trp RNA-binding Attenuation Protein-mediated Long Distance RNA Refolding Regulates Translation of trpE inBacillus subtilis *

Expression of the trpEDCFBA operon is regulated at both the transcriptional and translational levels by thetrp RNA-binding attenuation protein (TRAP) ofBacillus subtilis. When cells contain sufficient levels of tryptophan to activate TRAP, the protein binds to trpoperon transcripts as they are being synthesized, most often causing transcription termination. However, termination is never 100% efficient, and transcripts that escape termination are subject to translational control. We determined that TRAP-mediated translational control of trpE can occur via a novel RNA conformational switch mechanism. When TRAP binds to the 5′-untranslated leader segment of a trp operon read-through transcript, it can disrupt a large secondary structure containing a portion of the TRAP binding target. This promotes refolding of the RNA such that thetrpE Shine-Dalgarno sequence, located more than 100 nucleotides downstream from the TRAP binding site, becomes sequestered in a stable RNA hairpin. Results from cell-free translation, ribosome toeprint, and RNA structure mapping experiments demonstrate that formation of this structure reduces TrpE synthesis by blocking ribosome access to the trpE ribosome binding site. The role of the Shine-Dalgarno blocking hairpin in controlling translation oftrpE was confirmed by examining the effect of multiple nucleotide substitutions that abolish the structure without altering the Shine-Dalgarno sequence itself. The possibility of protein-mediated RNA refolding as a general mechanism in controlling gene expression is discussed.

Studies on the regulation of protein synthesis have shown that the RNA secondary structural features present in the 5Ј-UTR 1 dramatically influence translation initiation in both prokaryotic and eukaryotic organisms (for recent reviews see Refs. [1][2][3][4][5][6][7]. In prokaryotic mRNAs, a conserved stretch of 4 -6 nucleotides called the Shine-Dalgarno (SD) sequence is usually found 4 -11 nucleotides upstream of the initiation codon. The SD sequence base pairs with the 16 S rRNA present in the 30 S ribosomal subunit and thereby correctly positions the initiation codon in the ribosome (8,9). Translational control mechanisms have been identified in prokaryotes that involve blocking the SD sequence either by RNA secondary structure (9 -12) or by a bound protein (13)(14)(15)(16). In the known translational control mechanisms that occur by formation of SD blocking hairpins, formation of the inhibitory structure is spontaneous and does not require protein factors.
Expression of the Bacillus subtilis tryptophan biosynthetic genes is regulated in response to changes in the intracellular level of tryptophan at both the transcriptional and translational levels (for a recent review, see Ref. 17). Six of the seven trp genes are clustered in the trpEDCFBA operon. Transcription of the trp operon is regulated by an attenuation mechanism in which tryptophan-activated trp RNA-binding attenuation protein (TRAP) binds to 11 closely spaced (G/U)AG repeats (seven GAG and four UAG) (18 -25). TRAP binding to the 11 trinucleotide repeats in the nascent trp leader transcript prevents or disrupts formation of an RNA secondary structure, the antiterminator, thereby allowing formation of an overlapping Rho-independent terminator and hence causing termination of transcription before RNA polymerase reaches the trp structural genes. In the absence of TRAP binding, formation of the antiterminator prevents formation of the terminator, resulting in transcriptional read-through into the trp structural genes. In addition to regulating transcription of the trp operon, TRAP also regulates translation of trpE. Previous in vivo studies demonstrated that TRAP is responsible for regulating trpE translation 10 -15-fold (19,25). RNA structure predictions of the trp operon read-through transcript indicated that the most thermodynamically stable conformation of the leader RNA segment would contain a large secondary structure that includes a portion of the TRAP binding site in the 5Ј-half of the stem. It was proposed that TRAP binding to these repeats would disrupt the large secondary structure and promote refolding of the leader RNA such that the trpE SD sequence would be sequestered in an RNA hairpin (19,25). It was also shown that multiple nucleotide substitutions predicted to destabilize the SD blocking hairpin, without altering the SD sequence itself, reduced the ability of TRAP to regulate TrpE synthesis in vivo (25). Thus, the ability of TRAP to alter the conformation of trp operon read-through transcripts could partially explain the TRAP-dependent translational control of trpE expression that was previously observed (19,25).
The one unlinked trp gene, trpG, is a part of a folic acid biosynthetic operon (26). Expression of trpG is regulated by a translational control mechanism in which tryptophan-activated TRAP can bind to nine trinucleotide repeats (seven GAG, one UAG, and one AAG) that surround and overlap the trpG ribosome binding site. TRAP binding to these repeats represses TrpG synthesis by directly blocking ribosome access to the trpG ribosome binding site (15,16).
The crystal structure of TRAP complexed with L-tryptophan shows that TRAP is composed of 11 identical subunits arranged in a single ring, with one molecule of tryptophan bound between adjacent subunits (23,27). The RNA binding motif of TRAP consists of 11 repeated KKR motifs that line the periphery of the TRAP complex. This finding suggests that TRAP-RNA interaction proceeds through a mechanism in which one KKR motif binds to one (G/U)AG repeat, thereby wrapping the RNA around the outside of the TRAP complex (28).
In the present study, we performed experiments in vitro to elucidate the molecular mechanism responsible for TRAP-mediated translation control of trpE. Our cell-free translation and RNA structural studies demonstrate that TRAP binding to trp operon read-through transcripts does in fact promote refolding of the untranslated trp leader such that the trpE SD sequence, which is located more than 100 nucleotides downstream from the TRAP binding site, becomes sequestered in a stable RNA hairpin. Moreover, we found that formation of this SD blocking hairpin inhibits TrpE synthesis by blocking ribosome access to the trpE ribosome binding site.

EXPERIMENTAL PROCEDURES
Plasmids and Bacterial Strains-Plasmid pPB22 carrying the wild type B. subtilis trp promoter and leader (WTtrpL) has been described (20). The plasmid pINT-SDtrpL contains several trp leader point mutations (SDtrpL) that destabilize the predicted trpE SD blocking RNA hairpin without disrupting the trpE SD sequence itself (25). Plasmid pHD1 was constructed by subcloning the 730-base pair EcoRI-HindIII fragment containing the SDtrpL from pINT-SDtrpL into the EcoRI-HindIII sites of the pTZ18U polylinker (U.S. Biochemical Corp.). A description of plasmid pTRP-H3B2 has been published (29). pHD12 was constructed by subcloning the 2.4-kilobase pair PvuII fragment containing the trp operon promoter and leader as well as the trpED structural genes from pTRP-H3B2 into the unique SmaI site of pTZ18U. The plasmid pHD15 contains the WTtrpL, trpE, and the 5Ј-end of trpD. This plasmid was constructed by simultaneously ligating the 730-base pair EcoRI-HindIII fragment from pPB22 containing the WTtrpL and the 5Ј-end of trpE, as well as the 1.7-kilobase pair HindIII-BamHI fragment from pHD12 containing the 3Ј-end of trpE and the 5Ј-end of trpD into the EcoRI-BamHI sites of the pTZ18U polylinker. Plasmid pHD16 was constructed in the same manner as pHD15 except that the 730-base pair EcoRI-HindIII fragment containing the SDtrpL from pHD1 was used in place of the WTtrpL. The B. subtilis integration vector, ptrpBG1-PLK, used for the generation of trpEЈ-Јlac translational fusions was described previously (25). The plasmids pHD22 and pHD24, which contain trpEЈ-ЈlacZ translational fusions, were constructed by subcloning the trp leader containing EcoRI-HindIII fragments from pPB22 or pHD1 into the EcoRI-HindIII sites of ptrpBG1-PLK, respectively. The two plasmids, pHD22 and pHD24, were linearized with ScaI and separately integrated into the amyE locus of B. subtilis strain W168 (prototrophic). Transformation was by natural competence (30); selection was for chloramphenicol resistance (5 g/ml). Integration was confirmed by screening for the absence of amylase production by iodine staining (31). The resulting strains, PLBS127 (wild type trp leader, trpEЈ-ЈlacZ) and PLBS129 (SD trp leader, trpEЈ-ЈlacZ), contain both an intact trp operon in the natural chromosomal locus and a trpEЈ-ЈlacZ translational fusion under control of the trp promoter with either a wild type or a mutant trp leader region. For in vivo TrpE protein labeling, pHD15 was transformed into the Escherichia coli T7 overexpression strain K38(pGP1-2) (32) such that expression of plasmid-borne trpE is under control of T7 RNA polymerase. Exclusive labeling of TrpE was carried out by a published procedure (32).
␤-Galactosidase Assay-Cells were cultured in minimal Spizizen salts medium (33) containing 0.2% acid-hydrolyzed casein, 0.2% glucose, and 5 g/ml chloramphenicol in the presence or absence of 50 g/ml tryptophan. Each culture (8 ml) was harvested in mid-exponential phase (Klett 110, green filter number 54) by centrifugation, washed with cold 10 mM Tris-HCl (pH 7.5), and resuspended in 4 ml of Z buffer (34). Samples (0.1 ml) were diluted 10-fold with Z buffer. Ten l of fresh lysozyme solution (10 mg/ml) was added, and the mixtures were incubated for 5 min at 37°C prior to the addition of 10 l of 10% Triton X-100. ␤-Galactosidase activity was subsequently assayed as described (34).
In Vitro Protein Synthesis-TRAP was purified as described previously (20). Preparation of TRAP-deficient S30 extract followed a published procedure (16). The RNA used in this analysis was synthesized in vitro using the Ambion MEGAscript kit and plasmid pHD15 linearized with BamHI as template. Translation reactions (50 l) contained 72 mM Tris-HCl (pH 7.5), 72 mM NH 4 Cl, 10 mM magnesium acetate, 0.1 mM EDTA (pH 7.5), 2.4 mM dithiothreitol, 2 mM ATP, 0.1 mM GTP, 0.08 mM calcium folinate, 0.2 mM diisopropylfluorophosphate, 20 mM phosphoenolpyruvate, 35 units/ml pyruvate kinase, 1 mM L-tryptophan, 0.1 mM concentration of the remaining amino acids (minus methionine), 4 l of S30 extract (30 mg of total protein), 800 units/ml DNase I, 500 units/ml RNasin, 2 pmol of unlabeled transcript, and 10 Ci of [ 35 S]methionine. To reduce endogenous mRNA and DNA, the S30 extract plus DNase I was preincubated for 15 min at 37°C prior to the addition of the remaining components. Concentrations of TRAP used in various reactions are indicated in the appropriate figure legend. Final reaction mixtures were incubated at 37°C for 30 min. Reactions were terminated by the addition of an equal volume of 2ϫ SDS sample buffer (16). Samples (5 l) were heated at 95°C for 3 min and electrophoresed through a 15% SDS-polyacrylamide gel. Radiolabeled protein bands were quantified with a PhosphorImager (Molecular Dynamics, Inc.) and the ImageQuant software package.
Computer Predictions of RNA Secondary Structures-Predictions of RNA secondary structures within the wild type and mutant trp leaders were performed using the MFOLD program (35). 2 Primer Extension Inhibition and Toeprint Analyses-Primer extension inhibition experiments were carried out to map the position of the 3Ј-ends of stable RNA secondary structures. Gel-purified transcripts used in this analysis were synthesized with the Ambion MEGAscript in vitro transcription kit from HindIII-linearized pPB22 (wild type trp leader) or HindIII-linearized pHD2 (SD blocking hairpin mutant) as template. Reaction mixtures (20 l) contained 0.5 pmol of ␥-32 P-endlabeled primer complementary to nucleotides 245-265 relative to the start of trp operon transcription, 0.2 pmol of in vitro generated mRNA, 3 g of TRAP, 1 mM L-tryptophan, and 0.375 mM dNTPs in toeprint buffer (40 mM Tris-HCl, pH 8.0, 200 mM KCl, 4 mM MgCl 2 , 1 mM dithiothreitol). The mixture was incubated at 37°C for 10 min to allow TRAP⅐RNA complex formation and to allow the end-labeled primer and the transcript to anneal. After the addition of 10 units of Moloney murine leukemia virus reverse transcriptase (U.S. Biochemical), incubation was continued at 37°C for 10 min. Samples were extracted with phenol/chloroform followed by ethanol precipitation. Samples were resuspended in 5 l of water followed by the addition of 3 l of standard sequencing stop solution.
The 30 S ribosomal subunit toeprint reactions followed a published procedure (16), except that toeprint buffer was used (see above). The transcripts and end-labeled primer used in the analysis are described above. Samples were fractionated through standard 6% sequencing gels. Control sequencing reactions were carried out with the Sequenase version 2.0 sequencing kit (U.S. Biochemical) using the same plasmids and end-labeled primer described above.
RNA Structure Mapping-The transcripts used in this analysis were synthesized using the Ambion MEGAscript in vitro transcription kit and plasmid pPB22 or pHD2 linearized with HindIII as template. Titrations of RNases and chemical reagents were routinely performed to determine the amount of each reagent that would prevent multiple cleavages or chemical modifications in any one transcript so that we could minimize the potential of secondary rearrangements in short RNA segments. RNA samples were partially digested with RNase T1 (Life Technologies, Inc.) or RNase V1 (Amersham Pharmacia Biotech). Reaction mixtures (0.1 ml) contained 20 pmol (2 g) of TRAP, 1 pmol of transcript and 1 mM L-tryptophan in TKM buffer (40 mM Tris-HCl, pH 8.0, 250 mM KCl, 4 mM MgCl 2 ) (22). TRAP⅐RNA complexes were allowed to form for 10 min at 37°C, at which time 1.5 units of RNase T1 or 1 ϫ 10 Ϫ3 units of RNase V1 was added, and the samples were further incubated for 10 min at 37°C. Samples were immediately extracted with phenol/chloroform, and the RNA was recovered by two successive ethanol precipitations. Chemical modification reactions using DMS or CMCT were performed as described previously (22). TRAP⅐RNA complexes were allowed to form for 10 min at 37°C prior to the addition of 0.5 l of DMS to the mixtures. Following a 4-min incubation at 37°C, reactions were terminated, and the RNA was recovered as described (36). CMCT modification was performed by adding 20 g/ml CMCT (final concentration) and incubating at 37°C for 30 min. Reactions were terminated, and the RNA was recovered as described for DMS modification. RNA pellets were dried and resuspended in primer extension buffer. Hybridization mixtures contained 1 pmol of RNA and 2 pmol of ␥-32 P-end-labeled primer in primer extension buffer (U.S. Biochemical). Mixtures were heated to 80°C for 3 min and immediately placed on ice for 15 min. Following the addition of 0.5 mM dNTPs (final concentra-tion), primer extension was initiated by adding 10 units of Moloney murine leukemia virus reverse transcriptase (5 l final volume). After 10 min at 42°C, reactions were terminated by the addition of 3 l of standard sequencing stop solution. Samples were fractionated through standard 6% sequencing gels. Control sequencing reactions were carried out using the same plasmids and end-labeled primer described above.

TRAP Regulates TrpE Synthesis-Previous in vivo
experiments demonstrated that TRAP can regulate translation of trpE, the first structural gene of the trpEDCFBA operon, approximately 13-fold (19,25). It was also shown that a B. subtilis strain containing several mutations in the trp leader that were predicted to destabilize the SD blocking hairpin (SDtrpL), without altering the SD sequence itself, reduced the ability of TRAP to regulate TrpE synthesis (25). To confirm these in vivo observations, we constructed two B. subtilis strains containing trpEЈ-ЈlacZ translational fusions that were controlled by the wild type (WTtrpL) or SDtrpL trp leader and analyzed ␤-galactosidase expression when each strain was grown in the presence and absence of exogenous tryptophan. We observed minimal expression in the WTtrpL strain PLBS127 grown in the presence of tryptophan (Table I). The effect of exogenous tryptophan on expression of WTtrpL trpEЈ-ЈlacZ can be assessed from the ϪTrp/ϩTrp ratio, which was 345. Note that this ratio reflects both transcriptional and translational regulation. Comparable experiments were performed with the SDtrpL strain PLBS129. In this case the ϪTrp/ϩTrp ratio was only 19, significantly lower than that observed for the strain carrying the wild type trp leader (Table I). Moreover, comparison of ␤-galactosidase expression of the two strains grown in the presence of tryptophan allows us to assess the level of TRAP-mediated translational control. The SDtrpL/WTtrpL ratio of 12.5 is in good agreement with previously published in vivo results (19,25).
To analyze the TRAP-dependent translational regulation of trpE directly, we performed RNA-directed cell-free translation experiments. An in vitro system utilizing a TRAP-deficient B. subtilis S30 extract was used to determine if TRAP binding to trp operon leader RNA inhibits TrpE synthesis. The in vitro synthesized transcripts used in this analysis contained the wild type trp leader and the entire trpE coding sequence. As an in vivo control for the size of TrpE, we carried out exclusive labeling of plasmid encoded TrpE protein (Fig. 1A, lane 1). When we used the RNA-directed cell-free translation system, we observed a major protein species that was the same size as in vivo labeled TrpE (Fig. 1A, lane 2). No in vitro translation products were produced without the addition of trpE RNA (Fig.  1A, lane 11). When the trpE transcript was preincubated with increasing amounts of tryptophan-activated TRAP prior to the addition of the remaining components of the translation system, a corresponding decrease in TrpE translation was observed ( Fig. 1, A (lanes 3-6) and B). When the preincubation step was omitted, the addition of increasing amounts of TRAP to the translation system resulted in a similar decrease in TrpE synthesis ( Fig. 1, A (lanes 7-10) and B). Note that it was not possible to perform the control experiment in which TRAP is added in the absence of tryptophan, since tryptophan is required for TrpE synthesis. When taken together with previously published findings, these results demonstrate that TRAP binding to the (G/U)AG repeats located between nucleotides 36 and 91 of the untranslated trp leader results in a substantial reduction of TrpE synthesis.

RNA Secondary Structures Predicted to Form in trp Operon
Read-through Transcripts-To develop a detailed model of the TRAP-dependent trpE translational regulatory mechanism, we analyzed the RNA structures predicted to form in the leader segment of trp operon read-through transcripts using free energy minimization (35). In this analysis, we included nucleotides 1-210 relative to the start of transcription. The most thermodynamically stable RNA secondary structure predicted to form in the trp leader is shown in Fig. 2A (⌬G 0 ϭ Ϫ37.3 a The values shown are averages of six or more independent experiments Ϯ S.D.

FIG. 1. Regulation of TrpE synthesis by TRAP in cell extracts.
A, the S30 extract used in this analysis was produced from a TRAPdeficient strain of B. subtilis. The TRAP concentration (in M) used in each reaction is indicated at the top of each lane. trpE mRNA was preincubated with increasing amounts of tryptophan-activated TRAP prior to the addition of the remaining components of the in vitro translation system (lanes 3-6). The preincubation step was omitted for the samples corresponding to lanes 7-10. The position of the TrpE polypeptide is shown and is identical in size to the in vivo labeled control (lane 1). TrpE synthesis was not observed in the control sample in which trpE mRNA was omitted (lane 11). B, relative levels of TrpE polypeptide were plotted as a function of the TRAP concentration in each reaction. The amount of TrpE synthesized in the absence of TRAP (Fig. 1A, lane 2) was arbitrarily set to 100. Closed circles correspond to samples that were preincubated with tryptophan-activated TRAP prior to the addition of the remaining components of the in vitro translation system, whereas open circles correspond to samples in which the preincubation step was omitted. kcal/mol). Interestingly, the last six (G/U)AG repeats that comprise the TRAP binding target are contained in the 5Ј-half of the base of this structure, while the first five triplet repeats are predicted to be single-stranded. Moreover, the position of the TRAP binding site suggested that TRAP binding would disrupt the base of this structure. We also analyzed the RNA structures predicted to form if tryptophan-activated TRAP was bound to the (G/U)AG repeats from positions 36 -91 of the trp leader. We determined the structures predicted to form between nucleotides 92 and 210 and also between 1 and 210, except that in this case we removed nucleotides 36 -91 from the analysis. In each case, two secondary structures were predicted to form in the transcript downstream from the TRAP binding site (Fig. 2B). One of these structures consists primarily of the Rho-independent terminator present at the apex of the unbound structure ( Fig. 2A), while another entirely new stem-loop structure contains the trpE SD sequence in the 3Ј-half of the stem (⌬G 0 ϭ Ϫ12.4 kcal/mol). Note that when TRAP is not bound to the transcript, the nucleotides that comprise the 5Ј-half of the SD blocking hairpin would be base-paired with a segment of the TRAP binding target, making these two structures mutually exclusive ( Fig. 2A). Thus, these structures provide the basis for a molecular model that could explain the observed TRAP-dependent regulation of TrpE synthesis. In this model, binding of tryptophan-activated TRAP to its recognition target located between nucleotides 36 and 91 of trp operon read-through transcripts would disrupt the structure predicted to form in the naked RNA. This would allow the trp leader transcript to refold such that the nucleotides that were paired with the TRAP binding site would be able to participate in the formation of an RNA secondary structure that would sequester the trpE SD sequence, ultimately leading to a reduction in TrpE synthesis by preventing ribosome access to the trpE ribosome binding site. Remarkably, TRAP binding would repress trpE translation by altering the conformation of the transcript more than 100 nucleotides downstream from the 3Ј-end of the TRAP binding site.

TRAP-mediated Long Distance Refolding of the trp Leader Transcript Inhibits Ribosome Binding to the trpE Ribosome
Binding Site-To determine if TRAP binding promotes formation of the trpE SD blocking hairpin structure, we performed primer extension inhibition experiments using an in vitro synthesized transcript containing the wild type trp leader. The presence of a bound protein or a stable RNA secondary structure blocks primer extension by reverse transcriptase, resulting in a toeprint band at a position corresponding to the 3Јboundary of the bound protein or at a position near the 3Ј-end of the RNA duplex. A prominent block for reverse transcriptase that we observed in all cases corresponds to the base of the terminator structure, indicating that the terminator is a stable structure and that it is not influenced by the binding of TRAP or 30 S ribosomal subunits (Fig. 3). In addition, a band at position 94 was detected in the presence of bound TRAP but not in its absence (data not shown). This band corresponded to the position of the TRAP toeprint and is in excellent agreement with the known TRAP binding target that ends at position 91. Three prominent RNA structural toeprint bands were also detected at positions 196, 200, and 201 when TRAP was bound to the transcript (Fig. 3, lane 1), which correspond to positions at or near the base of the predicted SD blocking hairpin structure (Fig. 2B). These three bands were very faint when TRAP was not bound to the transcript (Fig. 3, lane 2). In the absence of TRAP binding, three prominent RNA toeprint bands were detected at positions 176, 179, and 180 (Fig. 3, lane 2), which correspond to positions near the base of the secondary structure predicted to form in the trp leader of naked read-through FIG. 2. trpE translational control model. A, under tryptophanlimiting conditions, TRAP is not activated and is unable to bind to the trp leader transcript. In this case, the predicted secondary structure of the trp leader of a trp operon read-through transcript is such that the trpE SD sequence is single-stranded, thereby allowing efficient translation. B, under excess tryptophan conditions, TRAP is activated. When tryptophan-activated TRAP binds to the (G/U)AG repeats present between nucleotides 36 and 91 in the leader of a trp operon read-through transcript, the predicted secondary structure of the downstream RNA is altered such that the trpE SD sequence is sequestered in a stable RNA hairpin. This RNA secondary structure inhibits TrpE synthesis by preventing 30 S ribosomal subunit interaction with the trpE message. The nucleotides between 171 and 184 that are responsible for sequestering the trpE SD sequence are shown in boldface type in both A and B. In the absence of bound TRAP, these nucleotides base-pair with a segment of the TRAP binding site. The (G/U)AG repeats and the trpE AUG initiation codon are indicated in boldface type. Numbering is from the start of transcription. Note that the RNA structures shown in B contain some modifications from those predicted from MFOLD due to the RNA secondary structure mapping data obtained during the course of these studies. transcripts ( Fig. 2A). These three bands were absent when TRAP was bound to the transcript (Fig. 3, lane 1). One additional RNA toeprint was detected at A 214 . Although the secondary structure responsible for this reverse transcriptase stop is not known, it is clear that it can form in the presence or absence of bound TRAP and that it does not prevent ribosome binding (Fig. 3). While it was apparent that the intensities of the bands corresponding to the base of the SD blocking hairpin were significantly reduced in the absence of TRAP, the fact that they were still detectable indicates that the two structures are in equilibrium when TRAP is not bound, although it is clear that the structure shown in Fig. 2A is thermodynamically favored. These results demonstrate that TRAP binding does in fact promote refolding of trp operon read-through transcripts. Moreover, the resulting structure would be capable of sequestering the trpE SD sequence.
Since it is well established that the SD sequence of mRNA base-pairs with the 3Ј-end of the 16 S rRNA present in 30 S ribosomal subunits (1,8), our model predicts that formation of the SD blocking hairpin would interfere with ribosome binding. To test this prediction, we performed a trp leader toeprint analysis using B. subtilis 30 S ribosomal subunits. In the absence of TRAP, we observed a prominent tRNA fMet -dependent toeprint band 15 nucleotides downstream from the first nucleotide in the AUG initiation codon (Fig. 3, lane 5), precisely the distance that was previously observed for the B. subtilis trpG ribosome toeprint (16). In addition to the ribosome toeprint, we observed several bands just below the terminator structure. One likely explanation for the presence of these bands is that binding of 30 S ribosomal subunits alters the large RNA secondary structure shown in Fig. 2A.
We were also interested in determining whether TRAP binding competes with the ability of ribosomes to bind to the trpE message. When TRAP was allowed to bind to the trp leader prior to the addition of 30 S ribosomal subunits, the ribosome toeprint signal was severely reduced (Fig. 3, compare lanes 3  and 5). Thus, as predicted, the ability of TRAP to promote formation of the SD blocking hairpin interferes with ribosome binding. Interestingly, we also observed an appreciable decrease in the signal at positions 196, 200, and 201 (Fig. 3,  compare lanes 1 and 3), suggesting that TRAP and ribosome binding are in competition with each other or that ribosomes are still able to bind to a transcript containing the SD blocking hairpin, albeit at a reduced level. When 30 S ribosomal subunits were added prior to the addition of TRAP, the intensity of the ribosomal toeprint band was reduced, while the intensity of the bands at positions 196, 200, and 201 decreased dramatically (Fig. 3, compare lanes 1, 4, and 5). These results indicate that interaction of the 16 S rRNA with the SD sequence interferes with the ability of TRAP to promote formation of the SD blocking hairpin, again consistent with our trpE translational control model.
Structure of the trpE Shine-Dalgarno Blocking Hairpin-The toeprint results presented above are consistent with the trpE translational control model in which TRAP binding to trp operon read-through transcripts promotes refolding of the trp leader RNA such that a newly formed secondary structure prevents ribosomes from interacting with the trpE SD sequence. To obtain more direct evidence for the TRAP-dependent RNA conformational switch mechanism, we probed the structure of trp leader read-through transcripts in vitro with structure-specific enzymatic and chemical reagents in the presence or absence of bound TRAP. trp operon read-through transcripts were subjected to partial digestion or chemical modification using RNase T1, RNase V1, DMS, or CMCT. The sites of nuclease cleavage or chemical modification were mapped by primer extension using the same end-labeled primer that was used in the toeprint analysis. Cleavage or chemical modification of specific nucleotides would give rise to a primer extension band 1 nucleotide shorter than the corresponding band in the sequencing lane. Thus, the patterns of cleavage or modification provide direct evidence of the trp leader RNA secondary structures that form in the presence or absence of bound TRAP. The results of the structure mapping experiments are shown in Fig.  4 and summarized in Fig. 5. As a control for toeprint bands that are caused by RNA secondary structure blocks to reverse transcriptase, primer extension experiments were performed in the presence or absence of bound TRAP without RNase or chemical treatment (the identical experiment described in Fig. 3, lanes 1  and 2).
The RNase T1 and RNase V1 results were most informative in determining the structure of the SD blocking hairpin. RNase T1 preferentially cleaves following unpaired G residues, whereas RNase V1 cleaves double-stranded RNA. In the absence of TRAP, all of the G residues located in the SD RNA segment between positions G 192 and G 201 were cleaved by RNase T1, whereas cleavage was severely reduced or not detected when TRAP was bound to the transcript, indicating that these residues are base-paired in the presence of bound TRAP but not in its absence (Fig. 4, compare lanes 7 and 8). We also observed enhanced RNase T1 cleavage at position 185 when TRAP was bound to the transcript, indicating that this G residue is unpaired when TRAP is bound to the RNA.
Note that G 185 is in the loop of the SD blocking hairpin (Fig.  5). In addition, the cleavage pattern of G 165 suggests that this residue is paired in the absence of TRAP and that it can be paired or unpaired when TRAP is bound to the transcript. Finally, the RNase T1 cleavage pattern indicated that G 148 and G 140 are base-paired when TRAP is bound to the transcript but not in its absence (data not shown). Thus, all of the RNase T1 data are consistent with the predicted RNA structures (Fig. 2, A and B).
The results obtained with RNase V1 are in good agreement with the RNase T1 (see above) and the DMS and CMCT results (see below). When TRAP was bound to the transcript, appreciable RNase V1 cleavage was detected at positions 174, 175, 179 -183, and 191-205, suggesting that these nucleotides were base-paired (Fig. 4, lane 9, and Fig. 5). The absence of cleavage at positions 177-178 and 185-189 suggests that these nucleotides were unpaired. Note that these residues correspond to the UU bulge and part of the loop of the SD blocking hairpin, respectively (Fig. 5). The RNase V1 cleavage pattern differed dramatically when TRAP was not bound to the transcript (Fig.  4, lane 10). The reduction in cleavage that was observed throughout the SD region suggests that this RNA segment is generally unstructured in the absence of TRAP binding. However, significant cleavage was detected between positions 175 and 183. These residues correspond to positions near the base of the structure predicted to form when TRAP is not bound.
To more precisely determine the structure of the RNA surrounding the trpE SD sequence in the presence and absence of bound TRAP, chemical modification experiments with DMS and CMCT were carried out. DMS methylates N 1 of adenine and N 3 of cytosine when the residues are single-stranded, whereas CMCT modifies unpaired G and U residues at the N 1 and N 3 positions, respectively. DMS-and CMCT-modified residues are unable to serve as templates for reverse transcriptase. The data obtained using these chemical reagents complement the enzymatic cleavage analysis. In the absence of bound TRAP, all of the A and C residues between A 188 and A 208 were modified by DMS, indicating that these residues were single-stranded (Fig. 4, lane 4, and Fig. 5). No appreciable DMS modification was detected between C 173 and C 187 , indicating that, as predicted, this RNA segment is structured when TRAP is not bound (Fig. 2A). Modification of the adenine residues at positions 190, 191, and 198 was significantly reduced when TRAP was bound to the trp leader, indicating that these nucleotides were base-paired (Fig. 4, lane 3, and Fig. 5). In contrast, increased or similar levels of DMS modification were detected at A 168 , C 187 , A 188 , A 194 , C 202 , A 203 , and A 204 , indicating that these residues were unpaired when TRAP was bound. The lack of DMS modification between A 168 and C 187 suggests that this RNA segment is structured when TRAP is bound (Fig.  4, lane 3). Finally, the band at position A 200 is a SD blocking hairpin toeprint band (Fig. 4, compare lanes 1 and 3). Note that with the exception of A 194 , the DMS results are consistent with the structures shown in Figs. 2 and 5. Recall that the RNase V1 results indicated that A 194 was paired when TRAP was bound to the trp leader transcript (Fig. 4, lane 9).
Results from the CMCT experiments were generally consistent with the structural mapping experiments described above and the structure of the SD blocking hairpin (Fig. 5). In the absence of TRAP, we detected CMCT modification at positions U 183 , U 184 , and U 189 , suggesting that these residues were unpaired (Fig. 4, lane 6). However, the low level of CMCT modification of U 183 and U 184 , combined with the finding that RNase V1 digestion resulted in a modest cleavage of these residues, suggests that they can pair in the absence of bound TRAP (Fig. 4, lane 10). The lack of detectable CMCT modifica-  3 and 4) or CMCT (lanes 5 and 6), or cleaved by RNase T1 (lanes 7 and 8) or RNase V1 (lanes 9 and 10) were detected by primer extension as described in the Fig. 3 legend. Lanes 1 and 2 correspond to controls that reveal RNA structural toeprint bands (Fig.  3). Note that the bands observed in lanes 3-10 are 1 nucleotide shorter than the corresponding bands in the A, C, G, or U sequencing lanes. Positions of the trpE SD sequence and the AUG initiation codon are shown. Numbering at the left corresponds to the DNA sequencing ladder and is from the start of transcription. tion of the remaining C and G residues between U 174 and G 185 indicates that these nucleotides were paired. In the presence of bound TRAP, we detected appreciable CMCT modification at residues U 177 , U 178 , and U 189 , indicating that these nucleotides were unpaired (Fig. 4, lane 5). Note that these residues are present within the side bulge (U 177 and U 178 ) and the loop (U 189 ) of the SD blocking hairpin (Fig. 5). We also detected low level modification of residues G 195 and U 196 , indicating that these nucleotides can be unpaired when TRAP is bound. However, the RNase T1 and RNase V1 results are consistent with these residues being paired (Fig. 4, lanes 7 and 9). We also observed a band that would appear to correspond to C 181 . Since CMCT is not useful for detecting cytosine residues, it is possible that this band is an RNA structural toeprint band caused by a secondary rearrangement in the transcript. Finally, the bands at positions 199 and 200 are toeprints of the SD blocking hairpin (Fig. 4, compare lanes 1 and 5). When taken together with the in vitro translation results, the RNA structural studies demonstrate that TRAP binding to the trp leader readthrough transcript is responsible for promoting formation of the trpE SD blocking hairpin, and that formation of this structure decreases TrpE synthesis by interfering with ribosome binding to the trpE SD sequence.
The SDtrpL Mutations Abolish Formation of the Shine-Dalgarno Blocking Hairpin-As mentioned above, we confirmed previously published results (25) that the changes in the SDtrpL transcript reduced the ability of TRAP to regulate TrpE synthesis in vivo (Table I). Computer predictions of the SDtrpL transcript suggested that instead of the SD blocking hairpin, a different secondary structure could form that contained the trpE SD sequence in the loop of the hairpin (structure not shown). To determine if the reduction in translational control could be attributed to the inability of the SD blocking hairpin to form, we performed RNA structural studies on the SDtrpL transcript. We found that the nucleotide substitutions did not alter the RNA structural toeprint between positions 176 and 180, indicating that the large secondary structure could form in the absence of TRAP and that TRAP binding disrupted the structure (Fig. 6, lanes 1 and 2). However, we did not detect any TRAP-dependent RNA toeprint bands that corresponded to the base of the SD blocking hairpin. Instead, we observed a prominent RNA structural toeprint band at position 204 in the presence and absence of bound TRAP. These results indicate that, as predicted, a stem-loop structure can still form in the vicinity of the trpE SD sequence in the SDtrpL transcript; however, in this case formation of the structure was not dependent on TRAP binding.
We also performed a 30 S ribosomal toeprint analysis using the SDtrpL transcript. As was previously seen for the wild type transcript, we observed a prominent ribosomal toeprint signal centered around A 219 (Fig. 6, lane 5). We also found that TRAP binding decreased the intensity of the ribosome toeprint (Fig. 6,  lanes 3-5); however, the reduction was less severe than what was previously observed with the wild type transcript (compare Fig. 3, lanes 3-5, with Fig. 6, lanes 3-5). This finding indicates that while the new TRAP-independent RNA structure has a less pronounced effect on ribosome binding compared with the TRAP-dependent SD blocking hairpin, TRAP binding still affects the ability of 30 S ribosomal subunits to interact with the trpE SD sequence in vitro.
The toeprint results described above suggested that TRAP binding to the SDtrpL transcript would not significantly alter the RNA structure surrounding the trpE SD sequence. To test this hypothesis, we performed RNase T1 and RNase V1 structure mapping experiments on the mutant trp leader. The cleavage patterns indicated that, as predicted, the SD blocking hair-pin was not present in the SDtrpL transcript. Furthermore, we found that TRAP binding had little effect on the structure of the transcript downstream of U 180 (Fig. 7). However, as was previously observed in the RNA structural toeprint analysis (Fig. 6), TRAP binding did eliminate the RNA toeprint signal between A 176 and U 180 , indicating that TRAP binding disrupts the large RNA secondary structure in the mutated trp leader. SDtrpL mapping experiments with DMS and CMCT were consistent with the nuclease results (data not shown). Thus, the in vitro toeprint and structure mapping analyses of the SDtrpL transcript are consistent with the in vivo results, which demonstrated that TRAP-mediated translational control was significantly reduced in the mutated transcript (Table I) (25). DISCUSSION Expression of the B. subtilis trpEDCFBA operon is regulated by TRAP at both the transcriptional and translational levels, while TRAP is only known to regulate trpG expression at the level of translation (17). While it is clear that TRAP-mediated formation of the SD blocking hairpin is responsible for regulating TrpE synthesis, RNA refolding would not be required for this structure to form in all cases. When cells are growing under conditions of tryptophan excess, TRAP would be activated and most likely bind to the message as it is being synthesized. In most cases, this would promote termination in the leader region (transcription attenuation); however, in some instances RNA polymerase will escape termination despite TRAP binding since transcription termination is never 100% efficient. In other situations, TRAP might bind prior to transcription of the trpE SD sequence but not in time to promote termination. Both of these scenarios would result in a TRAPbound read-through transcript that would not require RNA refolding to sequester the trpE SD sequence in the SD blocking hairpin. It is more likely that the TRAP-mediated RNA refold- ing mechanism that we observed in vitro using preexisting read-through transcripts would occur in vivo when cells were initially growing under tryptophan limiting conditions. Under these conditions, a relatively high percentage of TRAP molecules would not be activated, resulting in increased transcriptional read-through. The leader segments of these transcripts would then be able to fold into the structure shown in Fig. 2A, resulting in efficient TrpE synthesis. Eventually, either by synthesis or transport, B. subtilis would build up a sufficient level of tryptophan to activate TRAP. Tryptophan-activated TRAP would then bind to the trp leader and promote RNA refolding and formation of the SD blocking hairpin, ultimately leading to a reduction in trpE translation. It should also be pointed out that all of the coding sequences within the trpED-CFBA operon overlap by several nucleotides except for trpC and trpF. However, in this case the two coding sequences are still only separated by 4 bases (29). This gene organization suggests that translational coupling plays a role in trp operon expression. Thus, TRAP-mediated formation of the trpE SD blocking hairpin may reduce translation of every message within the polycistronic transcript. Indeed, preliminary results suggest that TrpD synthesis is regulated by formation of the trpE SD blocking hairpin. 3 It is also likely that translational inhibition will lead to decreased message stability of the trp operon transcript, since a reduction of ribosome density on the mRNA would probably result in increased nucleolytic attack. In addition, it is possible that translational inhibition of trpE could lead to transcriptional polarity by allowing increased access of Rho termination factor.
While all of these mechanisms may contribute to control trp operon expression, the results of our in vitro study demonstrate that TRAP has the ability to regulates TrpE synthesis by promoting RNA refolding. Our cell-free translation experiments demonstrated that increasing levels of TRAP resulted in a corresponding decrease in TrpE synthesis (Fig. 1). Computer predictions suggested that TRAP binding to nucleotides 36 -91 of trp operon read-through transcripts would disrupt the base of a thermodynamically favored RNA structure by virtue of the fact that 6 of the 11 (G/U)AG repeats that comprise the TRAP binding target are present in the 5Ј-half of the structure. Since the first five triplet repeats are predicted to be single-stranded and it is known that TRAP is specific for single-stranded RNA (37), it is likely that disruption of the structure occurs by a mechanism in which TRAP initially binds to the first five repeats and then subsequently interacts with the repeats present in the secondary structure, perhaps due to breathing of the imperfect stem. Computer predictions further suggested that once TRAP was bound, the nucleotides between positions 171 and 184 would be available to participate in the formation of a new RNA hairpin that would sequester the trpE SD sequence in the stem of the structure and thereby block ribosome access to the trpE ribosome binding site (Fig. 2). Thus, this mechanism could account for at least some of the observed TRAP-dependent reduction in TrpE synthesis that was observed in vivo (Table I) (19,25) and in vitro (Fig. 1). Results from primer extension inhibition experiments demonstrated that TRAP binding does promote refolding of trp leader transcripts and that the resulting structure inhibits ribosome binding (Fig. 3). Moreover, results from RNA structure mapping experiments demonstrate that the TRAP-dependent SD blocking hairpin contains the SD sequence in the 3Ј-half of the stem (Figs. 4 and 5). A few discrepancies exist in the structure mapping data when one compares the RNase V1 results with those for DMS, CMCT, and RNase T1. However, RNase V1 cleavage does not occur at every paired residue, and in addition to cleaving nucleotides in an RNA duplex, RNase V1 can cleave the first few bases in a single-stranded RNA segment that is adjacent to an RNA duplex as well as in single-stranded segments in which the nucleotides are stacked (38). Thus, sometimes results from RNase V1 cleavage are not entirely straightforward.
A trpEЈ-ЈlacZ translational fusion containing several changes in the trp leader predicted to destabilize the SD blocking hairpin without altering the SD sequence itself was described previously (25). It was determined that these changes reduced the ability of TRAP to regulate trpE translation (Table  I) (25). Our RNA structural studies reveal the structural basis for this observation. Instead of the TRAP-dependent SD blocking hairpin, a TRAP-independent structure can form in the vicinity of the trpE SD sequence (Figs. 6 and 7). Despite the finding that formation of this structure occurs in the presence or absence of bound TRAP, we found that TRAP binding resulted in a modest reduction in ribosome binding (Fig. 6), presumably due to TRAP-dependent stabilization of the structure. Thus, the approximate 13-fold translational regulation that was observed in vivo should be viewed as a lower limit of translational control (Table I) (19,25).
In prokaryotes, the ability to shut off translation of particular transcripts in response to environmental signals allows the organism to rapidly divert specific compounds into the synthesis of other important molecules and would also conserve energy by preventing the synthesis of proteins that are no longer required for growth. This appears to be the case for the trp operon of B. subtilis. Since anthranilate synthase, the enzyme responsible for catalyzing the initial biochemical step specific to tryptophan biosynthesis, is a complex of TrpE and TrpG polypeptides (39), blocking translation of trpE when a sufficient level of tryptophan is present in the cell provides a rapid response to changing tryptophan levels. This allows for more efficient utilization of chorismic acid in the synthesis of phenylalanine, tyrosine, and folic acid (40,41 3 and 4) were detected by primer extension as described in the Fig. 3 legend. Note that the bands observed in lanes 1-4 are 1 nucleotide shorter than the corresponding bands in the A, C, G, or U sequencing lanes. Positions of the trpE SD sequence and the AUG initiation codon are shown. Numbering at the left corresponds to the DNA sequencing ladder and is from the start of transcription.
TrpE activity and is another mechanism allowing the bacterium to sense the level of tryptophan in the cell.
The diversity of translational control mechanisms illustrates the importance of regulating protein synthesis. A few examples exist in which translation of particular genes is controlled by binding of a regulatory protein to the gene's SD sequence. In these cases, the RNA-binding protein directly blocks ribosome access to the respective ribosome binding site (13,14,16). Numerous examples also exist in which RNA secondary structures are responsible for regulating translation by sequestering the respective SD sequence (for reviews, see Refs. 1, 3, 5, 9, and 12). For example, translation of the IS10 transposase mRNA (9) and the bacteriophage MS2 maturase gene (12) is controlled by formation of RNA structures that block ribosome binding. Interestingly, in both of these mechanisms the kinetics of RNA folding rather than RNA-binding proteins are important for the observed regulation. In addition, it was recently shown that expression of the E. coli gnd gene is regulated at the translational level by a long range interaction between the gnd ribosome binding site and an internal complementary sequence lying between codons 71 and 74 of the gnd mRNA. Again, it does not appear that a protein factor is involved in this long range interaction (11). Another interesting translational control mechanism was identified for the bacteriophage Mu mom gene. In this case, an RNA hairpin that sequesters a portion of the mom SD sequence is disrupted when the Com protein binds just upstream of the secondary structure. Thus, Com protein serves as a translational activator by altering the conformation of the RNA surrounding the mom SD sequence such that ribosomes can gain easier access to the mom ribosome binding site (36,42). It has been proposed that translation of the E. coli S10 operon is regulated by a mechanism in which binding of L4 to the leader segment of the nascent S10 operon transcript promotes formation of an RNA structure that prevents ribosome binding, whereas a translationally active conformation forms in the RNA in the absence of L4 binding. However, further experimentation will be required to substantiate this mechanism, since the precise binding site for L4 is not yet known (5). A translational regulatory mechanism has been proposed for the E. coli L10 operon that is remarkably similar to the mechanism that we demonstrated for the B. subtilis trp operon (43). In this case, it is thought that binding of the L10-(L12) 4 complex to the untranslated L10 operon leader promotes sequestration of the L10 SD sequence in a stable secondary structure. However, mRNA structural studies have failed to confirm the predicted structural switch (44). While the trpE translational control mechanism is the first example in which an RNAbinding protein was found to promote refolding of the transcript to sequester a particular SD sequence, it is reasonable to speculate that this will prove to be a common regulatory mechanism employed by many bacterial species. Indeed, it is quite possible that this regulatory strategy has been overlooked in many situations, since in the case of trpE of B. subtilis, the SD sequence is more than 100 nucleotides downstream from the 3Ј-end of the TRAP binding site.
It is also reasonable to speculate that protein-mediated RNA refolding will regulate eukaryotic translation as well. Translation initiation of the majority of eukaryotic mRNAs occurs via a cap-dependent ribosomal scanning mechanism (2). Initiation requires recognition of the mRNA 5Ј-cap by eukaryotic initiation factor 4, loading of a 40 S ribosomal subunit, and unwinding of RNA secondary structure. Scanning of 40 S subunits is dependent on melting the 5Ј-UTR secondary structure. It is well documented that extensive secondary structure in the 5Ј-UTR inhibits translation initiation both in vivo and in vitro (45,46). A few cellular, picornavirus and other viral mRNAs initiate translation by a cap-independent internal initiation mechanism (47). Internal initiation is directed by the binding of ribosomes to an internal ribosome entry site element within the 5Ј-UTR. The internal ribosome entry site elements are organized in highly conserved stem-loop structures, which are absolutely critical for its function (48). Thus, in eukaryotes, translation could be regulated by protein-mediated refolding of RNA by altering the RNA structure surrounding internal ribosome entry site elements or by interfering with the cap-dependent ribosomal scanning mechanism. It is also plausible that protein-mediated RNA refolding will be responsible for altering the stability of many prokaryotic and eukaryotic mRNAs by either creating or eliminating recognition targets for endonucleases or by creating or eliminating barriers to exonucleases.