Structural Analysis of Late Intermediate Complex Formed between Plasmid ColIb-P9 Inc RNA and Its Target RNA

The antisense Inc RNA encoded by the IncIα ColIb-P9 plasmid replicon controls the translation of repZencoding the replication initiator and its leader peptiderepY at different rates with different mechanisms. The initial loop-loop base pairing between Inc RNA and the target in therepZ mRNA leader inhibits formation of a pseudoknot required for repZ translation. A subsequent base pairing at the 5′ leader of Inc RNA blocks repY translation. To delineate the molecular basis for the differential control, we analyzed the intermediate complexes formed between RepZ mRNA and Inc RNA54, a 5′-truncated Inc RNA derivative. We found that the initial base pairing at the loops transforms into a more stable intermediate complex by its propagation in both directions. The resulting extensive base pairing indicates that the inhibition of the pseudoknot formation is established at this stage. Furthermore, the region of extensive base pairing includes bases different in related plasmids showing different incompatibility. Thus, the observed extensive base pairing is important for determining the incompatibility of the low-copy-number plasmids. We discuss the evolution of replication control systems found in IncIα, IncB, and IncFII group plasmids.

The antisense Inc RNA encoded by the IncI␣ ColIb-P9 plasmid replicon controls the translation of repZ encoding the replication initiator and its leader peptide repY at different rates with different mechanisms. The initial loop-loop base pairing between Inc RNA and the target in the repZ mRNA leader inhibits formation of a pseudoknot required for repZ translation. A subsequent base pairing at the 5 leader of Inc RNA blocks repY translation. To delineate the molecular basis for the differential control, we analyzed the intermediate complexes formed between RepZ mRNA and Inc RNA 54 , a 5-truncated Inc RNA derivative. We found that the initial base pairing at the loops transforms into a more stable intermediate complex by its propagation in both directions. The resulting extensive base pairing indicates that the inhibition of the pseudoknot formation is established at this stage. Furthermore, the region of extensive base pairing includes bases different in related plasmids showing different incompatibility. Thus, the observed extensive base pairing is important for determining the incompatibility of the low-copy-number plasmids. We discuss the evolution of replication control systems found in IncI␣, IncB, and IncFII group plasmids.
Antisense RNAs bind to the complementary regions (or "sense" strands) of their target RNAs and exert a variety of regulatory functions (1,2). Some of them are encoded in the autonomous replication regions of plasmids and negatively control their replication and copy number (3). The copy number of the plasmid is proportional to the intracellular concentration of its antisense RNA, which, in turn, down-regulates the frequency of plasmid replication, thereby establishing a negative feedback loop for the maintenance of constant copy number. These antisense RNAs, when expressed in a bacterial cell, repress the replication of identical or closely related plasmids introduced by means of transfer or transformation. This phenomenon is called incompatibility (Inc) 1 and is used to classify plasmids: i.e. if two different plasmids are incompatible in a single bacterial cell, they belong to the same Inc group (4).
There are essentially two types of antisense RNAs that show the highest binding rate constants of ϳ10 6 /M/s (1). High copy number plasmids including pMB1 and ColE1 encode one type (designated here as class H for high copy) of such antisense RNAs. These antisense RNAs have the size of ϳ100 bases, form three stem-loops with a single-stranded 5Ј leader, and bind to the preprimer RNAs for the leading strand DNA synthesis, thereby inhibiting formation of mature primer RNAs (1). The other type of rapid binders (class L for low copy) are encoded by low-copy-number plasmids including the members of IncFII, IncI␣ (IncI 1 ) and IncB groups (1,2). These antisense RNAs are ϳ70 -100-base long, fold essentially into a single large stemloop of ϳ50 bases, and control the expression of the plasmid replication initiator (rep) genes at the translational level. Specifically, they block the translation of a rep leader peptide (RLP) coupled to the translation of rep (5)(6)(7)(8)(9), by binding to the vicinity of the ribosome-binding site (RBS) for the RLP (10 -12). In the case of IncI␣ ColIb-P9 and IncB pMU720 plasmids, the translation of RLP (repY and repB) induces formation of a pseudoknot between the target stem-loop and a sequence preceding the rep (repZ and repA, respectively) RBS, which is required for rep translation (13)(14)(15). The antisense RNA additionally inhibits the pseudoknot formation by binding to the loop of the target stem-loop structure (16). Because of the different mechanisms, the antisense RNA represses rep expression more efficiently than RLP expression, thereby establishing the constant level of Rep protein and hence the plasmid copy number (16). We have been studying this unique differential control exerted by the antisense Inc RNA of the ColIb-P9 plasmid.
It is established that the process of binding of antisense RNA I (A) to its target RNA II (T), encoded by ColE1 (class H), comprises a series of reactions producing more stable intermediates, where C s is the stable final product, C** is characterized by its altered RNase sensitivity and C* is the unstable complex, the formation of which is inferred from the inhibition kinetics. The C* complex appears to consist of multiple contacts at loops and 5Ј-single stranded leader, with the inhibitory dissociation constant (K i ) of 11 nM (17,18). In the C** complex, base pairing in each loop-loop interaction may stack on the stems of the antisense and target RNAs, which altogether provides the equilib-* This research was supported by a grant-in-aid from the Ministry of Education, Science, Sports and Culture (Monbu-sho) 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  rium dissociation constant (K d ) of 0.1 nM (18). Different incompatibilities among ColE1-type plasmids (class H) can be explained at the level of either C* or C** formation (17,18). The C* complex formed in class L was characterized only in the IncI␣ ColIb-P9 system: Formation of C* is mediated by base pairing at the 5Ј-rGGCGG-3Ј sequence in the target loop ( Fig. 1, step f), and shows a K i of 6 -10 nM (16). It is proposed that this initial interaction is stimulated by a specific loop structure conferred by the target loop sequence, 5Ј-rUUGGCG-3Ј (Fig. 1, step e) (16,19), which is conserved in the target loop for class L antisense RNAs (20,21). Formation of C* can directly compete with the pseudoknot required for ColIb-P9 repZ translation, explaining how the antisense Inc RNA represses repZ translation at the level of loop-loop interaction (Fig. 1, steps b and f) (16).
No detectable incompatibility between IncI␣ and IncFII plasmids (22) can be explained at the level of C* formation, because one of the bases critical for C* formation in IncI␣ ColIb-P9 is altered in IncFII plasmids (16). Despite sharing the bases critical for C* formation, IncI␣ and IncB plasmids are compatible for several generations although they show weak incompatibility thereafter (22). Therefore, the reduced stability of the complex formed later than C*, most likely C**, must be responsible for the observed compatibility between IncI␣ and IncB plasmids.
In this report, we focus on the secondary structure of a late intermediate complex (C**) formed between Inc RNA and its target, encoded by IncI␣ ColIb-P9 (highlighted in Fig. 1, step g). We find that the initial transient interaction (C*) between RepZ mRNA and Inc RNA 54 , a derivative of Inc RNA lacking its 5Ј-single stranded leader, transforms into a more stable complex altering the sensitivity of the target RNA region to a variety of RNases. Our results show that the region of extensive base pairing encompasses the bases different from the IncB plasmids, consistent with the idea that C** is an important determinant of plasmid incompatibility. Furthermore, we find that the extensive intermolecluar base pairing is separated at the site of initial base pairing, suggesting a stacking trend between each intermolecular helix and the stem of Inc RNA or its target. This unusual structure at the late intermediate step provides insights into the molecular basis for the unique differential control by Inc RNA.
Binding of Inc RNA to RepZ mRNA in Vitro-Binding of Inc RNA to RepZ mRNA was conducted in the binding buffer (50 l) containing 10 mM MgCl 2 , 100 mM NaCl, and 20 mM Tris-HCl (pH 7.6) at 37°C and analyzed by denaturing polyacrylamide gel electrophoresis in the presence of 8.3 M urea (16,19). Inc RNA-Rep RNA hybrid was detected as a complex that was stable during the denaturing gel electrophoresis and migrated anomalously compared with free RNA species.
Secondary Structural Analyses on the Late Complexes between Rep RNA 293 and Inc RNA 54 or Ps 14 -40 nM 5Ј 32 P-Rep RNA 293 (14) was incubated with indicated amounts of Inc RNA 54 and Ps 14 in 5 l of binding buffer (19) containing 0.25 g of yeast tRNAs at 37°C for 30 min and digested partially with 1 l of appropriately diluted RNases at 37°C for 10 min. The reaction was terminated by adding 5 l of stop solution (19) on ice. 4 l of the sample was resolved by 8% polyacrylamide gel electrophoresis (8.3 M urea). Gel was dried for autoradiography or phosphorimaging analysis with the Fuji BAS2000 BioImage analyzer. For the measurement of K d , 0, 25, 50, 100, and 400 nM Inc RNA 54 or Ps 14 was employed for the cleavage experiments as above, and the amount of RNase T1 products cleaved at G-327 and G-328 was quantitated together with the cleavage product at G-360 as an internal standard. K d was calculated as [Inc RNA 54 or Ps 14 ] yϭ1/2 Ϫ 20 nM, where y is the relative intensity of the G-327 and G-328 bands.

Inc RNA 54 Binds to Stem-Loop I in Two
Steps-We previously synthesized three RNA species, Inc RNA 54 , Ps 14 , and SI 54 (see under "Experimental Procedures") and examined their effects on the interaction between Inc RNA and Rep RNA 293 , a 293-base RNA corresponding to the repZ mRNA leader (16). Inc RNA 54 is an Inc RNA derivative lacking the 5Ј single-stranded leader and did not form a stable complex with Rep RNA 293 under the denaturing gel conditions. Ps 14 is a single-stranded 14-base portion of RepZ mRNA with a 7-base sequence complementary to loop I, and its binding to the target stem-loop (designated structure I) mimics the repZ mRNA pseudoknot. SI 54 is a 54-base RNA corresponding to the stem-loop I of RepZ mRNA. Fig. 2 shows a reproduced result of the kinetic analyses on the inhibition of Inc RNA binding to Rep RNA 293 by these three inhibitory RNAs. As shown by the filled squares in Fig. 2, Step g of Inc RNA pairing, the major focus of this study, is highlighted by a dotted box.
A and B, 32 P labeled Rep RNA 293 or Inc RNA reduced the fraction of unhybridized species, due to its hybridization to excess unlabeled Inc RNA or Rep RNA 293 , respectively. As shown by open symbols in Fig. 2, A and B, the presence of inhibitory RNAs decreased the initial rate of the hybridization with K i values of 6 -10 nM, as reported previously (16). Thus, the three RNAs formed transient, reversible complexes with the partners, and competed with the rate-limiting step of the hybridization reaction (complex C*; see the second steps in Fig.  2, C and E).
During the course of these analyses, we realized that the binding between Inc RNA and Rep RNA 293 was significantly retarded within 2 or 3 min in the presence of Inc RNA 54 and SI 54 , carrying a large stem-loop structure (Fig. 2, A and B). Because Inc RNA 54 and SI 54 are present in large excess over 32 P-Rep RNA 293 and 32 P-Inc RNA, respectively, we hypothesized that this secondary inhibition is due to their rapid, irreversible formation of a different complex with the complementary region in the 32 P-Rep RNA 293 or 32 P-Inc RNA. We presumed that such a complex did not withstand the denaturing gel electrophoresis for the detection of Inc RNA-Rep RNA hybrid (see the third steps in Fig. 2, C and E). In contrast, the secondary inhibition was not observed with the single-stranded Ps 14 RNA in this time frame (Fig. 2A). However, the preincubation (Ͼ15 min) of excess Ps 14 (400 nM) with 32 P-Rep RNA 293 completely blocked Inc RNA binding (data not shown). Thus, the initial base pairing with Ps 14 may transform into an irreversible complex more slowly than the complex with Inc RNA 54 (Fig. 2D).
Secondary Structure Analyses of the Late Complex between RepZ mRNA and Ps 14 -To examine the possibility that the initial base pairing (C*) between Inc RNA and its target transforms into an irreversible intermediate interaction corresponding to C**, we analyzed the secondary structure of 5Ј-32 P labeled Rep RNA 293 using RNases V1 (specific to doublestranded RNA), T1, U2, PhyM, and CL3 (specific to an unpaired G, A, U/A, or C residue, respectively), after the preincubation (30 min) of 32 P-Rep RNA 293 with Inc RNA 54 or Ps 14 . Fig. 3 shows the gel electrophoretic patterns of the cleaved products of Rep RNA 293 . Fig. 4 describes the deduced secondary structure of the target stem-loop I in RepZ mRNA. Cleavage experiments with 32 P-Rep RNA 293 alone (Fig. 3, lanes 1, 7, 13, 16, and 20) confirm the previously proposed structure of stemloop I (14,19), as shown in Fig. 4A. G-327 and G-328 at the top of the loop were strongly cleaved by RNase T1. These two bases are most critical for both the pseudoknot formation (13,14) and Inc RNA binding (19).
In the presence of Ps 14 , the structure of the top of stem-loop I was altered substantially. The RNase T1 cleavage at G-327 and G-328 was lost with excess Ps 14 (Fig. 3, lane 4; Fig. 4C). Alternatively, the enzyme cleaved G-321 strongly. In addition, U-318, A-319, A-322 were cleaved by RNase PhyM, and A-323 and C-320 were cleaved by RNase CL3. However, the cleavage patterns of the stem I region were not changed. These results indicate that Ps 14 and Rep RNA 293 form a complex by base pairing between their complementary sequences. This complex is very similar in structure to RNA 206 , a derivative of Rep RNA 293 truncated 3Ј-terminally at the same residue as Ps 14 (14). RNA 206 was unable to bind Inc RNA (19), consistent with our finding that the preincubation of Rep RNA 293 with Ps 14 blocked its binding to full-length Inc RNA.
The results shown in Fig. 3, lanes 2-4, indicate that the loss of RNase T1 cleavage at G-327 and G-328 (located at the top of structure I) is a good indicator of formation of the late complex with Ps 14 . In order to measure the equilibrium dissociation constant (K d ) for this Ps 14  The reaction was terminated at indicated times and analyzed on a 16% polyacrylamide gel as described (16). B, the binding between 2.5 nM Rep RNA 293 and 0.25 nM 32 P-Inc RNA was challenged with 10 nM target RNA derivative, SI 54 , as described (16). C-E, schematic representation of inhibitory complexes formed with Inc RNA 54 (C), Ps 14 (D), or SI 54 (E). Inhibitory or equilibrium dissociation constants for each complex are shown based on the previous report (16) or this study. and conducted partial cleavage with RNase T1. The amount of cleavage products at G-327 and G-328 was quantitated relative to that of cleavage product at G-360 as an internal standard (see Fig. 3). As shown in Fig. 5, the cleavage signal at G-327 and G-328 was reduced to 50% when [Ps 14 ] is 80 nM, giving the K d value of 60 nM. Thus, the formation of the extended pseudoknot is probably a slow process. These results support the idea that the pseudoknot induced in vivo by repY translation is unfolded by the inhibitory secondary structure (III) before it reaches the more extended pseudoknot, as illustrated in Fig. 1, steps Secondary Structure Analyses of the Late Complex between RepZ mRNA and Inc RNA 54 -In the presence of Inc RNA 54 , the structure of stem-loop I was altered more dramatically (Fig.  4B). The RNase T1 cleavage at G-327 and G-328 was significantly reduced by increasing the amount of Inc RNA 54 (Fig. 3,  lanes 5 and 6). However, the RNase T1 cleavage at these residues remained even with excesses Inc RNA 54 , whereas it was lost completely with excess Ps 14 (Fig. 3, lanes 4 and 6). Consistent with this observation, only RNase PhyM weakly cleaved at U-326 adjacent to guanine residues, among other single-strand specific RNases (Fig. 3, lane 21). The RNase V1 cleavage at C-312ϳU-314 completely disappeared and instead shifted to loop bases at G-321 and A-322 (Fig. 3, lane 11). These results suggest that Inc RNA 54 and Rep RNA 293 form a novel X-shaped complex with two intra-and intermolecular helices, as shown in Fig. 4B. It is conceivable that the initial base pairing propagates in both directions until it meets with the stems of Inc RNA and the target and, thereafter, the long intermolecular helix was separated at the center of the loops. This separation would allow the intramolecular stems to stand side by side, being stimulated by a stacking trend between each of inter-and intramolecular helices. Furthermore, the loss of RNase V1 sensitivity at C-312ϳU-314 strongly suggests that the region of extended base pairing covers the 21-base Inc loop region plus a few bases each from the upper stem and includes the bases different from IncB pMU720 plasmid (Fig. 4B). Because the bases in the 5Ј single-stranded leader of the antisense RNAs are identical between IncI␣ and IncB plasmids, we propose that the stability of the complex as depicted in Fig. 4B is responsible for determining different incompatibility between IncI␣ and IncB plasmids (see under "Discussion").
Next we attempted to estimate the K d value for this late Inc RNA 54 complex by plotting the relative intensity of the RNase T1 cleavage products at G-327 and G-328 against [Inc RNA 54 ].  5 shows that K d for the Inc RNA 54 complex is much smaller than that for the Ps 14 complex and is close to the detection limit for this experiment. Nevertheless, we could tentatively estimate the K d value of ϳ3 nM. Although determining the accurate K d value awaits further experiments, the results in Fig. 5 suggest that the late Inc RNA 54 complex is more stable than the transient initial interaction (C*) and corresponds to the C** complex. Thus, the whole structural data support our model that the initial base-paring between the stem-loops of Inc RNA 54 and the target in Rep RNA 293 (C*) rapidly transforms into a more stable, irreversible complex (C**), leading to block full-length Inc RNA binding, as observed in Fig. 2A.
Why Is the Late Inc RNA 54 Complex Unstable?-The intermolecular base pairing observed in the late Inc RNA 54 complex was quite extensive as shown in Fig. 4B. Nevertheless, it was not stable in denaturing gels, probably because the interacting loops are closed by stable stems: annealing of two strands of RNA generates torsion as they form a helical structure. Fixing the ends of these strands by a stable stem would remove the way to release such torsion, thereby providing a stress into the structure of the resulting complex. To test this idea briefly, we prepared Inc 21 , a 21-base RNA corresponding to the loop of Inc RNA, and allowed it to bind to Rep RNA 293 . Inc 21 does not form any secondary structure (data not shown) and therefore would wind up to loop I without storing a stress. As expected, this RNA rapidly formed a stable complex, persistent in denaturing gels (Fig. 6). Because the Inc RNA 54 -Rep RNA 293 complex was separated in the same denaturing gel conditions (16), we conclude that having a loop closed by a stable stem indeed destabilized the loop-loop complex. The stable binding of Inc 21 to RepZ mRNA also supports our previous proposal that there is little requirement in the structure of Inc RNA for its efficient binding to RepZ mRNA (16,19).

Molecular Basis for Differential Control of repY and repZ by
Inc RNA-The plasmid ColIb-P9 Inc RNA controls translation of repY and repZ with different mechanisms (12). By using Inc RNA 54 lacking the 5Ј leader of Inc RNA and Ps 14 , a part of RepZ mRNA complementary to loop I, we previously showed that the initial loop-loop base pairing between Inc RNA and the target in the repZ mRNA leader can compete with formation of the pseudoknot required for repZ translation (16). In this report, we show that Inc RNA 54 not only reduces the initial rate of the Inc RNA-RepZ mRNA hybridization by its transient base pairing to loop I, but also stops the reaction by its subsequent irreversible binding to the loop (Fig. 2). Intermolecular base pairing found in the latter irreversible complex was quite extensive, spanning probably Ͼ 20 base pairs and forms two helices separated at the site of the initial base pairing (Figs. 3  and 4). The equilibrium dissociation constant for this complex was tentatively estimated to be ϳ3 nM (Fig. 5), a value smaller than that for the transient initial interaction, 6 or 10 nM (16). Nevertheless, the late complex was unstable in denaturing gels, probably because of a structural stress stored during the complex formation (Fig. 6). Based on these results, we propose that the stepwise model for antisense RNA binding proposed for the ColE1 system (class H) (Ref. 18; see the Introduction) also applies to a class L antisense RNA, but with completely different molecular mechanisms. We believe that the late Inc RNA 54 complex characterized here as C** corresponds to an intermediate complex formed between IncB pMU720 RNA T (a 5Ј-truncated derivative) and its target that was analyzed by native gel electrophoresis (Ref. 23; see below).
In view of differential control of repY and repZ, several important functions could be assigned to the late intermediate step, as depicted in Fig. 1, step g. First, it is conceivable that the observed irreversible base pairing would trap the initial transient interaction in a bound state, thereby driving the binding reaction. Second, for the control of repZ, the inhibition of the pseudoknot formation should be established at this stage, because the bases critical for the pseudoknot formation (G-327ϳG-330) are still masked (Fig. 4B). Furthermore, the presumed side-by-side configuration of the complex would bring the ends of the stems closer, thereby allowing the Inc 5Ј leader (absent in Inc RNA 54 ) to base pair efficiently with a region distal to the repY RBS (Fig. 1, steps g and h).
The antisense RNA I of IncB pMU720 replicon is identical to the Inc RNA of IncI␣ ColIb-P9 except at the bases indicated in Fig. 4B. Therefore, the partitioning of these plasmids into different incompatibility groups (22) should be explained at the level of the C** complex formation as analyzed in this study. It is also noteworthy that the mutant IncI␣ Inc RNA with a loop mutation (G329ЈA in ColIb-P9) abolishing the C* formation (16) showed no detectable incompatibility (or interaction) against both wild-type IncI␣ and IncB plasmids (22). Thus, the mutation in the C* region has more dramatic effect on plasmid incompatibility than that located outside affecting only C** formation, consistent with our previous results (19).
Complex C** Formed by Other Class L Antisense RNAs-Although the presence of C* has not yet been demonstrated for IncFII plasmids, an intermediate complex, called the "extended kissing complex," was extensively analyzed with the CopA antisense RNA encoded by the R1 plasmid of IncFII group (24 -27). This complex, formed between a derivative of CopA lacking its 5Ј leader and the CopA target, was often characterized by native gel electrophoresis and gave the K d of 7.4 nM (26). In light of our study, we believe that this kissing complex corresponds to C** for the following reasons. First, 5Ј-truncated CopA had been incubated with the target mRNA for more than 5 min to form this complex (24 -27); this time is long enough to allow progression to C**, based on our results (Fig. 2). Second, reduced but remaining RNase T2 sensitivity at a loop G residue found in the 5Ј-truncated CopA in complex with the target (24) compares very well with the weak RNase T1 sensitivity at G-327 and G-328 found in the structure I of RepZ mRNA in complex with Inc RNA 54 (Fig. 3). Third, a measurement by native gel electrophoresis may underestimate the K d value, because the complex can dissociate in the gel. In fact, Hjalt and Wagner (26) realized this problem and measured K d differently by incubating the target RNA with various concentrations of the 5Ј-truncated CopA and measuring the amount of free form of the former that can bind intact CopA RNA: K d thus obtained was 1.6 nM consistent with our preliminary estimation for C** (ϳ3 nM ; Fig. 5). These values are significantly higher than the K d obtained for C** in ColE1, 0.1 nM (18). This fact agrees with the idea that C** in ColIb-P9 (or R1) is destabilized by a structural stress imposed by the propagation of a single initial loop-loop interaction (Fig. 6). C** in ColE1 should be free of such a stress, because it is composed of multiple loop-loop interactions involving a smaller number of base pairs (18).
Finally, the secondary structure of the extended kissing complex was analyzed with Pb(II)-induced cleavage experiments (27). Malgren et al. (27) appear to assume that the extended intermolecular base pairing occurs only at the 3Ј side of the target loop, even though Pb(II) sensitivity was also reduced at the 5Ј side. Using four different RNases that have potential to cleave all four kinds of nucleotides plus a double-strand specific RNase V1, we unambiguously showed that both sides of the partially base paired target loop were protected by Inc RNA 54 (Fig. 4B). In sharp contrast, the 5Ј side of the target loop in complex with Ps 14 showed dramatically increased sensitivity to RNases T1 and PhyM (Fig. 4C). Another new aspect concerns the weak single-stand specific RNase sensitivity at U-326ϳG-328 (Fig. 3). These results raised the interesting possibility that the intermolecular helix is separated at the top of structure I, resulting in a stacking trend between each separated helix and the stem of Inc RNA or structure I (Fig. 4B). We already discussed the biological significance for this unique configuration as above.
Evolution of Control of Plasmid Replication by Class L Antisense RNAs-The whole pathway of Inc RNA binding as shown in Fig. 1 appears to fit best for its differential control of repY and repZ expression. Antisense RNAs encoded by IncFII plasmids also follow a similar binding pathway (27). Nevertheless, the translation of rep genes encoded by these plasmids does not require formation of a pseudoknot (9,28), unlike the case for IncI␣ (13) and IncB (15) plasmids. Instead, IncFII plasmids additionally encode a transcriptional repressor for the rep mRNA, located upstream of the repressible promoter (3) (see Fig. 7). How did these differences arise during the course of evolution?
We believe that the difference between these related but distinct types of plasmids arose at least partly from difference in RNA secondary structure that blocks the RBS for respective rep genes, as depicted in Fig. 7B. In the case of IncI␣/IncB-type, the secondary structure that blocks the rep RBS is very tight. Therefore the translation of rep is possible only when the pseudoknot induced during the RLP translation termination transiently keeps the rep RBS accessible to the ribosome (6,14,16). On the other hand, the inhibitory secondary structure found in IncFII plasmid replicons is less tight around the rep RBS. Thus, the refolding of the inhibitory secondary structure, unfolded during the RLP translation termination, would be slow enough to allow the ribosome to bind the rep RBS. However, this kind of translational coupling could enable leaky, unregulated translation initiation of rep, as demonstrated for a mutant ColIb-P9 replicon lacking the pseudoknot and carrying less tight structure III (rep inhibitory stem-loop) (6). It is therefore conceivable that the ancestor of IncFII-type plasmids acquired an additional control system at the transcriptional level, in order to complement this regulatory defect.
It would be difficult to solve the evolutionary relationship between these regulatory systems. Nevertheless, we favor the model that the pseudoknot-mediated control in IncI␣/IncB plasmids is more ancient and was lost in the ancestor of IncFII plasmids during recombination between plasmids. In support of this scenario, the "mosaic" structure of large, self-transmissible plasmids is not uncommon (20), and there are even some cases of naturally occurring hybrid replicons carrying the control region of IncI␣/IncB-type and the rep gene of IncFII-type, or vice versa (9). Evolution of plasmids by recombination may have been stimulated by horizontal transfer of plasmids between a variety of bacteria encoding different restriction-modification systems (29). On the other hand, certain types of plasmids that are transferred only between limited species of enterobacteria and not exposed to harsh natural environment might retain the intact plasmid genome and hence the "molecular fossil" of replication control system. The ongoing genomewide sequence analyses for different groups of self-transmissible plasmids may answer whether ColIb-P9, originally isolated from Shigella sonnei, is one such plasmid.