Nucleic Acids Research

Bacterial strains and plasmids

Plasmid pMS2000 contains full-length MS2 cDNA under control of the thermoinducible P_L promoter of phage [lambda] (12) and bacteria transformed with pMS2000 produce viable phage. Vector plasmid pD carries the linker fragment TCTAGAgttaacggtaccCTTAAG (lower case letters refer to non-phage sequence) in place of the phage cDNA XbaI-BfrI fragment (1303-1901). Cells transformed with this vector do not produce phage. Random PCR-directed mutagenesis was carried out on the phage XbaI-BamHI (1303-2057) cDNA fragment subcloned into vector pTZ19 and the resulting plasmid was designated pTN1. In plasmids used for measurements of replicase gene expression the MS2 cDNA fragments are located behind the P_L promoter and the lacZ gene is fused at the BamHI site to the replicase gene (8).

Clones containing the phage [lambda] P_L promoter were grown in Escherichia coli K12 strain M5219 (M72 trpA_am, lacZ_am, Sm^r/[lambda]dbio₂₅₂, cI₈₅₇[Delta]H1), which encodes the thermosensitive [lambda] repressor (cI₈₅₇) and the transcriptional antitermination factor N (13). Escherichia coli AB259 F⁺ cells (Hfr3000, Thi^-, Su^-) were used for amplification of phages. Escherichia coli JM109 [F[prime] traD36 proAB lacI^q[Delta](lacZ) M15/endA1] was the host strain for plasmid pTN1. Bacteria were grown on LB broth containing 10 g/l bactotryptone, 5 g/l yeast extract and 10 g/l NaCl and included an antibiotic if required.

Construction of vector plasmid pD

To construct plasmid pD 3 pmol phosphorylated oligonucleotides CTAGAgttaacggtaccC and TTAAGggtaccgttaacT were combined, incubated at 70°C for 5 min and kept at room temperature for an additional 15 min to produce a double-stranded (ds)DNA linker with sticky XbaI and BfrI ends. The linker was then ligated to XbaI/BfrI-digested plasmid pMS2000 DNA.

Mutagenesis and cloning

The oligonucleotides used in this study, donated by Dr E.Stankevich, were synthesized on a Gene Assembler and purified on C18-Pep RPC HR 5/5 columns (Pharmacia). Complementary primers p52 and p53 were used to randomize the region of interest on MS2 cDNA by two-step PCR. Oligo p52 was synthesized as CGCCGGnnnnnnnAACATGAGGATTACCCATGT (identical to the 1732-1764 sequence of the MS2 genome) and oligo p53 as CGACATGGGTAATCCTCATGTTnnnnnnnCCGG (complementary to the 1734-1766 sequence of the MS2 genome), where n stands for any nucleotide. In the first step two overlapping mutant MS2 cDNA fragments were created using primer pairs p52 and p109 (complementary to the 2042-2062 sequence of the MS2 genome) and p53 and reverse primer CAGGAAACAGCTATGAC (M13/pUC reverse sequencing primer; Fermentas). Reaction mixtures (100 µl) contained 5 µg/ml pTN1 plasmid DNA, 0.5 µM each primer pair, 200 µM dNTPs, 0.1 mg/ml BSA, 2.5 mM MgCl₂, 10 µl 10× PCR buffer (100 mM Tris-HCl, pH 8.8, 500 mM KCl, 0.8% Nonidet P40) and 2 U Taq DNA polymerase (Fermentas). The reaction was carried through 25 cycles: denaturation at 95°C for 45 s, annealing at 40°C for 45 s and elongation at 72°C for 90 s. The desired DNA fragments were then purified electrophoretically in a 6% polyacrylamide gel. In the second step, the two overlapping MS2 cDNA fragments from the first PCR were combined and PCR with flanking p109 and the reverse primer was carried out as above, except that the p52 and p53 primers as well as the pTN1 plasmid were omitted from the reaction mixture.

Mutants of the infectious clone were constructed by digesting the final PCR product with XbaI and BfrI restriction endonucleases and then ligating them to the corresponding restriction sites in the vector plasmid pD. The ligated DNA was used to transform E.coli M5219 cells. Transformed bacteria were seeded onto nitrocellulose filters (HATF, pore size 45 µm; Millipore) on top of LB agar plates containing kanamycin (50 µg/ml) and grown for 16-20 h at 28°C. This was done to screen the phage-producing clones by replica plating. Plaque-forming colonies were identified by replica plating the nitrocellulose filters on a lawn of E.coli AB259 F⁺ cells on 0.8% agar and they were then inoculated into 3 ml LB broth containing kanamycin (50 µg/ml) and grown overnight at 28°C. Dideoxynucleotide chain termination sequencing (14) of the target region of dsDNA was carried out with T7 DNA polymerase (Pharmacia) using primer p108 (identical to the 1628-1648 sequence of the MS2 genome).

Monitoring phage evolution and competition

Mutant phages were obtained after growth of the infectious phage cDNA clones overnight at 28°C. Diluted samples of the lysate containing ~10⁵ plaque-forming units (p.f.u.) of the phage were then inoculated into 3 ml E.coli AB259 suspension (at an OD₆₇₀ of ~0.2) and the cultures were shaken overnight at 37°C. Diluted samples of the obtained lysate (~10⁵ p.f.u. phage) were used for the next infection and so on. The infections were continued for 20-40 growth cycles. If the RNA sequence after various passages showed genetic heterogeneity in the phage population, plating was used to separate individual genomes for subsequent sequence analysis.

To test the fitness of a mutant phage it was mixed with wild-type phage in a p.f.u. ratio of 1:1 and subsequently the mixture was allowed to complete two passages on F⁺ bacteria. The phage RNA from the initial and the evolved samples was sequenced. The disappearance of the mutant sequence from the population is defined as a failure of the mutant to compete with the wild-type, whereas the preservation of mutant RNA in the evolved sample is defined as good competition.

Sequence analysis of phage RNA

Phage RNA was prepared from 1 ml lysate as described by Olsthoorn et al. (12) and dissolved in 20 µl water. Samples of 1-3 µl were taken for reverse transcription using primer p109 at 42°C for 30 min in 10 µl 50 mM Tris-HCl, pH 8.3, 4 mM KCl, 8 mM MgCl₂ and 10 mM DTT containing 1 U AMV reverse transcriptase (Pharmacia). The cDNA product from the reverse transcription reaction was PCR amplified with primers p109 and p180 (identical to the 1200-1217 region of MS2 genome) and then precipitated with 1 vol ethanol in the presence of 2.5 M ammonium acetate (15). The amplified cDNA (bases 1200-2062 of the MS2 genome) was directly sequenced with T7 DNA polymerase. Routinely, two oligonucleotides p108 and p105 (complementary to the 1572-1585 region of the MS2 genome) were used as primers for sequencing segments on either side of the MJ interaction, which included bases 1700-1900 and 1300-1500 on MS2 cDNA respectively. Other primers were also sometimes used to check the sequence of bases 1500-1700 in the phage genome.

Detection of replicase and lysis gene expression

Escherichia coli M5219 cells that harbour the specified plasmids with MS2 cDNA sequences behind the P_L promoter were grown until an A₆₇₀ of ~0.25 was reached. The P_L promoter was then induced by shifting the cultures to 42°C. To measure replicase expression, cultures were put on ice after 30 min induction and the [beta]-galactosidase activity was determined in 0.1 ml samples according to standard procedures (16). To monitor lysis gene expression the cell density at A₆₇₀ was recorded (17).

The experimental system

A complete cDNA copy of the MS2 genome is present on a plasmid under transcriptional control of the thermoinducible promoter P_L of phage [lambda] (12). The plasmid is maintained in E.coli M5219 F^- host cells, which produce phage spontaneously. Adverse mutations lower the titer of the phage. However, upon passaging mutants can quickly restore high viability by base changes resulting from the low copying fidelity of phage RNA replicase (18). These (pseudo)revertants may predominate through natural selection if compensation for the inflicted defect occurs.

The mutants

To disrupt the MJ interaction in the phage RNA, substitutions were introduced in the infectious MS2 cDNA clone in the 3[prime]-side of the LDI (Fig. 1a and Table 1). The 1738-1744 sequence codes for non-essential amino acids of the lysis protein (19) and was therefore preferred over the coat segment of the long-distance base pairing as the target for base changes. The mutations lowered phage fitness, as revealed by the decreased titers of the infectious MS2 cDNA clones (Table 1) and by their failure to compete with the wild-type during co-infections (not shown). From ~20 mutants sequenced, five (MJ2, MJ5, MJ19, MJ22 and MJ3) were taken for evolutionary analysis. The choice was based on our wish to have few as well as multiple substitutions present. MJ2 and MJ5, containing only two and three base changes respectively, would provide the phage with a reasonable chance to somehow re-establish the original regulatory LDI. Conversely, the multiple substitutions in MJ3, MJ19 and MJ22 are more likely to fully rearrange the structure and they render restoration of a MJ variant unlikely (17). This would force the phage to find an alternative to retain replicase expression under control. From the many mutants the specific choice of MJ5, MJ19 and MJ22 was made because these showed a low initial titer, which usually speeds up evolution. MJ3 was a random choice. The course of the evolutionary adaptation was followed for many growth cycles, as described in Materials and Methods.

Reconstruction of an MJ interaction

Several descendants of mutants MJ2 and MJ5 had reformed the MJ interaction (Fig. 1b, left). In one of them this led to wild-type phage (revertant MJ2.5) and this result is inconclusive. A more informative revertant is MJ2.7, which shows that the secondary structure of the phage RNA but not the sequence is important. Here, three base changes re-established the LDI, while only one of them is a true reversion. The G1737U substitution probably resulted in an extra U-A base pair for the interaction, albeit at the expense of the coat terminator (CT) hairpin.

Two revertants (MJ5.4 and MJ5.6) arose from mutant MJ5. Descendant MJ5.6 evolved a nearly native LDI but still preserved the U.G pair, where in the wild-type it is C-G. On the other hand, the C1732U second site substitution in the phage turned the mismatch of the CT hairpin into an A-U base pair. The CT hairpin possibly stabilizes the MJ interaction by coaxial stacking (8) and thus strengthening of this helix may help to suppress replicase translation. Perhaps, closing the mismatch is sufficient to compensate for the new U.G pair in the MJ interaction. Less straightforward is revertant MJ5.4. Besides two reversions in the LDI, this phage also acquired an A1746C substitution in the replicase initiator hairpin R33, a change that we cannot yet explain.

Progeny phages without a MJ interaction

The progeny of MJ19, MJ3 and MJ22, as well as one descendant of MJ2, did not restore the MJ interaction (right column of Fig. 1b), but their evolutionary adaptations strikingly resembled each other in that most of them had the U1817C base change some 50 nt downstream of the replicase start (MJ2.2, MJ19.2, MJ3.1 and MJ22.3). Pseudorevertants MJ2.2 and MJ19.2 also had in common the A1746C substitution. This base change was also seen in MJ5.4, and we will take up this mutation in the Discussion. Probably, all these second site mutations provide good compensation for absence of the MJ interaction.

Replicase control in pseudorevertants is restored

To verify this question the 1365-2057 cDNA fragment of MJ19.2 and also the same region of the original mutant MJ19 were fused to the lacZ gene in an expression vector under transcriptional control of the P_L promoter (8). In these constructs the start codon of the coat gene is absent and we therefore measured the degree to which the replicase start was buried by structure. [beta]-Galactosidase values for these clones (Table 2, first column) showed that the mutations in MJ19 indeed increased accessibility of the replicase start 3-fold. On the other hand, the mutations evolved in MJ19.2 decreased replicase synthesis ~6-fold and thus fully compensated for the inflicted activation of the replicase initiation site in mutant MJ19. In fact, MJ19.2 produced even less replicase than did the wild-type. These results demonstrate that evolution of pseudorevertants was directed towards suppression of replicase translation. The data, however, do not yet answer the essential question, i.e. whether the evolved mutations affected coat-replicase coupling. To resolve this point, the 1365-1628 sequence was deleted from the experimental constructs and replicase-[beta]-galactosidase fusion protein in the resulting clones was measured (Table 2, second column). Because the removed region involves that part of the coat gene contributing to the MJ interaction, replicase expression in the shortened variants is no longer repressed by coat gene sequences and thereby goes up as compared with the corresponding parental clones. This increase gives a quantitative measure for coupling (coupling factor, Table 2). In the wild-type, deletion of the coat cistron fragment raises replicase synthesis 9-fold (6). In MJ19 we expect an impaired long-range control due to disruption of the MJ interaction and this expectation is born out by the low coupling factor of three. The most important finding is that the suppressor mutations in MJ19.2 have restored the dependence of replicase gene translation on that of the coat. The coupling factor is back to eight. Thus, although the phage did not rebuild the MJ interaction, it did manage to restore coupling between the coat and replicase genes.

Table 2. Relative expression of the replicase gene

RlacZ	MS2 cDNA fragment		Coupling factora
	1365-2057	1628-2057
Wild-type	1.0	9.0b	9
MJ19	3.1	9.2	3
MJ19.2	0.5	4.0	8

The efficiency of replicase gene translation is measured as activity of replicase-[beta]-galactosidase fusion protein (RlacZ) using clones that contain either the 1365-2057 or 1628-2057 MS2 cDNA fragment fused to the lacZ gene at the BamHI site. The constructs lack the start of the phage coat cistron. In plasmids which begin at position 1365 of MS2 cDNA the coat gene segment that couples translation of replicase in wild-type phage is present (6). The RlacZ activity of the wild-type clone (8) is set at 1.0 (note that coupling is not absolute; 4). In constructs starting at position 1628 most of the coat gene is absent and replicase synthesis is uncoupled. Values are averaged from five experiments.
^aCoupling is calculated as the ratio between the RlacZ activities of the corresponding clones starting at 1628 and 1365.
^bData of Berkhout and van Duin (6).

The question arises as to how replicase translation is regulated in MJ19.2. As mentioned above, A1746C is not understood, but U1817C makes sense in the context of the structural model shown in Figure 1a, as it stabilizes the VD interaction. It then falls into the same category as C1732U, found in MJ5.6. Both U1817C and C1732U stabilize flanking stems that can apparently strengthen the core interaction in an indirect way. Alternatively, it is clear that ribosome binding to the replicase start requires melting of hairpin R32. Perhaps, melting of R32 is somehow antagonized by the stronger VD interaction.

To verify that MJ19.2 did not contain substitutions outside the monitored 1300-1900 sequence, its 1303-1901 cDNA fragment was cloned back in the full-length MS2 cDNA copy. The titer of this infectious MJ19.2 clone had increased ~25 times as compared with the parental MJ19 clone. Also, phage produced from the MJ19.2 plasmid competed well with the wild-type during co-infections. These results show that the two suppressor mutations that re-establish translational control are also entirely responsible for the high viability of MJ19.2.

In wild-type RNA translation of the lysis gene is also coupled to coat gene reading (20). This coupling is enforced by a local stem-loop structure that occludes the lysis start (21). It is worthwhile noting that in both mutant MJ19 and revertant MJ19.2 coupling is still intact, i.e. in clones lacking the coat start signals lysis protein is not produced, whereas the protein is made when coat gene translation takes place (not shown).

Molecular basis for restoring the coupling

A1772G and A1746C were found one and three times respectively. How they restore coupling is not clear. Computer analysis (23) indicates that A1746C and the mutations accumulated in MJ22.2 set up different pairings between the replicase start and the coat gene, but we have no further support for such structures. The suppressor mutations U1817C and C1732U occur four times and once respectively. They make sense as they stabilize the structures flanking the MJ pairing. It is difficult to say whether they act by stabilizing the MJ LDI directly through, for instance, coaxial stacking (8) or whether their presence physically interferes with ribosomes diffusing to the replicase start. In this respect it is worth mentioning that the replicase initiator hairpin R33 by itself is too weak to shield the replicase start (24).

Whatever the mechanism, for the VD interaction we have independent evidence that it is involved in repressing replicase translation. We made two mutations, G1421A and C1424A, in the VD structure at wobble positions in the coat gene. They disrupt the LDI and stimulate replicase translation, measured as a lacZ fusion, ~2-fold in a partial construct containing the 1365-2065 MS2 cDNA fragment. When these two mutations were introduced into an infectious clone, some of the revertants had acquired the U1817C suppressor mutation.

The results presented in this study show great similarity to a recent paper in which the importance of translational coupling between the coat and the lysis genes was analysed by the same evolutionary approach (17). From both studies it is clear that the coupling is not a fortuitous result of folding of the RNA. Rather, as in both studies coupling was restored by suppressor mutations, the mechanism must add to the fitness of the phage.

Purpose of the translational coupling

One problem in the life cycle of phage is that the genome must be able to switch from being a template for translation to being a template for replication. An RNA molecule can never be template for both processes at the same time, since they move in opposite directions. This problem is most likely avoided while ribosomes and replicase compete for a common site on the plus strand, i.e. the start of the coat protein cistron (25). Since lysis and replicase expression is coupled to coat gene translation, a replicase blocking the coat ribosome entry site effectively clears ribosomes from the three distal genes. The proximal maturation gene is not accessible to ribosomes in complete RNA, only on short non-equilibrated growing strands (26).

It is amazing to see how easy it is to re-invent these seemingly complex control mechanisms and how many different solutions there are. It is also remarkable that the perturbations caused by five to six substitutions can sometimes be neutralized by just two suppressor mutations.