In Vitro Analysis of the Interaction between the FinO Protein and FinP Antisense RNA of F-like Conjugative Plasmids*

The FinO protein regulates the transfer potential of F-like conjugative plasmids through its interaction with FinP antisense RNA and its target, traJ mRNA. FinO binds to and protects FinP from degradation and promotes duplex formation between FinP and traJ mRNA in vitro. The FinP secondary structure consists of two stem-loop domains separated by a 4-base spacer and terminated by a 6-base tail. Previous studies suggested FinO bound to the smooth 14-base pair helix of stem-loop II. In this investigation, RNA mobility shift analysis was used to study the interaction between a glutathione S-transferase (GST)-FinO fusion protein and a series of synthetic FinP andtraJ mRNA variants. Mutations in 16 of the 28 bases in stem II of FinP that are predicted to disrupt base pairing did not significantly alter the GST-FinO binding affinity. Removal of the single-stranded regions on either side of stem-loop II led to a dramatic decrease in GST-FinO binding to FinP and to the complementary region of the traJ mRNA leader. While no evidence for sequence-specific contacts was found, the results suggest that FinO recognizes the overall shape of the RNA and is influenced by the length of the single-stranded regions flanking the stem-loop.

The FinO protein regulates the transfer potential of F-like conjugative plasmids through its interaction with FinP antisense RNA and its target, traJ mRNA. FinO binds to and protects FinP from degradation and promotes duplex formation between FinP and traJ mRNA in vitro. The FinP secondary structure consists of two stem-loop domains separated by a 4-base spacer and terminated by a 6-base tail. Previous studies suggested FinO bound to the smooth 14-base pair helix of stemloop II. In this investigation, RNA mobility shift analysis was used to study the interaction between a glutathione S-transferase (GST)-FinO fusion protein and a series of synthetic FinP and traJ mRNA variants. Mutations in 16 of the 28 bases in stem II of FinP that are predicted to disrupt base pairing did not significantly alter the GST-FinO binding affinity. Removal of the single-stranded regions on either side of stem-loop II led to a dramatic decrease in GST-FinO binding to FinP and to the complementary region of the traJ mRNA leader. While no evidence for sequence-specific contacts was found, the results suggest that FinO recognizes the overall shape of the RNA and is influenced by the length of the singlestranded regions flanking the stem-loop.
RNA-protein interactions are important in the post-transcriptional regulation of RNA metabolism and expression. An excellent example is the FinOP fertility inhibition system, which controls the transfer frequency of F-like conjugative plasmids. Expression of the plasmid transfer genes is positively regulated by the TraJ protein, which is required for activation of transcription from the major transfer operon promoter, pY. Expression of traJ is negatively regulated by the combined actions of the finO and finP gene products. FinP is a plasmidspecific ϳ79-base antisense RNA molecule, complementary to part of the 5Ј-untranslated region (UTR) 1 of traJ mRNA (1-3), which includes the ribosome binding site and first two codons of traJ. The FinP secondary structure (GenBank accession number U01159) (4) consists of 2 stem-loop (SL) domains, separated by a 4-base spacer and terminated by a 6-base tail (Fig. 1). The 5Ј-UTR of traJ mRNA forms the mirror image of FinP, with the traJ RBS and start codon being localized to SLIc (complemen-tary to FinP SLI). The traJ leader also contains an extensive single-stranded region and a third stem-loop, SLIII, at its 5Ј end. Binding of FinP to the traJ UTR is believed to sequester the traJ RBS, preventing its translation and repressing plasmid transfer.
FinP requires a protein cofactor, FinO, to exert its negative effect on traJ expression. FinO is a 21.2-kDa basic protein (5), which has been shown to bind FinP SLII and traJ mRNA in vitro, increasing the rate of duplex formation (6). Binding of FinO to FinP also protects FinP from RNase E-mediated degradation (7,8), increasing its steady-state concentration to a level that allows sequestration and reduced expression of the traJ mRNA (9). The F plasmid is derepressed for transfer (i.e. it transfers constitutively) due to the insertion of an IS3 element within its finO gene (10,11). This effect can be overcome by supplying a functional finO gene from a related plasmid such as R6 -5 (5) or R100 (12) in trans, resulting in a 100 -1000fold reduction in F plasmid transfer.
The focus of this study is the specificity of the RNA-protein interaction between FinP antisense RNA and the FinO protein.
Eight alleles of FinP have been described for F-like conjugative plasmids (Fig. 2), with sequence identity residing in the stem, spacer and tail regions and variability in the two loops (2,13). Two alleles of FinO exist (13), which show very little sequence variation, but are classified on the basis of their levels of repression of F-like plasmids which is tied to their levels of expression (5). FinO is not plasmid-specific, which suggests that the loops are unimportant for RNA-protein recognition. The FinO protein does not share homology with any of the protein sequence motifs found in other RNA-binding proteins (14 -16), providing no clues to the amino acid residues involved in RNA recognition.
This report presents a preliminary characterization of the RNA targets recognized by FinO. RNA mobility shift analysis was used to determine the binding affinity of a glutathione S-transferase (GST)-FinO fusion protein for a series of synthetic FinP and traJ mRNA variants. High affinity binding by GST-FinO was shown to be dependent on the presence of SLII, flanked on either side by single-stranded regions. Unexpectedly, mutations that disrupted base pairing in SLII were tolerated, as was the introduction of internal loops. Furthermore, decreasing the length of the 3Ј tail reduced GST-FinO binding while altering the sequence had no effect. The same structural features were recognized in traJ mRNA, suggesting that FinO does not make sequence-specific contacts with the RNA, but recognizes its overall shape. from these BamHI-linearized pLJ5-13 mutant plasmids, resulting in the addition of 7 bases (GGGGAUC) to their 3Ј ends derived from the BamHI site in the vector. All other RNAs were transcribed from gelpurified PCR products prepared with Vent TM DNA polymerase (New England Biolabs) using the appropriate primers and DNA templates, as detailed in Tables I and II. pOX38-Km has been described previously (17).
Preparation of RNAs-Uniformly labeled RNAs were prepared by in vitro transcription of DNA templates in reaction mixtures containing 1ϫ T7 RNA polymerase buffer (Boehringer Mannheim), 1 unit/l RNAguard (Amersham Pharmacia Biotech), 10 Ci of [␣-32 P]UTP (Mandel Scientific), 0.5 mM GTP, ATP, and CTP, and 0.02 mM UTP at 37°C for 1 h. DNA was removed with 10 units of RNase-free DNase I (Boehringer Mannheim) for 15 min at 37°C, and the labeled RNA was gel-purified as described (8). RNA was stored at Ϫ20°C for 1 week without noticeable degradation. A step of RNA denaturation-renaturation did not improve or modify the binding affinity of the protein; therefore, gelpurified RNA was used directly for gel-shift analysis.
Gel-shift Analysis of RNA-Protein Interactions-pGEX-FO2 (6) was used for expression of a protein fusion between GST and R6 -5 FinO. GST-FinO was purified using glutathione-agarose affinity as described (6), except that the cells were lysed using a French press, rather than by sonication (18). The ability of GST-FinO to bind each RNA was assessed by RNA mobility shift analysis (6). 7.5 fmol of 32 P-labeled RNA was incubated with increasing amounts of GST-FinO (0 -31.9 pmol) in a total volume of 30 l for 30 min at room temperature in buffer containing 50 mM Tris-HCl, pH 8.0, 1 mM EDTA, 10 mM NaCl, 100 g/ml bovine serum albumin (RNase/DNase-free; Boehringer Mannheim), 7.6 units of RNAguard (Amersham Pharmacia Biotech), 10% glycerol. Samples were loaded onto a continuously running (150 V) 6% or 8% nondenaturing polyacrylamide gel in 1ϫ TBE buffer and electrophoresed at room temperature for 1.5 h. Bands were visualized by autoradiography and phosphorimaging and quantified using a Molecular Dynamics PhosphorImager 445 SI. The equilibrium association constant (K a ) for GST-FinO binding was calculated from the protein concentration that caused 50% of the labeled RNA to shift in the gel (6,19). For RNA variants that gave more than one band shift, K a values were calculated by considering all bound RNA as a single species. Except where noted, K a values were calculated from three independent determinations. The percent binding relative to FinP was calculated as 100 ϫ (K a variant/K a FinP).

RESULTS
We have previously shown that a fusion of FinO from the related plasmid, R6 -5, to GST is active in vivo and in vitro (6,18). Furthermore, GST-FinO and FinO obtained from thrombin cleavage of the fusion protein bind to FinP RNA with equal affinity. 2 In the present study, the relative equilibrium association constants for GST-FinO binding to FinP variants were determined by performing gel-shift assays in which radiolabeled RNA was combined with increasing amounts of purified GST-FinO protein. Labeled RNA was synthesized by in vitro run-off transcription using PCR-generated templates or cloned fragments (see "Experimental Procedures"). Transcripts prepared from cloned fragments contained 7 additional bases (GGGGAUC) at their 3Ј ends derived from the BamHI site in the vector used to linearize the plasmids. To ensure that the 7 bases from the vector did not alter the binding constants, FinP RNA and the quadruple mutant, FinP C41:U/C46:U/G66:U/ G71:U, with or without the additional 3Ј bases, were synthe-2 M. J. Gubbins and L. S. Frost, unpublished data. sized. The presence of the 7 extra bases at the 3Ј end had no effect on GST-FinO binding (data not shown).
A comparison of GST-FinO binding affinities to FinP, SLI (nucleotides 1-34, Fig. 1) and SLII (nucleotides 35-79), which were transcribed from PCR-generated templates, is shown in Fig. 3. The K a for GST-FinO binding to FinP was 2.0 ϫ 10 7 M Ϫ1 , 50-fold higher than that for SLI and 2-fold higher than SLII (summarized in Table III). These values are higher than previously reported (6) and likely reflect an increase in the fraction of active GST-FinO in preparations using the French press method for cell lysis, rather than sonication. Since the fraction of GST-FinO that is active was not determined, these values may still be underestimated. As a negative control, incubation of either FinP, SLI or SLII with GST alone did not result in the formation of shifted species in the gel (data not shown). This indicated that the complexes formed in the presence of the fusion protein were the result of RNA interaction with FinO.
In agreement with the results of van Biesen and Frost (6), these results suggest that the major determinants for FinO binding reside in SLII, which prompted us to look for differences between SLI and SLII. The most notable structural difference between SLI and SLII is an A-A mismatch within stem I (Fig. 1). SLI A12:U, which creates an A:U base pair (bp) at this site, results in a fully duplexed SLI. Its ability to bind GST-FinO was tested and when compared with SLI with the natural A-A mismatch, was not increased (Table III). Thus, this 11-bp helix was not sufficient for recognition. A second obvious difference between stem-loops I and II is the sequence of the loops. To determine if the loops contributed to the specificity of GST-FinO binding, FinP RNAs from ColB2 and R100 -1 were synthesized. ColB2 loop I is identical to F, but differs at 6 of the 7 bases comprising loop II (Fig. 2). R100 -1 differs from F at 1 base in loop I and 2 bases in loop II, which is also 1 base smaller than F loop II. The K a for GST-FinO binding to ColB2 was almost identical to F and was slightly increased for R100 -1 (Table III). These results indicate that the sequences of the loops are unimportant for FinO binding. The sequences of the two FinP stems, although highly conserved among the finP alleles (Fig. 2), differ significantly from each other and could account for the preference of GST-FinO for SLII. The effect of mutations in stem II of full-length FinP on GST-FinO binding were examined, and the binding data are summarized in Table III. Spontaneous single base mutations on the 5Ј side of stem II (G42:A, G45:A, C46:U) reduced the corresponding affinity constants by no more than 30% when compared with wild-type. In the 5Ј upper half of stem II, A47:G (naturally present in R100 -1) and G49:A did not significantly reduce GST-FinO binding, whereas G48:A and A51:G reduced binding by about 50%. Each of these mutations is expected to disrupt Watson-Crick base pairing, reducing the helical length of stem II. The C41:U/C46:U double mutant, which could lead to formation of two non-Watson-Crick G:U base pairs and maintain the helicity of SLII, showed a 30% reduction in binding constant. To disrupt these potential interactions, FinP variants G66:U/G71:U, C46:U/G66:U/C71:U and C41:U/C46:U/ G66:U/G71:U were constructed (Table III). None of these mutations significantly altered GST-FinO binding, suggesting that a continuous duplex is not important for efficient binding.  (Table III).
GST-FinO Binding Is Enhanced by Single-stranded Regions on Either Side of SLII-Since the stem II point mutants examined did not significantly alter the binding affinity of GST-FinO, we turned our attention to the conserved single-stranded regions of FinP: the 5Ј leader, spacer, and 3Ј tail. Removal of 3 bases from the 5Ј leader following the initiating G required for T7 RNA polymerase transcription (FinP ⌬A2,U3,A4) had little effect on binding affinity (Table IV); however, the spacer and tail regions proved to be important, as shown in Fig. 4 and summarized in Table IV. Deletion of either the spacer or 3Ј tail from SLII had minor to moderate effects on GST-FinO binding, reducing the K a values by 1.3-fold and 5.5-fold, respectively. Deletion of both the spacer and 3Ј tail from SLII had a profound effect, reducing the binding constant 14-fold.
In order to determine whether FinO had a nonspecific affinity for single-stranded RNA, the ability of polyuridylic acid to compete with FinP for GST-FinO binding was tested. The results (data not shown) indicate that the GST-FinO/FinP interaction was unaffected by the presence of a large molar excess of polyuridylic acid. These results suggest that the singlestranded spacer and 3Ј tail make important contributions to a higher order structure recognized by FinO.
The nucleotide sequences of synthetic stem-loops I and II were arbitrarily chosen (6) such that SLI includes the 5Ј leader and stem-loop I, whereas SLII includes the spacer, stem-loop II and 3Ј tail (Fig. 1). Since GST-FinO binding to SLII was strongly dependent on the presence of both single-stranded flanking sequences, the observed low affinity of GST-FinO for SLI might be due to the absence of a 3Ј single-stranded region. To test this hypothesis, SLI variants were constructed that had either the spacer or tail sequence added to their 3Ј ends. When assayed for GST-FinO binding, the K a values for SLI-spacer and SLI-tail were increased 3-fold and 10-fold, respectively ( Fig. 5; Table IV). These results clearly demonstrate the im-  portance of the 3Ј flanking sequence to the structure recognized by GST-FinO.
The Length, but Not Sequence, of the FinP 3Ј Tail Is Important for GST-FinO Binding-The effect of decreasing the length of the 3Ј tail following SLII on GST-FinO binding was examined since addition of a 7-base extension had no effect (see above). Shortening of the SLII 3Ј tail from 6 bases (GAUUUU) to 4 bases (GAUU) decreased the K a 3-fold to 0.38 ϫ 10 7 M Ϫ1 ( Fig. 6; Table IV). A further reduction in tail length to 2 bases (GA) decreased the K a another 6-fold to 0.06 ϫ 10 7 M Ϫ1 . These results suggest that a minimum 6-base 3Ј tail is necessary for efficient binding by GST-FinO. To determine whether the presence of the 3Ј tail reflected a sequence-specific or general requirement for additional bases flanking SLII, variants SLII-GAAAAA, spacer-SLII-GACA, and FinP G74:C/A75:C were constructed (Table IV). The sequence of each variant was chosen to avoid introducing any other obvious secondary structural features. Comparison of the variant pairs (SLII-GAAAAA with SLII-tail; spacer-SLII-GACA with spacer-SLII-GAUU and FinP G74:C/A75:C with wild-type FinP) in Table IV shows that these base transversions in the 3Ј tail had minor effects on GST-FinO binding. These results indicate that the length, but not sequence, of the FinP 3Ј tail is important for high affinity binding by GST-FinO.
GST-FinO Recognizes the Same Structural Features in traJ mRNA-The results of an earlier study (6) showed that GST-FinO binds to a truncated 184-base version of the sense mRNA, traJ, with a K a similar to that for FinP. The sequence and secondary structure of the first 117 bases of F traJ are shown in Fig. 1. The secondary structure of traJ RNA between nucleotides 33 and 111 is almost identical to FinP, with the following exceptions: SLIc of traJ mRNA has an additional mismatch (A-C), which is paired in FinP SLI; SLIIc of traJ is 2 bases shorter than SLII of FinP, resulting in a 6-base traJ spacer, as compared with 4 bases for FinP.
To further characterize the interaction between GST-FinO and F plasmid-encoded traJ184 (previously called TraJ211, Ref. 6), a series of 3Ј truncated traJ variants were created and their binding to GST-FinO was compared with that of FinP. As seen in Fig. 7, addition of increasing amounts of GST-FinO led to the conversion of low molecular weight bands to one or more higher molecular weight bands (see "Discussion" for further comments). GST-FinO bound traJ184 almost as well as FinP (95%), with a K a of 1.9 ϫ 10 7 M Ϫ1 . Deletion of 74 bases from the 3Ј end of traJ184, creating traJ110 (see Fig. 1), yielded a modest 25% reduction in GST-FinO binding to 1.4 ϫ 10 7 M Ϫ1 . Removal of SLIc from traJ110, which results in a transcribed  product of 77 bases (traJ77; similar to SLII of FinP, Fig. 1) reduced GST-FinO binding by 48% to 0.73 ϫ 10 7 M Ϫ1 . This is similar to the 45% reduction in the binding constant observed for the removal of SLI from FinP (see Fig. 3; Table III). Further mutation of traJ77 to eliminate the spacer residues 3Ј to SLIIc (A72 to C77), creating traJ71, decreased the K a for GST-FinO binding another 53% to 0.34 ϫ 10 7 M Ϫ1 . As with FinP, this result suggests that SLIIc and its 3Ј flanking sequence are important determinants for high affinity GST-FinO binding. DISCUSSION This report describes the structural features of RNA recognized by the FinO protein. RNA-binding proteins have generally been shown to target single-stranded regions caused by loops, bulges, and mismatches or between helical stems, rather than the duplexed regions themselves (14,20,21). Duplexed regions are necessary, however, for spacing and presentation of the single-stranded nucleotides in the correct orientation (20). The hairpin loops represent an obvious single-stranded region of FinP available for protein binding. Three lines of evidence suggest that FinO does not make sequence-specific contacts with bases in the FinP loops. First, although the loop sequences vary between finP alleles (2), FinO is exchangeable among F-like plasmids (5,22). Second, the traJ mRNA loops are complementary in sequence to FinP, and yet GST-FinO bound traJ and FinP with nearly equal affinity (Figs. 3 and 7). Third, GST-FinO bound F, ColB2, and R100 -1 FinPs with the same relative affinity, even though the sequences of the loops vary considerably, especially within loop II.
The results reported by van Biesen and Frost (6) and in this study indicate that the sequence between nucleotides 35 and 79 (SLII) of FinP is sufficient and necessary for high affinity GST-FinO binding. Stem II is fully duplexed in FinP and presents a relatively poor sequence-specific target for FinO. Attempts to disrupt the helical nature of SLII by introduction of internal loops and base substitutions were freely tolerated, indicating that a continuous duplex is not necessary for FinO binding. In agreement with this, single mutations at three of the sites tested (C41, C46, G49) and at G50, had no effect on FinP repressor activity from the related conjugative plasmid R1, as measured by conjugation frequency (23). Thus, the mutant FinP RNAs are also fully stabilized in vivo by FinO (9). These results indirectly demonstrate that FinO binding is not affected by these base substitutions in vivo, in accord with the results obtained in this study, in vitro.
Earlier studies (6) showed that GST-FinO could bind to a FinP/traJ184 duplex or a duplex formed between SLI and the complementary sequence of traJ mRNA. However, the present results indicate that the 14-bp duplex, SLII alone, was not sufficient for high affinity binding by GST-FinO. These apparently conflicting results can be reconciled by the finding that single-stranded regions adjacent to the duplexed RNA were necessary for binding by GST-FinO. In this respect, FinO resembles the stem-loop binding protein (SLBP), which binds to the 3Ј end of histone mRNA in mammalian cells (24). Efficient binding by SLBP requires at least three nucleotides each 3Ј and 5Ј of a stem-loop. FinO requires at least 6 nucleotides 3Ј to SLII and as many as 4 nucleotides 5Ј to SLII, although the length of the 5Ј spacer was not examined in this study. Like FinO, SLBP does not have a strict loop size requirement, suggesting that specific contacts do not involve the loop. However, unlike the interaction between FinO and FinP, the sequence of the stem and flanking regions is important for SLBP binding.
Since no evidence was obtained for sequence-specific contacts between GST-FinO and FinP, our data suggest that FinO FIG. 6. The length of the SLII 3 tail is important for RNA recognition. The effect of SLII 3Ј tail-length on GST-FinO binding was examined by gel shift analysis as detailed in the legend to Fig. 4. b, percentage of binding relative to FinP (as described under "Experimental Procedures"). Four traJ variants of different length, indicated by their names, were synthesized and subjected to gel shift analysis using the conditions described in the legend to Fig. 4, except that samples were resolved on a nondenaturing 6% polyacrylamide gel. b, percentage of binding relative to FinP (as described under "Experimental Procedures"). recognizes the overall shape of the RNA conferred by a stemloop structure, flanked on either side by single-stranded regions. Congruent with this, GST-FinO recognizes the same structural features in traJ mRNA. The requirement for a 6-nucleotide flanking region 3Ј to SLIIc is fulfilled by the traJ spacer, which is 2 bases longer than the FinP spacer (Fig. 1). In addition the traJ spacer, which differs from the FinP 3Ј tail at 4 of the 6 bases, can serve as a functional 3Ј flanking region, indicating that sequence is unimportant for binding by FinO. Addition of the FinP tail sequence (GAUUUU) to the 3Ј side of SLI conferred moderate GST-FinO binding, although the binding constant was 2-fold lower than that for SLII with the equivalent 3Ј tail. These results suggest that the RNA conformation recognized by GST-FinO can be adopted quite efficiently by three different sequences: traJ mRNA with its 6-base single-stranded spacer, SLI with the FinP 3Ј tail, and SLII with the 3Ј tail.
The functionally related RNA-binding protein, Rom, has also been shown to recognize RNA in a structure-dependent, rather than sequence-dependent fashion. Rom binds to an unstable complex formed between the complementary hairpin loops of RNA I and RNA II of ColE1 plasmids (25). The results of three independent studies indicate that Rom is capable of binding and stabilizing any complex formed by pairs containing fully complementary loop sequences (26 -28). Furthermore, Rom has been shown to extend the average half-life of general pulselabeled Escherichia coli mRNA (29). Like Rom, FinO might have many targets in the cell, but the role of FinO in promoting duplex formation can occur only when it is bound to complementary RNAs. GST-FinO binds pBR322 antisense RNA I with an affinity constant of 1.4 ϫ 10 7 M Ϫ1 , 70% of that for FinP. 3 The secondary structure of pBR322 RNA I consists of a 9-base leader followed by three stem-loop domains, with a 2-base spacer between stem-loops I and II (30,31). Although the region of RNA I bound by FinO has not been determined, it is conceivable that FinO has a nonspecific affinity for the third stem-loop, which lacks a 3Ј tail, but is longer than 14 bp. Further experiments are needed to define the binding site and establish the specificity of this interaction.
Interestingly, binding of GST-FinO to traJ184 gave rise to four shifted bands of different mobility, with the largest complex being retained in the well (Fig. 7). The fastest migrating complex was converted to more slowly migrating complexes with increasing GST-FinO concentration. GST-FinO interaction with traJ110 resulted in three distinct complexes, whereas two complexes were formed with either traJ77 or traJ71. One possible explanation for the formation of multiple complexes may be stepwise binding of GST-FinO monomers to multiple sites on its RNA target. The bands retained in the well likely represent complexes formed due to GST-FinO aggregation. Cooperative binding has been reported for the human immunodeficiency virus Rev-RRE interaction (32,33,34,35) and between p24 (a 220-amino acid truncated polypeptide of the protein kinase, PKR) and human immunodeficiency virus dsTAR RNA (36). A closer look at GST-FinO binding to its antisense RNA targets (Fig. 3) shows that only one complex was formed between GST-FinO and either SLI or SLII, whereas two complexes were formed with full-length FinP (more readily distinguishable by under-exposure of this gel; data not shown). Although the strongest GST-FinO binding was achieved with SLII, low affinity binding was observed between GST-FinO and SLI with its attached 4-base spacer. Thus, binding of GST-FinO to its primary binding site, SLII, may promote binding to the low affinity site, SLI, resulting in the more slowly migrating complex observed at high GST-FinO concentrations. Similarly, the higher molecular weight complexes observed with the traJ RNA variants may represent successive GST-FinO binding to stem-loops IIc, Ic, and III. Further experiments are needed to determine the specificity of FinO binding to these potential low affinity sites. In addition, since the physiological FinO concentration is not presently known but is thought to be low, based on mRNA levels (5), the relevance of the higher molecular weight complexes, which are formed at high FinO concentrations, remains to be established.