An RNA pseudoknot as the molecular switch for translation of the repZ gene encoding the replication initiator of IncIalpha plasmid ColIb-P9.

Translation initiation of the repZ gene encoding the replication initiator of plasmid ColIb-P9 is not only negatively regulated by the action of the antisense Inc RNA encoded in the leader region, but is also coupled to the translation and termination of a transcribed leader sequence, repY, a positive regulatory element for repZ gene expression. This translational coupling depends on base pairing between two complementary sequences, 5'-rGGCG-3' and 5'-rCGCC-3', which are located upstream of and in the middle of repY, respectively, and have the potential to form a pseudoknot with the stem-loop structure I. Another stem-loop called structure III near the 3'-end of repY sequesters both the 5'-rCGCC-3' sequence and the repZ ribosome-binding site. Here we show that the RepZ mRNA leader sequence synthesized in vitro indeed contains several stem-loop structures including structures I and III, but not the pseudoknot. However, disruption of structure III, without changing the repZ ribosome-binding site, by means of base substitution and deletion induces base pairing between the two short complementary sequences distantly separated, resulting in the formation of a pseudoknot. When the pseudoknot is allowed to form in vivo due to the same mutations, a maximum level of repZ expression is obtained comparable to one observed in the absence of Inc RNA. These results strengthen our previously proposed model that the pseudoknot induced by the translation and termination of the repY reading frame functions as the molecular switch for translational initiation of the repZ gene.

Translation initiation of the repZ gene encoding the replication initiator of plasmid ColIb-P9 is not only negatively regulated by the action of the antisense Inc RNA encoded in the leader region, but is also coupled to the translation and termination of a transcribed leader sequence, repY, a positive regulatory element for repZ gene expression. This translational coupling depends on base pairing between two complementary sequences, 5-rGGCG-3 and 5-rCGCC-3, which are located upstream of and in the middle of repY, respectively, and have the potential to form a pseudoknot with the stemloop structure I. Another stem-loop called structure III near the 3-end of repY sequesters both the 5-rCGCC-3 sequence and the repZ ribosome-binding site. Here we show that the RepZ mRNA leader sequence synthesized in vitro indeed contains several stem-loop structures including structures I and III, but not the pseudoknot. However, disruption of structure III, without changing the repZ ribosome-binding site, by means of base substitution and deletion induces base pairing between the two short complementary sequences distantly separated, resulting in the formation of a pseudoknot. When the pseudoknot is allowed to form in vivo due to the same mutations, a maximum level of repZ expression is obtained comparable to one observed in the absence of Inc RNA. These results strengthen our previously proposed model that the pseudoknot induced by the translation and termination of the repY reading frame functions as the molecular switch for translational initiation of the repZ gene.
Evidence has been accumulating that RNA pseudoknots, stem-loop structures containing an extended double helix with a complementary sequence outside the loop (for review, see Refs. 1 and 2), play an important role in control of gene expression at the translational level. In the Escherichia coli ribosomal protein ␣ and rpsO operons, RNA pseudoknots act as the binding sites for proteins S4 and S15, respectively, being responsible for autoregulation of the respective operon genes (3,4). In some mammalian retroviruses and in the mRNA for the rat ornithine decarboxylase antizyme, RNA pseudoknots function as regulatory signals for ϩ1 or Ϫ1 frameshifting or translational read-through to allow translation of sequences downstream of stop codons (5)(6)(7). In IncI␣ ColIb-P9 and IncB pMU720 plasmids, possible RNA pseudoknots have been proposed to be the molecular switches for translation initiation of the replication initiator genes, thereby controlling plasmid copy number in the host cells (8 -10).
ColIb-P9 is a low-copy number and self-transmissible plasmid with a size of 93 kilobases (kb). 1 The basic replicon of ColIb-P9 consists of a 3-kb DNA fragment which contains sufficient information required for autonomous replication and copy number control (Ref. 11, also see Fig. 1). The frequency of ColIb-P9 replication is limited by the degree of expression of the repZ gene encoding a 39-kDa replication initiation protein, RepZ, which is thought to react with the replication origin (11). Previous studies revealed that repZ expression was negatively controlled by the antisense Inc RNA of about 70 bases, the product of the inc gene that governs the phenotype of plasmid incompatibility (11,12). Inc RNA is transcribed from the noncoding strand of the leader region of repZ and binds to RepZ mRNA at the complementary region to form an RNA-RNA duplex (12). Two promoter down mutations, inc1 and inc2, altered the Ϫ35 and Ϫ10 regions of the inc gene promoter, p inc , and increased the level of repZ expression without affecting significantly the amount of RepZ mRNA (Refs. 11 and 12; see In an effort to elucidate the molecular events required for repZ expression, we also isolated replication-defective (rep) mutations (8,13). Two types of rep mutations were isolated that were located in the leader region of repZ. One type including rep57 disrupted an upstream open reading frame called repY which encodes a polypeptide of 29 amino acids. rep57 was an amber mutation of repY codon-11 (13) and reduced the level of repZ expression without affecting the amount of RepZ mRNA (8,13). Changing the amino acid composition of the reading frame by frameshift mutations did not affect repZ translation, whereas repZ translation was abolished profoundly when the position of the repY stop codon, located 7 bases downstream of the repZ start codon, was shifted either by frameshift mutations or changing the stop codon to a sense codon, indicating that the translational termination event of the repY reading frame, not the RepY polypeptide itself, is required for repZ translation (9). Since the level of repY expression was inversely correlated with the amount of endogenous Inc RNA (13) and the 5Ј-end of Inc RNA was located 3 bases upstream of the repY * This work was supported by a grant-in-aid from the Ministry of Education, Science, and Culture of Japan. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18  RBS (12), we concluded that inhibition of repY translation by Inc RNA is at least one mechanism by which Inc RNA represses repZ translation (13).
Surprisingly, however, the other type of rep mutations were found to disrupt an intramolecular base pairing within RepZ mRNA between the 5Ј-rGGCG-3Ј and 5Ј-rCGCC-3Ј sequences, distantly separated at positions 327-330 and 437-440, respectively (Ref. 8,see Figs. 1B and 2C). rep2006 (G327A) and rep2041 (G330A) changed the first and forth guanine residues of the former, whereas rep2044 (G438A) altered the second guanine residue of the latter (8). Besides, compensatory base changes to each of the rep mutations in the other sequence restored the ability to produce RepZ (8). This intramolecular base pairing had the potential to form a pseudoknot with a stem-loop structure designated I which was predicted as the target site of Inc RNA. On the other hand, another stem-loop structure designated III sequestered the 5Ј-rCGCC-3Ј sequence together with the repZ RBS. Based on these findings, we proposed a regulatory model (Fig. 1B), in which translation and termination of repY induces the intramolecular base pairing, leading to form a novel RNA pseudoknot for repZ translation. Inc RNA inhibits the formation of the pseudoknot, both directly and indirectly by inhibiting repY translation. Consistent with this model, disruption of structure III by two mutations (⌬G451 and C459A) resulted in partial derepression of repZ expression (9). In addition, Inc RNA inhibited repZ expression more efficiently than repY expression (8). However, biochemical evidence for the proposed pseudoknot had not been provided.
Several other fundamental questions have arisen during the course of studies on the control of repZ translation: does pseudoknot formation simply expose the repZ RBS to the ribosome, or also enhance repZ translation through specific interaction with the ribosome? And why does a single base substitution in the proposed 5Ј-rGGCG-3Ј/5Ј-rCGCC-3Ј duplex affect profoundly the intramolecular base pairing, when the complementary sequences consist of 7 bp or more (see Fig. 2C)? What is the efficiency with which repY translation stimulate the initiation of repZ translation? To answer some of these questions, we set out to characterize biochemically the pseudoknot. In this report, we demonstrate that RepZ mRNA synthesized in vitro formed structures I and III, but not the pseudoknot. However, disruption of structure III by means of base substitution/deletion resulted in the formation of a unique RNA pseudoknot in vitro. Evidence is also presented indicating that the pseudoknot formed in vitro is generated in vivo, affecting directly the level of repZ expression.
Measurement of ␤-Galactosidase Activity-The ␤-galactosidase activity expressed from translational lacZ fusions was assayed as described (8). The specific activity of the enzyme was expressed as Miller units (23). The values reported here are the results of at least three inde- In B, the unfolding of structure III by the ribosome stalling at the repY stop codon results in formation of a pseudoknot by base pairing between the 5Ј-rGGCG-3Ј and 5Ј-CGCC-3Ј sequences distantly separated, and allows the ribosome to access the repZ RBS. Binding of Inc RNA to the loop of structure I directly inhibits formation of the pseudoknot (c), and the subsequent Inc RNA-RepZ mRNA duplex formation inhibits repY translation (d). Sites of mutations affecting these regulatory mechanisms are indicated by arrows in A (see Introduction for details on these mutations). pendent experiments, and the standard deviations were within 15% of each value.
Preparation of RNA-RNA corresponding to the RepZ mRNA leader region was synthesized in vitro using T7 RNA polymerase. As the template we used SmaI-digested pKA10, pKA10-A25, and pKA10-A28 for wild-type RNA 293 , A25 RNA 293 , and A28 RNA 290 , respectively, EcoO109I-digested pKA10 for RNA 120 , and NdeI-digested pKA16 and pKA16-A10 for RNA 206 and rep2044 RNA 206 , respectively. SmaI cleaves pKA10 and its derivatives at the multiple cloning site of the vector, 3Ј to the ColIb-P9 insert, whereas EcoO109I cleaves in the middle of the ColIb-P9 portion, 3Ј to the structure I region (see Fig. 1). NdeI cleaves pKA16 or its rep2044 derivative at the NdeI site introduced in the middle of the structure III region of ColIb-P9 (Table I).
For preparation of the 5Ј-end-labeled products, the reaction mixture of 50 l contained 40 mM Tris-HCl (pH 8.0), 6 mM MgCl 2 , 5 mM dithiothreitol, 2 mM spermidine, 2 mM each of ATP, CTP, and UTP, 0.5 mM GTP, 4.625 MBq of [␥-32 P]GTP (222 TBq/mmol), 40 units of ribonuclease inhibitor (Takara Shuzo, Kyoto), 10 g of T7 RNA polymerase (obtained from Shigeyuki Yokoyama, University of Tokyo), and 1 g of template DNA. The reaction mixtures were incubated for 60 min at 37°C, extracted with phenol/chloroform, and precipitated with ethanol, followed by gel purification (24). The samples were suspended in the standard binding buffer described by Tomizawa (25). All the radioactive materials were purchased from New England Nuclear Research Products.
RNase Digestion-The 5Ј-end-labeled RNA (0.1 g/reaction) was digested with serially diluted RNase T1 (Amersham Pharmacia Biotech), RNase derived from Bacillus cereus, RNase Bc (Amersham Pharmacia Biotech), or RNase V1 (Amersham Pharmacia Biotech) in 6 l of the standard binding buffer in the presence of 0.5 g of yeast tRNAs. After incubation for 10 min at 37°C, the reaction was stopped by adding 4 l of stop solution (95% formamide, 0.1% xylene cyanol, and 0.1% bromphenol blue) and placed on ice. 3 l of the sample was resolved by electrophoresis on a 8.3 M urea, 8% polyacrylamide gel.

Secondary Structures of RepZ mRNA Leader Region-To
characterize biochemically secondary structures in the RepZ mRNA leader region, we prepared in vitro RNA 293 labeled with 32 P at the 5Ј-end as described under "Experimental Procedures." This RNA corresponds to the wild-type ColIb-P9 sequence from positions 244 to 524 that covered the entire region containing structures I to III. After the RNA sample was partially digested by RNases T1 (specific for guanine residues in single-stranded regions), RNase Bc (specific for pyrimidines in single-stranded regions), or RNase V1 (specific for doublestranded regions), the resulting products were analyzed by electrophoresis on a denaturing polyacrylamide gel. The patterns of the autoradiography are shown in Fig. 2, A and B, and the cleavage sites identified are summarized in Fig. 2C, in which the deduced secondary structures are also presented. We observed three stem-loop structures, I, II, and III as predicted previously (13). In addition, we found three new stem-loop structures designated Ia, IIa, and IV, which could not be predicted computationally. Note that RNase Bc appeared to cleave any ribonucleotides in the 5Ј side of loop sequences in the assay condition employed in this study.
When we looked at structure I and III regions which contained the two complementary sequences, 5Ј-rGGCG-3Ј (position 327-330) and 5Ј-rCGCC-3Ј (position 437-440), required for the intramolecular base pairing, two characteristic cleavage patterns were observed. First, both G-327 and G-328 residues in structure I were strongly cleaved by RNase T1 while both G-330 and G-331 residues were hardly cleaved (Fig. 2B, lane 1). Additionally, C-324, U-325, and U-326 were weakly cleaved by RNase Bc. We inferred from these patterns that the 5Ј-rUGGC-3Ј (position 326 -329) comprised the loop of structure I while C-324 and U-325 base paired weakly with G-331 and G-330, respectively, as shown in Fig. 2C. Second, all 9 guanine residues except G-447 in the sequence from 436 to 462 were not cleaved by RNase T1, and C-440 was strongly cleaved by RNase V1 (Fig. 2B, lanes 1 and 3). Since G-445, G-455, and G-457 consist of parts of the repZ RBS (UAAG at position 442-447 as the Shine-Dalgarno sequence and GUG at position 455-457 as the repZ initiation codon) (9,11), these results indicate that the translational initiation signals of repZ are embedded within structure III (Fig. 2C). Likewise, the data showed that the 5Ј-rCGCC-3Ј sequence is folded in structure III as well. Thus, these data, taken together, revealed that the intramolecular base pairing leading to the formation of a possible pseudoknot did not take place in vitro in the wild-type RNA 293 . This was further supported by the observation that the RNase cleavage patterns of structure I region in RNA 120 , a deletion of the wild-type RNA 293 beyond position 353, were identical to those in RNA 293 (Fig. 2, A and B).
In Vitro Formation of a Pseudoknot Structure-On the basis of the above results, we disrupted structure III by substituting 8 bases of its 3Ј side sequence, 5Ј-rGCUUGUGGCAGG-3Ј, to adenine residues without changing the repZ RBS (the substituted bases are underlined in the sequence). The resulting multiple mutation was designated as A25. Secondary structures of A25 RNA 293 synthesized in vitro were analyzed in conjunction with mutation rep2041 (G330A) and rep2044 (G438A) defective in the intramolecular base pairing (8). When compared with the wild-type RNA 293 , RNase T1 sensitivities at G-327 and G-328 in A25 RNA 293 were substantially reduced, although the cleavage pattern of the sequence corresponding to the 3Ј side of the structure I stem was not changed (A25 in Fig.  3A). Furthermore, the RNase V1 sensitivity at C-440 disappeared, indicating that structure III of A25 RNA 293 was indeed disrupted by the base changes. On the other hand, when mutation rep2041 or rep2044 was introduced into A25 RNA 293 , the degree of the RNase sensitivities of G-327 and G-328 was restored to that observed in the wild-type RNA 293 (A26 and A32 in Fig. 3A). Since C-329 and G-330 had been shown to interact with G-438 and C-437 by base pairing, respectively (8), these results can be explained by suggesting that, in A25 RNA 293 , at least 5Ј-rGGCG-3Ј in the loop of structure I base pairs with its complementary sequence 5Ј-rCGCC-3Ј located in the unfolded structure III region to form an RNA pseudoknot as shown in Fig. 3C (panel A25).
However, we also observed in A25 RNA 293 that G-438 and G-433 were weakly cleaved by RNase T1 and V1, respectively, whereas G-447 became insensitive to RNases T1 and Bc (A25 in Fig. 3B, lanes 1-3). In addition, the degree of cleavage of G-438 and G-433 as observed in A25 was enhanced in both A26 and A32 RNAs 293 except that in the latter the RNase T1 sensitivity of G-438 was lost due to the base change by mutation rep2044 (Fig. 3B). These results can be explained by proposing that A25 RNA 293 contains a mixture of at least two types of base pairing. One type occurs between the 5Ј-rGGCG-3Ј and 5Ј-rCGCC-3Ј to form the pseudoknot, as mentioned above, and the other occurs between 5Ј-rUCGCU-3Ј (positions 431-435) and 5Ј-rAGCGA-3Ј (position 444 -448) to produce a new stem-loop structure designated IIIa, as shown in Fig. 3C. We do not know if these types of base pairings coexist in a single RepZ mRNA molecule, or occur separately in two groups of messages. By contrast, both A26 and A32 RNA 293 formed only structure IIIa.
To separate the pseudoknot from structure IIIa, we constructed mutant A28 from A25. This mutant deleted three bases, 5Ј-rUCG-3Ј (positions 433-435) at the 5Ј side of the structure IIIa stem of A25 without affecting the repZ RBS. When cleavage patterns of A28 RNA 290 were compared with those of A25 RNA 293 , the RNase T1 sensitivities at G-327 and G-328 were reduced as observed in A25 RNA 293 (A28 in Fig.  3A). By contrast, G-445 and G-447, and U-442 were strongly cleaved by RNases T1 and Bc, respectively (A28 in Fig. 3B,  lanes 1 and 2), an indication that structure IIIa was disrupted by the deletion. When mutation rep2041 was introduced into A28, G-327 and G-328 residues were strongly cleaved by RNase T1 while G-445 and G-447 were not (A29 in Fig. 3, A and B). Similarly, when mutation rep2044 was introduced into A28, both G-327 and G-328 were cleaved by RNase T1. In addition, G-445 and G-447 were also cleaved by RNase T1 as observed in A28 RNA 290 (A34 in Fig. 3, A and B). We inferred from these results that A28 RNA 290 predominantly formed the pseudoknot that was observed in A25 RNA 293 (Fig. 2C). Conversely, A29 RNA 290 does not form the pseudoknot, and instead, generates a small stem-loop structure designated IIIb (A29 in Fig. 3C). However, A34 RNA 290 produced neither the pseudoknot nor structure IIIb. It should be noted that both structures IIIa and IIIb cover partially the Shine-Dalgarno sequence for repZ translation.
A RNA polymerase and SmaI-or EcoO109I-digested pKA10 as template, respectively, labeled with 32 P at the 5Ј ends, and purified as described under "Experimental Procedures." These two RNA species (lanes 1-3 for RNA 293 and lanes 4 -6 for RNA 120 ) were partially cleaved with RNase T1 (lanes 1 and 4), RNase Bc (lanes 2 and 5), and RNase V1 (lanes 3 and 6), and the cleavage products were separated by 8% polyacrylamide gel electrophoresis in the presence of 8.3 M urea. In panel B, electrophoresis was carried out longer for the same samples shown in panel A. The lengths of cleavage products were measured by comparison with the RNase cleavage products of a sample of RNA 293 or RNA 120 , cleaved at every nucleotide with NaOH, and compared with the positions in the ColIb-P9 sequence. Positions of some of the cleavage products are resolved to the left in the ColIb-P9 coordinate, numbered as in Fig. 1. Those mentioned in the text are in bold and highlighted with arrows. Asterisks indicate undigested products. Locations of stemloop structures are shown by vertical bars to the right of the relevant cleavage products. C, deduced secondary structures of RNA 293 . Nucleotide sequence of RNA 293 is shown with the deduced secondary structures, named Ia to IV. Uppercase letters denote ribonucleotides derived from the ColIb-P9 portion, whereas lowercase letters denote those derived from the vector (pTZ19R) portion. Nucleotides are numbered as in Fig. 1. RNase T1, Bc, or V1 cleavage sites were indicated by filled, gray, or empty triangles, respectively. Nucleotide sequences that are complementary to each other and have a potential to form a pseudoknot with structure I are boxed. The initiation codons for repY and repZ are underlined. The termination codon of repY is double-underlined. The Shine-Dalgarno sequence for repZ translation is shown by asterisks. 3Ј-ends of shorter RNA species, RNA 120 and RNA 206 , are shown by arrows. The 5Ј-ends of these RNAs are identical to that of RNA 293 . lacked the sequence responsible for structure III, IIIa, or IIIb formation. When RNA 206 was digested by RNases, the cleavage pattern of the structure I region was found to be quite different from that of the wild-type RNA 293 . (i) Both G-327 and G-328 were not cleaved by RNase T1; (ii) instead, G-321 was strongly cleaved by RNase T1; and (iii) RNase Bc sensitivity at U-317 was lost (Fig. 4A, lanes 1-3). However, the cleavage patterns of the sequence corresponding to the stem region in structure I were identical to those of RNA 293 (Fig. 4A). Surprisingly, when rep2044 was introduced into RNA 206 , the cleavage patterns of structure I region became identical to those of RNA 293 (Fig. 4A, lanes 4 -6, and also see, Fig. 2B, lanes 1-3). These results indicate strongly that the 10 complementary bases in the top region of structure I had the potential to pair with those in the downstream sequence, generating a unique RNA pseudoknot as shown in Fig. 4B. Of further significance was the finding that this seemingly strong base pairing was abolished by introduction of a single mismatch mutation, rep2044. This phenomenon will be studied later in relation to the role of each base pair in repZ expression.
Effect of the Pseudoknot on repZ Expression-To assess the pseudoknot formed in vitro, we examined effects of mutations A25 and A28 on the level of repZ expression in the absence of inc and repY genes, the negative and positive regulatory elements, respectively, as the action of either or both of them might cause conformational changes in the RepZ mRNA leader region (see Fig. 3C for effects of A25 and A28 mutations on RepZ mRNA secondary structure). The RepZ activity was monitored by measuring the ␤-galactosidase activity of the repZ-lacZ translational fusions, in which codon 221 of the repZ gene was fused in-frame to codon 7 of the promoter-less lacZ gene as the reporter. The vector for this fusion gene was a mini-F plasmid, whose copy number was almost the same as that of ColIb-P9 (data not shown). A summary of the experiments is given in Table II. pKA340-W3 carrying inc1, a promoter-down mutation in the inc gene, showed 1,912 units of RepZ activity. When rep57 (C408T), an amber mutation of repY codon 11 (13), was introduced into pKA340-W3, the resultant plasmid pKA340-A33 exhibited only 2.6 units of the RepZ activity, a value 735-fold (1912/2.6) lower than that of pKA340-W3. In the case of pKA340-A25 carrying the A25 mutation, the level of RepZ activity was 223 units, 86-fold higher than that of pKA340-A33. This level was further elevated to 3,714 units in pKA340-A28 bearing the A28 mutation. The value of the RepZ activity in pKA340-A28 was twice as high as that of pKA340-W3. There results indicate that disruption of structures III and IIIa leads to a higher level of repZ expression.
On the other hand, introduction of mutation rep2041 or rep2044 into pKA340-A25 caused a considerable reduction of repZ expression, yielding a value of 14 or 16 units, respectively. Similarly, the presence of rep2041 in pKA340-A28 decreased the RepZ activity to 265 units. By contrast, when rep2044 was introduced into pKA340-A28, the resulting plasmid pKA340-A34 exhibited almost the same RepZ activity as that of pKA340-W3. Furthermore, we found that this activity was not affected by the action of either Inc RNA or repY translation (pKA340-A34ϩ or -A50 in Table II), an indication that the repZ gene was constitutively expressed.
Of importance was the observation that although mutations rep2041 and rep2044 reduced substantially the level of repZ expression in plasmids pKA340-A26 and -A32, their activities were significantly higher than that of pKA340-A33. In addition, the level of repZ expression in pKA340-A29 was about 18 times higher than those of pKA340-A26 and -A32. These facts and the constitutive expression of repZ in pKA340-A34 suggest strongly that the RepZ activities of these plasmids are inversely correlated with the calculated thermodynamic stabilities of structures III (Ϫ11.3 kcal/mol), IIIa (Ϫ5.3 kcal/mol), and

FIG. 2-continued
IIIb (Ϫ1.0 kcal/mol). If this were the case, it could be concluded that these secondary structures observed in vitro are indeed formed in vivo, acting as the negative elements for repZ expression due to sequestering of its Shine-Dalgarno sequence. By the same token, we would conclude that the pseudoknot observed in vitro in A28 RNA 290 is also formed in vivo, being responsible for the highest level of repZ expression in pKA340-A28. Furthermore, the data indicate that the role of repY in repZ expression is to disrupt structure III during the process of its translation and termination, inducing the intramolecular base pairing between the two complementary sequences to form the pseudoknot. Thus, the pseudoknot formed by the process of repY translation acts as the molecular switch for translation initiation of repZ expression, as proposed previously as shown in Fig. 1B (8, 9).
Bases Involved in Pseudoknot Formation in Vivo-In view of the finding that all the base pairs in the pseudoknot of RNA 206 were abolished by a single mismatch mutation, rep2044, we considered the role of each base pair in the pseudoknot formation. To address this problem and to characterize further the pseudoknot formed in vivo, the complementary sequences of pKA340-W3 were changed one by one and examined for repZ expression in the absence of Inc RNA. The results of the experiments are given in Table III. All the A:C mismatches in base pairings between 5Ј-rUGGCGG-3Ј (326 -331) and 5Ј-rCCG-CCA-3Ј (436 -441) reduced repZ expression to various extents, and the most severe reduction was observed at the four middle contiguous base pairs, 5Ј-rGGCG-3Ј/5Ј-rCGCC-3Ј, consistent with previous observations (8). On the other hand, mismatches of U:U or A:A at positions 332 and 435 and of A:G at positions 334 and 433 did not cause a significant reduction in repZ expression. Similarly, we observed that substitution of U-325 to C located outside the complementary sequence also did not affect the level of repZ expression. These results, taken together, indicated that 6 specific contiguous base pairings between 5Ј-rUGGCGG-3Ј (positions 326 -331) and 5Ј-rCCGC-CA-3Ј (positions 436 -441) were sufficient for the pseudoknot to be formed in vivo.
When the base pairings between the sequences at positions 327-331 and 436 -440 were restored by a compensatory base change, the level of repZ expression was restored to some extent. For example, the replacement of an A:C mismatch with a G:U pair resulted in a modest, but significant, increase in repZ expression. Furthermore, restoring to a canonical A:U pair increased repZ expression to a greater extent, yielding values of 38 to 93% of that of pKA340-W3. These results implied that maintaining the base pairings required for pseudoknot formation was more important than the base sequences themselves.
However, the base pairing between U-326 and A-441 seemed to have different requirements than the others we examined. Mutation of this base pair to C:A or G:A reduced repZ expression by a factor of 8.5 or 7.1, respectively, whereas the substitution to U:C decreased expression only by a factor of 3.2 (Table  III). In addition, we observed that the reduced repZ expression associated with the G:A mismatch was restored only 2.5-fold by replacement with the canonical G:C pair, while it was recovered to the level of pKA340-W3 by either the U:G or C:G pair. These results indicate that pyrimidine residues are required at position 326 for pseudoknot formation, probably participating in the formation of a unique loop conformation in structure I FIG. 3. Secondary structures of mutant RNA 293 derivatives with structure III disrupted. A and B, 5Ј-end-labeled RNA 293 or its mutant derivatives were synthesized, and cleaved with RNase T1, Bc, or V1, and subjected to electrophoresis as in Fig. 2. Mutant RNAs were generated from pKA10-A25 (inc1 rep57 A25) (labeled A25), pKA10-A26 (inc1 rep57 A25 rep2041) (labeled A26), pKA10-A32 (inc1 rep57 A25 rep2044) (labeled A32), pKA10-A28 (inc1 rep57 A25 A28) (labeled A28), pKA10-A29 (inc1 rep57 A25 A28 rep2041) (labeled A29), and pKA10-A34 (inc1 rep57 A25 A28 rep2044) (labeled A34). inc1 and rep57 mutations do not change the cleavage patterns of RNA 293 (data not shown). A part of the RNase T1 cleavage pattern corresponding to the structure I region is shown in panel A, whereas parts of RNase T1, Bc, and V1 cleavage patterns corresponding to the structure III region are shown in panel B (lanes 1-3, respectively). Representation schemes of the RNase cleavage products are the same as in Fig. 2, A and B, except that,  Requirement for repY Translation in the Induction of repZ Expression-We have previously shown that the translation and termination of the repY reading frame is essential for repZ translation, and that Inc RNA regulates the translation of both repY and repZ at different rates (8). In this paper, the role of repY translation in repZ expression was shown to involve disruption of structure III, inducing pseudoknot formation (Table  II). To investigate the effect of repY translation on repZ expression, we compared ␤-galactosidase activities between pKA340-W3 and pKA340-A52, carrying repZ-lacZ and repY-repZ-lacZ fusions in mini-F derived plasmids, respectively, where in both cases the lacZ gene was connected to the same The repY-repZ-lacZ fusion in pKA340-A52 was constructed by inserting a cytosine residue between positions 403 and 404 and copy number of this plasmid was the same as that of pKA340-W3. Fig. 5 shows the results of the experiments. pKA340-A52 exhibited 2151 units, a value almost identical to that of pKA340-W3. In addition, we had previously shown that no protein synthesis was initiated from the normal repZ initiation codon in the repY-repZ-lacZ fusions because of the lack of the termination event near the repZ initiation codon (8,9,13). Assuming that the ␤-galactosidase activity of the repY-repZ-lacZ fusion protein was not affected by the extra 25 amino acids derived from the repY frame, these results indicated that the ratio of repY translation to that of repZ translation was nearly one, implying that ribosome access to the repZ RBS and the subsequent steps in initiation of repZ translation are very efficiently coupled to repY translation and termination. Finally, we examined whether Inc RNA controls directly translation initiation of the repZ gene. When the inc1 mutation in both pKA340-W3 and pKA340-A52 was converted to the wild-type for restoring the synthesis of Inc RNA, both repY and repZ expressions decreased considerably, and the ratio of repY translation to repZ translation increased to 5 (pKA340-A63 and pKA340 in Fig. 5), consistent with previous results (8). Furthermore, when the copy number of the lacZ fusion genes was increased by re-cloning into the multicopy vector plasmid, pMB1, the repY expression increased in proportion to the gene dosage, whereas the repZ expression did not increase significantly (Fig. 5). These observations indicate that Inc RNA controls repZ translation not only indirectly by inhibition of repY translation, but also directly by uncoupling it from repY translation. In this way, a constant level of repZ expression is maintained irrespective of changes in plasmid copy number. DISCUSSION In this report, we have presented evidence that the novel RNA pseudoknot formed in the RepZ mRNA leader region is involved in translation initiation of the repZ gene, encoding the replication initiator protein of the ColIb-P9 plasmid. Structural analyses of the RepZ mRNA leader sequence synthesized in vitro revealed that it contained several stem-loop structures including structures I and III, but not the pseudoknot (Fig. 2). However, disruption of structure III, without changing the repZ RBS, by means of base substitution and deletion was found to induce new intramolecular base pairing between two distantly separated short complementary sequences located at positions 327-330 and 437-440. These interactions result in formation of a pseudoknot involving structure I (Fig. 3). We have also presented evidence suggesting that the pseudoknot observed in vitro is formed in vivo, affecting significantly the level of repZ expression (Table II). These results strengthen the previously proposed model that the pseudoknot induced by translation and termination of the repY reading frame functions as the molecular switch for translational initiation of the repZ gene.
Is the disruption of structure III sufficient for pseudoknot FIG. 4. RepZ mRNA pseudoknot as observed in RNA 206 . A, 5Јend-labeled RNA 206 and its rep2044 derivative were subjected to RNase digestion and electrophoresis as described in the legend to Fig. 2. Representation schemes of the RNase cleavage products are the same as in Fig. 2, A and B. Nucleotides at which RNase sensitivities are altered by rep2044 are highlighted by arrows. B, secondary structures of RNA 206 beyond A-300 are depicted using the same symbols as in Fig.  2C. Ps indicate the pseudoknot interaction.

TABLE II
Effect of disruption of structure III on repZ expression MC1061 (lacX74) cells carrying the indicated plasmid were grown to A 600 ϭ 0.4 and assayed for ␤-galactosidase activity as described under "Experimental Procedures." inc promoter-down inc1 mutation was employed for removing Inc RNA transcribed from each fusion plasmid. An amber mutation of repY codon-11, rep57, was used to remove (any) conformational change induced by the ribosome synthesizing RepY polypeptide.

Plasmid
Mutation repY Inc RNA Secondary structure a RepZ activity b  III Effect of pseudoknot mutations on repZ expression MC1061(lacX74) cells carrying the indicated plasmids were assayed for ␤-galactosidase activity as described in Table II. Inc-promoter down-mutation inc1 was employed to measure RepZ activity in the absence of Inc RNA. stem-loop for antisense RNA binding can also be predicted, and evidence for one in pMU720 is reported (10). However, no other stem-loop analogous to structure II of ColIb-P9 is predicted in pMU2200, and in fact, a 30-base long DNA segment encompassing structure II was able to be deleted without affecting the pseudoknot formation in ColIb-P9. 2 In addition, we previously constructed an IncFII R100:ColIb-P9 chimeric mini-plasmid where the ColIb-P9 repZ leader sequence upstream of 5Ј-rCGCC-3Ј was replaced with the corresponding region of a distantly related R100 plasmid containing its entire inc gene plus the initiation codon of repA6 in-frame to the truncated ColIb-P9 repY frame (22). The leader regions of rep genes of IncFII-type plasmids encode an antisense RNA and upstream open reading frame such as inc and repA6 in R100, respectively, but the rep genes of these plasmids do not require formation of a pseudoknot for coupling translation of upstream open reading frame to rep (26). Although the resultant chimeric plasmid did not produce repZ due to lack of the pseudoknot structure, the replacement of the Inc RNA-target loop of the R100 portion with that of ColIb-P9, including the 5Ј-rGGCG-3Ј sequence, was sufficient for repZ in the ColIb-P9 portion to be translated (22). These lines of evidence together suggest that only structure I and the exposed 5Ј-rCGCC-3Ј sequence are sufficient for the pseudoknot formation, and hence repZ expression. If so, how does the loop of structure I base pair specifically with a sequence preceding the repZ RBS to form a pseudoknot? And why does a single mismatch mutation such as rep2044 disrupt it profoundly, although possible base pairing occurs between 11 bases of complementary RNA sequences (Fig. 4)? We believe that a transient base paring between a subset of these sequences, 5Ј-rGGCG-3Ј and 5Ј-rCGCC-3Ј, is rate-limiting for the formation of the pseudoknot. Thus, interfering with this transient step by a single mismatch can be sufficient for preventing all subsequent steps in pseudoknot formation. In addition, a certain conformation of the loop region of structure I appears to play a role in stimulating the intramolecular base pairing by giving it higher affinity and specificity, as suggested here by the preference for a pyrimidine residue at a base 5Ј to the 5Ј-rGGCG-3Ј sequence for repZ expression (Table III). The same conformation in the loop of structure I may stimulate RepZ mRNA-Inc RNA interaction as well. Detailed analyses of the structural basis of the intermolecular interaction in the accompanying paper (30), as well as the analyses of the pseudoknot formation in vitro with different RNA species that inhibit binding of Inc RNA to the RepZ mRNA leader 3 support this hypothesis. 2 K. Asano and K. Mizobuchi, unpublished observations. 3 K. Asano and K. Mizobuchi, manuscript in preparation.
FIG. 5. Ratio of translational coupling between repY and repZ. repZ-lacZ translational fusion plasmids are depicted with the same scheme as in Fig. 1. Mutation sites introduced into each construct are shown by upward arrows. inc1 removes Inc RNA transcription, whereas ⍀C403 fuses repY to repZ in-frame. Gray boxes denote 5Ј-truncated lacZ reading frame. The replication origin of each plasmid, which determines its copy number in host cells, and Miller's units of ␤-galactosidase as measured with E. coli MC1061 (lacX74) cells carrying each construct are given to the right. Copy numbers of F or pMB1 plasmid, given in parentheses as per host chromosome, are according to Refs. 28 and 29, respectively.