Pseudoknot-Dependent Translational Coupling in repBA Genes of the IncB Plasmid pMU720 Involves Reinitiation

Abstract

Replication of the IncB miniplasmid pMU720 requires synthesis of the replication initiator protein, RepA, whose translation is coupled to that of a leader peptide, RepB. The unusual feature of this system is that translational coupling in repBA has to be activated by the formation of a pseudoknot immediately upstream of the repA Shine-Dalgarno sequence. A small antisense RNA, RNAI, controls replication of pMU720 by interacting with repBA mRNA to inhibit expression of repA both directly, by preventing formation of the pseudoknot, and indirectly, by inhibiting translation of repB. The mechanism of translational coupling in repBA was investigated using the specialized ribosome system, which directs a subpopulation of ribosomes that carry an altered anti-Shine-Dalgarno sequence to translate mRNA molecules whose Shine-Dalgarno sequences have been altered to be complementary to the mutant anti-Shine-Dalgarno sequence. Our data indicate that translation of repA involves reinitiation by the ribosome that has terminated translation of repB. The role of the pseudoknot in this process and its effect on the control of copy number in pMU720 are discussed.

pMU720, a miniplasmid derived from a large, low-copy-number, conjugative plasmid pMU707, belongs to incompatibility group B (14). Its basic replicon consists of a 3.25-kb DNA fragment that encodes the functions necessary for autonomous replication and control of copy number. The frequency of replication of pMU720 depends on the expression of the repA gene, whose product is required for initiation of replication. Expression of repA is negatively regulated at the translational level by a 71-base antisense RNA, RNAI, which is transcribed from the noncoding strand of the DNA helix that encodes the leader region of repA mRNA (28, 29). Interaction of RNAI with stem-loop I (SLI), its complementary region in repA mRNA, results in inhibition of translation of this mRNA (28, 34).

The translation initiation region (TIR) of repA is sequestered within a stable secondary structure, stem-loop III (SLIII) (Fig. 1). This prevents the binding of ribosomes to the TIR, blocking the independent initiation of repA translation. Consequently, translation of repA has to be activated by ribosomes translating and correctly terminating a polypeptide encoded by an upstream gene, repB (29). It is the process of repB translation per se, rather than the amino acid composition of this leader peptide, that is important for activation of repA translation (29). Unfolding of SLIII by ribosomes translating and terminating repB facilitates formation of a pseudoknot immediately upstream of the Shine-Dalgarno sequence (SD) of repA. The pseudoknot, which forms by base pairing between two complementary sequences, one in the loop of SLI (the proximal sequence) and the other adjacent to the repA SD (the distal sequence), is an essential enhancer of translation of repA, and its positioning vis-à-vis the SD of repA is critical for activity. Insertion of as few as 6 bases between the pseudoknot and the SD abolished activation of repA translation, and a significant decline in translation was observed in mutants carrying insertion of a single base (40). The low efficiency of the initiation of repA translation appears to be an inherent property of its TIR, as mutations that disrupted SLIII and largely uncoupled expression of repA from that of repB did not free it from its dependence on the pseudoknot (40). By contrast, a mutation that improves the repA SD by increasing its similarity to the consensus sequence resulted in a significant level of expression that was independent of the pseudoknot (40).

FIG. 1.

(A) The leader region of the repBA mRNA of pMU720 (28), showing the major secondary structures, SLI, SLII, and SLIII, repB and repA SDs (in boldface type) and start codons (in boldface type and boxed), repB stop codon (in boldface type and underlined), and proximal and distal pseudoknot sequences. (B) Diagrammatic representation of the proposed pathway of regulation of repA expression. Diagram a shows the nascent repBA mRNA, with SLIII sequestering repA SD, the repA start codon, and the distal pseudoknot sequence, thus preventing expression of repA. Diagram b shows the ribosome elongating RepB and disrupting SLIII. Diagram c shows that the ribosome terminating at the repB stop codon facilitates formation of the pseudoknot, which activates translation initiation of repA. Diagram d shows that the interaction of RNAI with SLI preempts formation of the pseudoknot by sequestering the proximal pseudoknot bases; it also indirectly inhibits the formation of the pseudoknot by hindering the access of ribosomes to the TIR of repB. Hatched boxes, repA translational signals; solid boxes, repB translational signals; stippled boxes, pseudoknot sequences.

Although RNAI regulates expression of both repA and repB, its control of repA is much tighter than that of repB (29). Mutational analyses indicate that RNAI controls translation of repB by binding to SLI to produce a complex that excludes the ribosomes from the repB TIR by steric hindrance (41). The inhibitory complex involves pairing of the 5′ tail of RNAI with its complementary region in repBA mRNA, which lies immediately upstream of the repB SD. RNAI inhibits expression of repA not only through its regulation of expression of repB but also directly by binding to the bases in SLI that are necessary for the formation of the pseudoknot, thus preventing the formation of this tertiary structure. The binding of RNAI to the pseudoknot bases in SLI is proposed to be an early step in the binding reaction between these RNA molecules, whereas the pairing involving the 5′ tail of RNAI is a late step in this reaction (35). Moreover, although translation of RepB is essential for the expression of repA (29), stringent control of repB expression is not essential for effective control of the copy number of pMU720 (41), indicating that RNAI achieves this control primarily by binding to the proximal pseudoknot sequence to preempt formation of the pseudoknot.

A similar model has been proposed for the regulation of expression of repZ, the gene for the replication initiator protein of the IncI₁ plasmid ColIb-P9, which is closely related to pMU720 (2-8). The existence of the pseudoknot and its role as an activator of translation of repZ were first described in this system (3, 7). Recently, it was shown in vitro that the pseudoknot forms only when SLIII is disrupted by base substitutions, a situation mimicking disruption of SLIII by the ribosome translating the leader peptide (5). Synthesis of replication initiator protein RepA of the IncL/M plasmid pMU604 is also negatively regulated by an antisense RNA and is activated by a pseudoknot whose formation depends on translation and correct termination of a leader peptide (10, 11). However, pMU604 is only distantly related to the IncB and IncI₁ plasmids, so its nucleotide sequence, including the region encoding elements that regulate expression of repA, is significantly different from that of the other two plasmids (10).

Pseudoknots are involved in both inhibition and activation of initiation of translation in eubacteria. In the case of the well-studied Escherichia coli ribosomal proteins S4 and S15, which act to repress their own translation, a pseudoknot that forms in their TIRs has been shown to be the site of binding. The binding of these proteins to their respective operators stabilizes the pseudoknot, which in concert with the repressor traps the 30S subunit on the mRNA (reviewed in references 18 and 37). The mechanism of action of pseudoknots that activate initiation of translation of the replication initiator proteins of IncL/M, IncB, IncI₁, and related plasmids is not understood. The observation that disruption of the secondary structures that sequester the TIRs of these genes does not result in pseudoknot-independent translation initiation led to a proposal that a direct interaction between the ribosome and the pseudoknot improves the recognition of an inherently weak TIR (7, 40). However, if this is so, then it is unlikely that this interaction involves recognition by the ribosome of specific bases of the pseudoknot, since the sequence of the pseudoknot of the IncL/M plasmid is significantly different from those of the IncB and IncI₁ plasmids (3, 10, 29). This notion is supported by the observation that single base changes in the distal pseudoknot sequence do not significantly inhibit activation as long as compensatory substitutions are made in the proximal pseudoknot sequence (4, 9, 29).

In pMU720, the stop codon of repB lies 7 nucleotides (nt) downstream of the start codon of repA. This distance is too short to allow a second ribosome to bind to the repA TIR while a ribosome is still terminating translation of repB. Since SLIII is a simple hairpin, it should reform quickly when the terminating ribosome is released from the repBA mRNA. However, the existence of the pseudoknot, which sequesters some of the bases that form the 5′ side of the stem of SLIII, might slow this process sufficiently to allow a ribosome from the intracellular pool to initiate translation of repA. Alternatively, the pseudoknot might facilitate reinitiation at the repA start by the ribosome that translated repB. In this paper, we describe the use of the specialized ribosome system developed by Brink et al. (15), and Hui and de Boer (20), and Hui et al. (21) in an effort to reach a better understanding of the molecular basis for the activation of repA expression by the pseudoknot. We find that translation of repA occurs mainly through reinitiation by the ribosome that has completed translation of repB. Possible ways that the pseudoknot might facilitate this process are discussed.

MATERIALS AND METHODS

Bacterial strains, plasmids, and phages.

The strains of E. coli K-12 used in this study are given below. JM101 [Δ(lac-proAB) supE thi F′ (traD36 proA⁺B⁺ lacI^qZΔM15)] (24) was used for cloning and propagating M13 derivatives. XL1 Blue MRF′ {Δ(mcrA)183 Δ(mcrCB-hsdSMR-mrr)173 endA1 supE44 thi-1 recA1 gyrA96 relA1 lac [F′ proAB lacI^qZΔM15 Tn10 (Tet^r)]} (Stratagene) was used to grow M13 derivatives which had undergone mutagenesis as described by Vandeyar et al. (38). JP3438 (thr-1 leuB6 thi-1 lacY1 gal-351 supE44 tonA21 hsdR4 rpoB364 recA56) was used for propagating pMU720 derivatives. JP7740 (W3110 ΔlacU169 tsx recA56) was used for β-galactosidase assays.

The bacteriophage vectors used to clone fragments for DNA sequencing and mutagenesis were M13tg130 and M13tg131 (22). The plasmids used are described in Table 1.

TABLE 1.

Plasmids used in this study

Plasmid	Relevant characteristicsa	Reference
pMU720	Gal, IncB, miniplasmid	14
pMU525	lac′ZYA′, Tp, IncW, low-copy-number translational fusion vector	28
pMU1550	repA-lacZ translational fusion carrying nt 1-789 of pMU720; Tp, IncW, IncB	29
pMU1578	repB-lacZ translational fusion carrying nt 1-730 of pMU720; Tp, IncW, IncB	29
pIZ513	pACYC184 derivative carrying the specialized rrnB operon under the control of tac promoter and a lacIq gene; Cm	19
pIZ514	pACYC184 derivative carrying a deletion mutant of the specialized rrnB operon under the control of tac promoter and a lacIq gene; Cm	19

^aAbbreviations: Cm, chloramphenicol resistance; Tp, trimethoprim resistance; Gal, ability to promote fermentation of galactose. Mutations introduced into the repA-lacZ and repB-lacZ fusion plasmids are described in Results.

Media, enzymes, and chemicals.

The minimal medium used was half-strength buffer 56 (26) supplemented with 0.2% glucose, thiamine (10 μg/ml), and necessary growth factors. Enzymes and chemicals of a suitable grade were purchased commercially and not purified further. [³⁵S]dATPαS (>1,000 Ci/mmol) for use in sequencing was obtained from Amersham Corporation. Ampicillin was used at a final concentration of 50 μg/ml, trimethoprim was used at 10 μg/ml in minimal medium and 40 μg/ml in nutrient medium, chloramphenicol was used at 10 μg/ml, kanamycin was used at 20 μg/ml, isopropylthiogalactoside (IPTG) was used at 1 mM, and 5-bromo-4-chloro-3-indolyl-β-d-galactopyranoside (X-Gal) was used at 25 μg/ml.

Recombinant DNA techniques.

Plasmid and bacteriophage DNA was isolated and manipulated as described by Sambrook and Russell (30). DNA was sequenced by using a model 377 DNA sequencer and ABI Big Dye terminators (Perkin-Elmer Corporation) or by the method of Sanger et al. (31), which was modified in that T7 DNA polymerase was used instead of the Klenow fragment and terminated chains were uniformly labeled with [³⁵S]dATPαS. Oligonucleotide-directed in vitro mutagenesis reactions were performed on single-stranded M13 templates by using the method of Vandeyar et al. (38). The oligonucleotides were purchased from GeneWorks Ltd. DNA sequencing was used to screen for and confirm the presence of mutations.

Construction of the lacZ fusion plasmids.

The construction of the translational fusions (pMU1550 and pMU1578) has been described previously (29). pMU1550 (repA-lacZ) and pMU1578 (repB-lacZ) are derivatives of pMU525 (28) carrying nt 1 to 789 and 1 to 730 of pMU720, respectively. Thus, in the repA-lacZ and repB-lacZ translational vectors, codon 23 of repA and codon 29 of repB, respectively, are fused in frame with codon 8 of lacZ. β-Galactosidase expression in these fusions is therefore dependent on transcription from the rep mRNA promoter and translation from the fused gene (repA or repB). Mutant derivatives of these translational fusions were constructed by replacing the wild-type (wt) 789-bp EcoRI-BglII fragment of pMU1550 or the 730-bp EcoRI-BglII fragment of pMU1578 with the corresponding fragment carrying the mutation to be tested.

Measurement of β-galactosidase activity.

β-Galactosidase activity of mid-log-phase cultures was assayed as described previously by Miller (25). Each sample was done in duplicate, and each assay was performed at least six times. E. coli JP7740 cells harboring appropriate plasmids were grown in minimal medium supplemented with 0.4% glucose and antibiotics selective for the plasmids being tested. Whenever the specialized ribosome system was used, duplicate sets of cultures were grown at 37°C to early log phase (optical density at 600 nm, ∼0.2), 1 mM IPTG was added to one set, and incubation of all cultures continued for an additional 3.5 h.

Prediction of RNA secondary structures.

The mfold (version 3.1) computer program of Mathews et al. (23) and Zuker et al. (43) was used to predict RNA secondary structures.

RESULTS

The SD, which is usually located 5 to 9 bases upstream of the start codon in E. coli mRNA, plays an important role in the binding of ribosomes to the TIR of genes. This sequence is complementary to a region at the 3′ end of 16S rRNA (the anti-SD, ASD), and base pairing between SD and ASD enhances productive interaction between the ribosome and TIR. Specialized ribosomes are ribosomes whose ASD has been mutated so that it is no longer complementary to the consensus SD, 5′ UAAGGAGGU 3′. Consequently, these ribosomes can be directed to translate a specific subpopulation of mRNA molecules whose SD has been altered to be complementary to the mutant ASD (15, 20, 21). The IPTG-inducible specialized ribosome system (19) was used to target the TIR of repB or repA or both repB and repA. This system comprises a p15A derivative carrying the rrnB operon encoding a 16S rRNA whose ASD has been altered from 5′ ACCUCCUUA 3′ to 5′ ACACACUUU 3′. Transcription of this rrnB operon is driven by the tac promoter, and the presence of the lacI^q gene on the rrnB plasmid ensures tight regulation of expression of the specialized rRNA.

Translation of repB by specialized ribosomes.

The effect of changing the SD and flanking sequences of repB from 5′ UAcGGGaua to UAAGuGuGa (SDB^b) or UAcGuGuGa (SDB^c) (bases shown in capital letters can pair with wt ASD, those shown in boldface type can pair with the specialized ASD, and those in plain lowercase type cannot pair with either ASD) on the expression of β-galactosidase from the repB-lacZ translational fusion in the presence and absence of specialized ribosomes was determined. Induction of the specialized ribosomes had no significant effect on the expression of β-galactosidase from the wt repB-lacZ fusion (Fig. 2), which shows that these ribosomes were unable to initiate translation at the TIR of repB when the SD was not altered to match the specialized ASD. Introduction of the SDB^b mutation, which creates a contiguous 7-base region of complementarity to the specialized ASD, resulted in a 2.6-fold increase in the expression of repB-lacZ in the absence of IPTG (from 523 to 1,341 U) and a further 2.8-fold increase upon the induction of the specialized ribosomes (from 1,341 to 3,697 U). Introduction of the SDB^c mutation, which has a contiguous 5-base region of complementarity to the specialized ASD, had no significant effect under noninduced conditions but resulted in a twofold increase in β-galactosidase activity (from 511 to 1,036 U) upon induction. These data show that although the repB SDB^b fusion is expressed at higher levels than the repB SDB^c fusion, the specialized ribosomes are able to initiate translation at the TIRs of both repB mutants. This is an important consideration because the dependence of repA translation on the pseudoknot is due, at least in part, to its poor SD (40). It is thus possible that investigation of the role of the pseudoknot in repA translation might require the use of a suboptimal specialized SD.

FIG. 2.

Effect of replacement of the natural SD of repB by the specialized SDs, SDB^b and SDB^c, on the expression of β-galactosidase from the repB-lacZ translational fusion in the presence (induced) and absence (uninduced) of specialized ribosomes. β-Galactosidase activities were measured by the method of Miller (25), and the values shown are averages of at least six independent determinations. A schematic representation of the lacZ fusions used is shown. The native and specialized SDs, including flanking sequences, are shown (bases shown in capital letters can pair with wt ASD, those shown in boldface type can pair with the specialized ASD, and those in plain lowercase type cannot pair with either ASD).

The expression of β-galactosidase from repB-lacZ carrying the SDB^b or SDB^c mutation in the absence of IPTG was not due to leaky expression of the specialized rrnB operon under noninduced conditions, as it was also seen in cells that contained pIZ514, which carries a deletion in rrnB, as well as in cells that did not contain the rrnB plasmid (data not shown). Thus, the wt ribosomes were able to initiate translation at the specialized TIR, whereas the specialized ribosomes were not able to initiate at the nonspecialized TIR.

Effects of SDB^b and SDB^c mutations on expression of β-galactosidase from the repA-lacZ fusion carrying the wt SD.

The effect of changing the SD of repB to SDB^b or SDB^c on the expression of β-galactosidase from the repBA-lacZ translational fusion in the presence and absence of specialized ribosomes was determined. Induction of the specialized ribosomes had no significant effect on the expression of β-galactosidase from the wt repBA-lacZ fusion (Fig. 3). Introduction of the SDB^b mutation resulted in a 2.4-fold increase in the expression of repA-lacZ (from 113 to 274 U), but this was reduced by 1.7-fold (from 274 to 164 U) upon induction of the specialized ribosomes. Similarly, induction reduced the expression of β-galactosidase from the repBA-lacZ fusion carrying the SDB^c mutation by 1.4-fold. Thus, although an increased level of translation of repB by wt ribosomes resulted in an increased expression of repA-lacZ, an increase in translation of repB due to induction of the specialized ribosomes reduced initiation at the repA start. These data indicate that translation and correct termination of repB by the specialized ribosomes not only does not facilitate translation initiation at the repA TIR that does not carry the specialized SD but also interferes with this process. This interpretation supports the reinitiation model of translational coupling, where the same ribosome translates both of the coupled genes, which predicts that efficient initiation at the repA TIR requires that it contains the same type of SD (either native or specialized) as that contained by repB.

FIG. 3.

Effect of replacement of the natural SD of repB with the specialized SDs, SDB^b and SDB^c, on the expression of β-galactosidase from the repA-lacZ translational fusion in the presence (induced) and absence (uninduced) of specialized ribosomes. The repA gene in these fusions carried its natural SD. β-Galactosidase activities were measured by the method of Miller (25), and the values shown are averages of at least six independent determinations. Also shown is a schematic representation of the lacZ fusions used.

Effects of SDB^b and SDB^c mutations on expression of β-galactosidase from the repA-lacZ fusion carrying the specialized SD, SDA^b.

If translational coupling of repBA genes involves reinitiation, then substitution of the native SD of repA by a specialized SD should result in an increase in the expression of repA upon induction of specialized ribosomes, provided that the repB SD is also specialized. To test this prediction, the SD of repA was changed from 5′ UAAGcGaca 3′ to 5′ UAAGuGuGa 3′ (SDA^b), and the effect of this substitution on the expression of β-galactosidase from the repBA-lacZ translational fusion carrying native or specialized repB SD was determined. The SDA^b mutation is predicted to leave the structure of SLIII unchanged, replacing a C-G base pair with a U-G base pair, but to lower its stability (ΔG₀ [calculated free energy of formation of SLIII at 37°C]) from −16.8 to −14.5 kcal/mol (Fig. 4).

FIG. 4.

Predicted secondary structures and ΔG₀ values of SLIII carrying the SDA^b, SDA^d, or C704A mutation. The substitute bases are indicated by arrows. Ps⁻ indicates the presence of the C704A substitution, which abolishes the formation of the pseudoknot.

Introduction of the SDA^b mutation had no significant effect on the expression of repA from the wt repBA-lacZ fusion, both in the presence and in the absence of specialized ribosomes (Fig. 5). However, replacement of repB SD by SDB^b or SDB^c resulted in an approximately twofold increase in the expression of repA upon induction (from 299 to 589 U and from 143 to 275 U, respectively). Figure 5 also shows that in the absence of induction the expression of repBA-lacZ from the fusion carrying the SDB^b mutation was 2.7-fold higher than from the fusion carrying the native repB SD (299 compared to 111 U), showing once again a correlation between the levels of repB and repA translation by the wt ribosomes.

FIG. 5.

Effect of replacement of the natural SD of repA with the specialized SD sequence, SDA^b (5′ UAAGuGuGa 3′), on the expression of β-galactosidase from the repA-lacZ translational fusion in the presence (induced) and absence (uninduced) of specialized ribosomes. The repB gene in these fusions carried either its natural SD or the specialized SDs, SDB^b and SDB^c. β-Galactosidase activities were measured by the method of Miller (25), and the values shown are the average of at least six independent determinations. Ps⁻ indicates the presence of the C704A substitution, which abolishes the formation of the pseudoknot. Also shown is a schematic representation of the lacZ fusions used.

In order to determine whether translational coupling of repBA genes carrying the specialized SD mutations was dependent on the pseudoknot, the C704A mutation, which almost completely abolishes the formation of the pseudoknot in the wt repBA-lacZ translational fusion (29), was introduced. The C704A substitution is predicted to lower the stability of SLIII (ΔG₀) from −16.8 to −16.1 kcal/mol for the wt SDA and from −14.5 to −13.8 kcal/mol for the SDA^b mutant (Fig. 4). Under induced conditions, introduction of the C704A substitution reduced the level of β-galactosidase activity from the repA-lacZ SDA^b fusion carrying the SDB^b (from 589 to 369 U) or SDB^c (from 275 to 190 U) mutation by 1.6- and 1.4-fold, respectively (Fig. 5). These data indicate that in the presence of specialized ribosomes, 60 to 70% of the translation initiating at SDA^b of repA was no longer dependent on the pseudoknot. When the specialized ribosome system was not induced, the C704A substitution had a more severe effect on repA expression from the SDA^b mutant carrying specialized SDB, reducing activity by ∼70% (from 299 to 75 U and from 143 to 46 U). This difference in the degree of dependence on the pseudoknot of specialized and wt ribosomes initiating repA translation in the SDA^b mutant is most likely due to the better match between SDA^b and the specialized ASD than between SDA^b and the wt ASD sequences.

Effects of SDB^b and SDB^c mutations on expression of β-galactosidase from the repA-lacZ fusion carrying a specialized SD, SDA^d.

Since the high level of pseudoknot-independent repA translation by the specialized ribosomes in the SDA^b mutant was thought to be due, at least in part, to the good fit between the SDA^b sequence and the specialized ASD, a specialized SDA with less similarity to the consensus sequence was used. This sequence, 5′ UAcGuGuca 3′ (SDA^d), was used to replace the SDA of repBA-lacZ, and the level of expression in the presence of native or specialized SDB was measured. The SDA^d mutation is predicted to destabilize SLIII, changing its ΔG₀ value from −16.8 to −10.4 kcal/mol, and was expected to lead to significant uncoupling of repBA translation. This was indeed found to be the case, with 74% of the initiation of repA translation by the specialized ribosomes being independent of repB expression (data not shown). Therefore, the U720G substitution was introduced to restore base pairing between bases at positions 711 and 720 in SDA^d, increasing the predicted stability of SLIII to a nearly wt level (ΔG₀ of −16.1 kcal/mol) (Fig. 4). Thereafter, SDA^d was always used in combination with the U720G substitution.

Introduction of the SDA^d mutation reduced the expression of repA-lacZ from the wt repBA-lacZ fusion 9.4-fold, from 113 to 12 U (Fig. 6; compare with the wt data shown in Fig. 3), showing that the mutant SD is poorly recognized by the wt ribosomes. The induction of specialized ribosomes increased the level of expression approximately twofold (to 25 U), but this still represented only ∼20% of the wt level. This increase in expression, which represented 13 U of β-galactosidase activity, was presumably due to de novo initiation at the repA TIR by the specialized ribosomes. Replacement of repB SD by SDB^b increased the expression of repA-lacZ in the presence of specialized ribosomes to ∼130% of the wt level (124 compared with 92 U). In the absence of specialized ribosomes, the expression of repA-lacZ in this mutant was ∼40% (50 compared with 113 U) of the wt level. Replacement of the repB SD by SDB^c resulted in an increase in the expression of repA-lacZ to approximately the wt level in the presence of specialized ribosomes (90 U) but to only ∼20% of the wt level (25 U) in the absence of specialized ribosomes. These data show that efficient expression of repA carrying the SDA^d mutation requires the presence of specialized SDB and induction of the specialized ribosome system, supporting the reinitiation model of repBA coupling.

FIG. 6.

Effect of replacement of the natural SD of repA by the specialized SD, SDA^d (5′ UAcGuGuca 3′), on the expression of β-galactosidase from the repA-lacZ translational fusion in the presence (induced) and absence (uninduced) of specialized ribosomes. The repB gene in these fusions carried either its natural SD or the specialized SDs, SDB^b and SDB^c. β-Galactosidase activities were measured by the method of Miller (25), and the values shown are the average of at least six independent determinations. Ps⁻ indicates the presence of C704A substitution, which abolishes the formation of the pseudoknot. Also shown is a schematic representation of the lacZ fusions used.

The C704A substitution was introduced to determine whether the expression of repA from repBA-lacZ fusions carrying the SDA^d mutation is dependent on the pseudoknot. This substitution is predicted to have only a small effect on the stability of SLIII containing the SDA^d mutation, changing its ΔG₀ value from −16.1 to −15.4 kcal/mol (Fig. 4). Introduction of the C704A substitution reduced the level of β-galactosidase activity from the repBA-lacZ SDA^d fusion carrying the SDB^b or SDB^c mutation four- to fivefold both in the presence (from 124 to 28 U and from 90 to 22 U, respectively) and absence (from 50 to 11 U and from 25 to 5 U, respectively) of IPTG (Fig. 6). Thus, 70 to 80% of the translation initiating at SDA^d of repA in mutants carrying specialized SDB is dependent on the pseudoknot.

DISCUSSION

The translational coupling between the repBA genes of the IncB plasmid pMU720 is unusual in that it is dependent on the formation of a pseudoknot immediately upstream of the SD of repA (29, 40). This requirement is, at least in part, the consequence of the poor affinity of the repA SD for the ribosome, because mutations that increase the complementarity of this sequence to the ASD cause a significant level of pseudoknot-independent translation (40). Similarly, the dependence of translational coupling between the repBA genes of the IncL/M plasmid pMU604 on the pseudoknot is due to the poor affinity of the repA TIR for the ribosome, since improving the spacing between the SD and the start codon or changing the start codon from GUG to the more efficient AUG resulted in a significant level of direct coupling of repBA (11). On the other hand, the unusual positioning of the stop codon of repB vis-à-vis the start codon of repA does not contribute to the dependence on the pseudoknot (29, 39). The crucial question that remains to be answered is how the pseudoknot activates translational coupling of the repBA genes. We have used a specialized ribosome system, which directs a subpopulation of ribosomes to translate mRNA molecules that carry an altered (specialized) SD, to determine whether the pseudoknot acts by facilitating the binding of ribosomes from the cellular pool to the repA TIR or by promoting reinitiation at the repA start by the ribosome that terminates the translation of repB. Since the presence of SD stimulates reinitiation by approximately 10-fold (16, 36), the prediction is that efficient reinitiation would occur only when the coupled genes carried the same type of SD, either native or specialized, because the ribosome translating the upstream gene has to be able to bind to the SD of the downstream gene. On the other hand, if the only role of the ribosome translating the upstream gene were to facilitate formation of the pseudoknot, which in turn allowed binding of a ribosome from the cellular pool to the repA TIR, then it would not matter whether repB and repA carried the same or different types of SD. Our data fits the first prediction, supporting the reinitiation model.

Reinitiation is believed to occur when the terminated ribosome or its 30S subunit remains bound to the mRNA and scans in the 5′ or 3′ direction until it encounters functional start signals (1). The scanning model is supported by data showing that in vitro the ribosome that has terminated translation of a short synthetic mRNA does not disengage from the mRNA but slides back to the start of the mRNA to reinitiate its translation (27). Scanning is considered to be a passive process, inhibited by stable RNA structures, because scanning ribosomes do not have the energy necessary to unfold such structures (1, 13, 32). Thus, it is expected that in the repBA system the movement of a ribosome that has terminated translation of repB and is scanning in the 5′ direction will be blocked by the pseudoknot. The simplest scenario is that the pseudoknot traps the scanning ribosome in a position that is optimal for base pairing between the 3′ end of its 16S rRNA and the repA SD and orients it for initiation of repA translation. This interpretation is consistent with the data showing that the positioning of the repA SD vis-à-vis the pseudoknot is crucial for efficient translation initiation (40), whereas the distance between the repB stop and repA start is not important as long as the terminating ribosome is in a position to facilitate formation of the pseudoknot (29). Although the elongating ribosome is able to melt stable RNA structures, its progress may be slowed. If so, then by slowing the flow of ribosomes elongating RepB, the pseudoknot may further enhance the efficiency of reinitiation at the repA start by extending the time available to the reinitiating ribosome, thus decreasing the probability of its dislodgment by the next ribosome elongating RepB. Increasing the complementarity between the repA SD and the 3′ end of 16S rRNA results in tighter anchoring of the ribosome to the mRNA, which is predicted to decrease the likelihood of its dislodgement and make reinitiation less dependent on the pseudoknot, as is indeed the case.

In the IncL/M plasmid pMU604, formation of the pseudoknot reduced the translation of repB 3.5-fold (11). This was not the case in pMU720, where the pseudoknot had only a small effect (1.3-fold decrease) on repB expression (data not shown). The major difference between these two systems lies in the distances separating repB SD from the pseudoknot, which are 33 nt in pMU604 and 66 nt in pMU720. It is thus likely that, because of its close proximity, the pseudoknot hinders the access of ribosomes to the repB TIR of pMU604, whereas in pMU720, it is too far away to do so. Inhibition of repB translation may be important for repA expression in pMU604, because the spacing between the pseudoknot and repA SD in this plasmid is suboptimal for efficient translation initiation at the repA start (11). Once the pseudoknot has formed, reducing the flow of ribosomes over the restart region would be expected to increase the probability of the translation initiation of repA.

In the repBA mutants carrying specialized repB SD and native repA SD, induction of the production of the specialized ribosomes, which increased repB expression, inhibited the translation of repA by wt ribosomes. There are a number of other reports of translational interference, where ribosomes translating the upstream gene of a polycistronic mRNA inhibit translation initiation at the start of the downstream gene (12, 16, 33, 42). In the case of the translationally coupled coat-L genes of bacteriophage MS2, negative interference could be demonstrated only when translation of the L gene was made independent of translation of the coat gene (12). Under those conditions, synthesis of the L protein increased with decreasing translation of the coat gene. The naturally low efficiency of translational coupling between genes V and VII of bacteriophage IKe could be increased by decreasing the translation of the upstream gene (42). On the basis of these findings, it was proposed that during the process of termination and reinitiation, the 30S ribosomal subunit loses many of its contact points with mRNA and so becomes more susceptible to dislodgement by the elongating ribosome entering the intercistronic junction. This model predicts that for all translationally coupled genes there is a level of translation of the upstream gene that is optimal for efficient coupling and that an increase in translation above this level results in interference (42). In pMU720, where there is no repB-independent translation of repA (29), translational coupling is very efficient in the presence of the pseudoknot, as under derepressed conditions caused by the removal of RNAI, repA was expressed at 70% of the level of repB expression (39) and a reduction in the expression of repB caused a decrease in the expression of repA (29). In repBA, as in the coat-L pair of coupled genes, translational interference could be detected only when the specialized ribosomes translating the upstream gene were not contributing to the translation of the downstream gene. Moreover, translational interference increased as the level of repB translation by the specialized ribosomes increased.

Our data suggest that there is little, if any, initiation of repA by ribosomes from the cellular pool. This is not because the pseudoknot acts as a barrier to the entry of ribosomes from the pool, because when translation of repA was uncoupled from that of repB by the combined effects of the disruption of SLIII, through the introduction of mutation S3.4 (GC718AA GC725CA), and the premature termination of repB, it was still dependent on the pseudoknot (40). Under these conditions, the pseudoknot must have been facilitating the binding of ribosomes from the cellular pool to the repA TIR. It is worth noting that under derepressed conditions, the expression of repA-lacZ from this mutant was approximately twofold lower than that from the wt (40), indicating that facilitated binding of ribosomes from the cellular pool is less efficient than reinitiation. Introduction of mutation S3.4 is predicted to replace SLIII with a less stable structure (ΔG₀ of −5.8 kcal/mol), whose formation would be preempted by the formation of the pseudoknot. Although it has been suggested that significant inhibition of translation requires structures with stabilities greater than −5 to −6 kcal/mol (17), it might be expected, given the weakness of the repA TIR, that a structure with a ΔG₀ of −5.8 kcal/mol would be inhibitory. Although the pseudoknot does not hinder entry of a ribosome from the cellular pool, such a ribosome would not have access to the TIR of repA while the reinitiating ribosome was present. Presumably, by the time the reinitiating ribosome clears the repA TIR the pseudoknot has been disrupted by the next ribosome elongating RepB and the process of reinitiation starts all over again. The existence of SLIII makes translation of repA dependent on that of repB, which prevents the binding of ribosomes from the cellular pool to the repA TIR. Thus, the system appears to be finely poised to allow no more than one molecule of RepA to be made each time a pseudoknot forms. Formation of the pseudoknot is controlled by the intracellular concentration of RNAI. The requirement that the translation of repB be completed before the pseudoknot can form expands the window for effective RNAI binding, so that at high copy number, when the concentration of RNAI is saturating, pseudoknot formation is blocked and RepA synthesis is switched off. Since RepA acts in cis and multiple RepA molecules are required for initiation of replication (T. Betteridge, J. Yang, A. J. Pittard, and J. Praszkier, unpublished data), control of replication is very tight as multiple rounds of pseudoknot formation are required.