revertants could have arisen, the more attractive possibility seems that the initial RNase III cut earmarks the RNA for destruction by the combined actions of PAP and exonucleases (31 ). On the other hand, there is the replicase trying to save the RNA from extinction. The revertant sequences that are the outcome of these competing actions seem to offer a glimpse of the dynamics and intermediates involved in RNA degradation and replication.
We owe many thanks to Dr Regnier for generously providing us with RNase III, the P1 lysate for the preparation of an rnc- strain and for valuable suggestions. This work was supported by NATO grant HT.LG 94.0481 and by a fellowship to J.K. from the RNA Research Foundation.
*To whom correspondence should be addressed. Tel: +31 71 527 4759; Fax: +31 71 527 4340; Email: duin_j@rulgca.leidenuniv.nl
+On leave from the University of Latvia
REFERENCES