The t rp RNA-binding Attenuation Protein (TRAP) from Bacillus subtilis Binds to Unstackedtrp Leader RNA*

TRAP ( t rp RNA-binding attenuation protein) is a tryptophan-activated RNA-binding protein that regulates expression of the trp biosynthetic genes by binding to a series of GAG and UAG trinucleotide repeats generally separated by two or three spacer nucleotides. Previously, we showed that TRAP contains 11 identical subunits arranged in a symmetrical ring. Based on this structure, we proposed a model for the TRAP·RNA interaction where the RNA wraps around the protein with each repeat of the RNA contacting one or a combination of two adjacent subunits of the TRAP oligomer. Here, we have shown that RNAs selected in vitro based on their ability to bind tryptophan-activated TRAP contain multiple G/UAG repeats and show a strong bias for pyrimidines as the spacer nucleotides between these repeats. The affinity of the TRAP·RNA interaction displays a nonlinear temperature dependence, increasing between 5 °C and 47 °C and then decreasing from 47 °C to 67 °C. Differential scanning calorimetry and circular dichroism spectroscopy demonstrate that TRAP is highly thermostable with few detectable changes in the structure between 25 °C and 70 °C, suggesting that the temperature dependence of this interaction reflects changes in the RNA. Results from circular dichroism and UV absorbance spectroscopy support this hypothesis, demonstrating thattrp leader RNA becomes unstacked upon binding TRAP. We propose that the bias toward pyrimidines in the spacer nucleotides of the in vitro selected RNAs represents the inability of Us and Cs to form stable base stacking interactions, which allows the flexibility needed for the RNA to wrap around the TRAP oligomer.

The tryptophan biosynthetic genes of Bacillus subtilis are negatively regulated in response to the intracellular level of L-tryptophan by a tryptophan-activated RNA-binding protein called TRAP 1 (trp RNA-binding attenuation protein; Ref. 1). TRAP regulates expression of the trp biosynthetic genes by binding to several RNA targets in a tryptophan-dependent manner. Two RNA binding sites for TRAP have been characterized, one located within the 204-nucleotide leader region preceding trpE (2)(3)(4)(5) and the second in a segment of RNA overlapping the ribosome binding site of trpG (6). 2 In both cases, TRAP binds to a series of GAG and UAG trinucleotide repeats generally separated by two or three spacer nucleotides (4,5,7). Recent experiments using synthetic RNAs have demonstrated that these trinucleotide repeats are crucial for TRAP binding, with GAG repeats being favored over UAG repeats (7,8). This work also indicated that a spacer of two nucleotides is optimal for TRAP binding and that spacers containing A or U residues are preferred over Gs or Cs.
The crystal structure of TRAP complexed with L-tryptophan has recently been solved and refined to 1.8-Å resolution (5). TRAP is an 11-mer of identical subunits arranged in a symmetrical ring. The secondary structure of TRAP is made up entirely of ␤-strands, ␤-turns, and random coils, which assemble into a novel quaternary structural arrangement consisting of 11 7-stranded antiparallel ␤-sheets. Each of these ␤-sheets contains four ␤-strands from one subunit and three ␤-strands from the adjacent subunit. This structural arrangement generates an extensive subunit-subunit interface, which is a major stabilizing force of the oligomeric structure.
Based on the multiple repeats in the TRAP binding sites and the 11 subunits of the TRAP oligomer, we have proposed that RNA binds to TRAP by wrapping around the protein, with each trinucleotide repeat of the RNA contacting one, or a combination of two adjacent subunits of TRAP (5). We have previously shown that the affinity of TRAP for an RNA containing residues 36 -92 of the trp leader (RNA 36 -92) increased linearly as the temperature was raised from 5°C to 47°C (9). This result implied a strong unfavorable enthalpic contribution (ϩ⌬H) to the binding free energy. Furthermore, the stability of the RNA⅐TRAP complex was found to be virtually insensitive to changes in ionic strength between 100 mM and 700 mM potassium glutamate, indicating that release of ions upon complex formation is not the driving force of the interaction.
In this report, we continue the analysis of the RNA binding site for TRAP using in vitro selection (SELEX, for systematic evolution of ligand by exponential enrichment (10) of RNAs that bound specifically to tryptophan-activated TRAP from a pool of RNAs containing 25 positions with random nucleotides. All RNAs selected contained multiple GAG or UAG repeats, confirming the importance of these trinucleotide motifs. Surprisingly, Ͼ90% of the spacer nucleotides of the RNAs selected were pyrimidines (Us and Cs). In view of our model for the TRAP⅐RNA interaction, one possible explanation for the bias toward pyrimidines in the spacers might be their lesser ability to form base stacking interactions relative to purines (11). If so, this would suggest that base stacking interactions play a role in the TRAP⅐RNA interaction. In agreement with this hypothesis, we have shown that RNA 36 -92 is stacked in solution and becomes unstacked when bound to TRAP. Further analysis showed that the temperature dependence of the TRAP⅐RNA 36 -92 interaction is nonlinear, with a maximum at 47°C. Since we have shown that the structure of the TRAP oligomer is stable at temperatures up to 67°C, the nonlinear temperature dependence of the TRAP⅐RNA interaction likely reflects changes in the RNA structure at higher temperatures (i.e. unstacking of the bases). Therefore, we propose that the spacer nucleotides in the TRAP binding site play two roles: to allow the proper spacing of the trinucleotide repeats to interact with the binding sites on TRAP, and to maintain the flexibility needed for the trp leader RNA to wrap around the TRAP oligomer.

EXPERIMENTAL PROCEDURES
Plasmids and Strains-Plasmids used in this study were propagated in Escherichia coli JM107. Plasmid pTZ18U36 -92 has been described previously (9). Plasmid pTZ18U2-64 contains residues ϩ2 to ϩ64 of the B. subtilis trp leader cloned downstream of the T7 promoter in pTZ18U (U. S. Biochemical Corp.). The insert was created by PCR using the Ϫ20 Universal primer and trpL64 (5Ј-GCAAGCTTACCCTATTCTCTAACT-CAAC-3Ј) as primers, and pTZ18UdR4 (12) as template to generate a product containing bases 2-64 of the trp leader flanked by EcoRI and HindIII sites. Plasmid pCB1, which contains 11 repeats of the sequence 5Ј-GAGAA-3Ј under control of a T7 promoter, was designed based on work described previously (13) and generated as follows. Two oligonucleotides, T7 oligo (5Ј-AGGAATTCAATTATAATACGACTCACTATA-3Ј) and GAGAA 11 oligo (5Ј-GAGGGCCC(TTCTC) 11 TATAGTGAGTCG-3Ј), which overlap by 12 bases (in bold), were annealed and the 3Ј-ends extended with the Klenow fragment of DNA polymerase I (14). The duplex DNA was digested with EcoRI and Bsp120I and ligated into similarly cut pSP64/p36T (13). Plasmids pUC118GAGUU and pUC118GAGAU contain 11 repeats of the sequence 5Ј-GAGUU-3Ј or 10 repeats of the sequence 5Ј-GAGAU-3Ј followed by 5Ј-GAGA-3Ј, respectively, cloned downstream of a T7 promoter. Each plasmid contains a fusion of the desired transcript to the hammerhead ribozyme (pUC118GAGUU) or a hairpin ribozyme (pUC118GAGAU), and both were modeled on the system described in Price et al. (15).
RNA Synthesis and TRAP Purification-The in vitro synthesis of 32 P-labeled trp leader RNA using T7 RNA polymerase has been described previously (9). To generate an RNA containing the sequence 5Ј-GAGAA-3Ј repeated 11 times ((GAGAA) 11 ), Bsp120I-linearized pCB1 was used for in vitro transcription (9). 32 P-labeled RNAs containing 11 repeats of 5Ј-GAGUU-3Ј (GAGUU) 11 ) or 10 repeats of GAGAU followed by 5Ј-GAGA-3Ј [(GAGAU) 11 ] were transcribed from BamHI-linearized pUC118GAGUU or pUC118GAGAU respectively (16). In each case, following transcription, MgCl 2 was added to a final concentration of 50 mM, and the reaction was heated to 80°C for 15 min. The mixture was then incubated at 37°C for 15 min to allow ribozyme cleavage to generate the 55-base TRAP-binding RNA and the 75-base ribozyme. All RNAs were gel-purified as described previously (9), and unlabeled RNAs were quantitated using extinction coefficients at 260 nm, based on the method of Puglisi and Tinoco (17). TRAP was purified as described previously (18) and quantitated using the extinction coefficient of 1280 M Ϫ1 cm Ϫ1 at 280 nm (12).
Filter Binding Analysis-Quantitative filter binding analyses were performed as described previously (9). For thermostability studies, aliquots of TRAP were incubated at 80°C, 85°C, or 90°C for 15 min and then cooled on ice prior to incubation with the RNA. To study the thermodynamics of the TRAP⅐RNA interaction, reaction mixtures were incubated at the desired temperature for 15 min prior to filtration. In all cases, data shown are the average of at least two experiments with standard deviations of Ͻ10% of the mean.
In Vitro Selection of RNAs That Specifically Bind TRAP-SELEX was performed as outlined by Tuerk (19) using an oligonucleotide pool containing 25 random positions (25N; Ref. 20). A pool size of 25N was chosen because it represents the size limit (4 25 ) that can be well represented by a reasonable amount of synthetic DNA (2 nmol). Moreover, we found that the observed dissociation constant (K obs ) for TRAP and an RNA containing the first five G/UAG repeats of the trp leader transcript (RNA 2-64) is 100 nM (data not shown). Since 25 random positions can accommodate five trinucleotide repeats each separated by two nucleotides, we reasoned it should be possible to efficiently select such RNAs from this pool.
The DNA template for SELEX contained 25 random bases flanked by a 5Ј fixed region containing a T7 promoter and a 3Ј fixed region to enable reverse transcription of the RNA (20). Two oligonucleotide primers (3.33 nmol each), which anneal to the 5Ј fixed region (T7Univ) and the 3Ј fixed region (RevUniv) were used for PCR with 500 pmol of oligonucleotide template (N25) in 50 mM KCl, 10 mM Tris, pH 8, 7.5 mM MgCl 2 , 2 mg/ml gelatin, and 75 units of Taq polymerase. Following an initial denaturation step at 93°C for 3 min, reactions were cycled 20 times, with each cycle consisting of 93°C for 30 s, 55°C for 10 s, and 72°C for 1 min. Transcription reactions using T7 RNA polymerase were performed as described by Tuerk (19). The reaction was then treated with 5 units of RQ1 DNase for 15 min to remove the DNA template, and the RNAs were purified on 10% denaturing polyacrylamide gels. The initial population of RNA contained approximately 500 pmol of random sequences.
RNAs that bound to TRAP were selected by filter binding as described by Tuerk (19). The TRAP concentration was 666 nM in the first two rounds, 333 nM in the subsequent five rounds, and then successively reduced in each of the subsequent rounds to 1.5 nM in the 12th round. The RNA concentration was 6 mM in the first two rounds, 3 mM in the next two rounds, and then successively reduced during further rounds to 0.75 nM in the 12th round. TRAP binding RNAs were eluted from the nitrocellulose filters as described previously (19). Eluted RNAs were reverse transcribed in 50-l reactions containing 0.5 mM dNTPs, 1 M RevUniv primer, and 500 units of reverse transcriptase (Life Technologies, Inc.). To eliminate nonspecific RNAs, counter-selections were performed before the 1st, 3rd, and 6th rounds. Counter-selections were done exactly as the binding reactions except in the absence of tryptophan. RNAs that passed through the filter were collected and used for further rounds of selection.
Differential Scanning Calorimetry-Differential scanning calorimetry analyses were performed on a Nano differential scanning calorimeter (Applied Thermodynamics) at a scan rate of 1°C/min at 2.7 atmospheres. TRAP was dialyzed into FBB (250 mM potassium glutamate, 16 mM HEPES, pH 8, 4 mM MgCl 2 , 100 ng/l yeast tRNA) and diluted to 1 mg/ml. For experiments involving L-tryptophan, both the sample and reference contained 1 mM L-tryptophan.
Circular Dichroism-Circular dichroism (CD) spectra were obtained on a JASCO model J-500C spectropolarimeter. TRAP was dialyzed into 50 mM potassium phosphate (pH 7.4) and diluted to 1 mg/ml. Spectra were collected using 1-mm pathlength quartz cuvettes. For experiments at elevated temperatures, TRAP was equilibrated at 70°C for 15 min prior to data acquisition. CD studies of trp leader RNA were performed using 1-cm pathlength cuvettes with 5 M RNA 36 -92. For RNA experiments containing TRAP and/or tryptophan, 5 M TRAP and/or 50 M L-tryptophan were included. Neither TRAP nor tryptophan show significant dichroism in this range (230 -300 nm) and therefore make no contribution to the RNA spectra.
Analysis of RNA 36 -92 Denaturation-Temperature-dependent UV absorption data of trp leader RNA 36 -92 (320 mM in FBB) were collected using a Beckman DU-640 spectrophotometer with a circulating water bath at a heating rate of 1°C min Ϫ1 . All data were normalized to the absorbance at 15°C.

RNA Binding
Site for TRAP-Two natural TRAP binding sites have been characterized and each contains multiple G/UAG trinucleotide repeats generally separated by two or three nucleotide spacers (4 -5). Both of these sites must not only allow for efficient TRAP binding but must also function for proper gene regulation. To examine the sequence requirements of the RNA target of TRAP in the absence of this other selective influence, we used SELEX (10). RNAs that specifically bound to tryptophan-activated TRAP were selected from a pool containing 25 random positions. To follow the selection process, we measured the affinity of the selected RNA pools for TRAP using filter binding (9). We were unable to detect specific binding of the original random population of RNAs using TRAP concentrations up to 2 M. By the fourth round of selection, the observed dissociation constant (K obs ) was 40 nM, which improved to 20 nM in the 6th round and 10 nM in both the 10th and 12th rounds. Since no improvement was seen between the 10th and 12th rounds, selections were stopped after 12 rounds.
Fifteen cDNA clones were sequenced from the 4th round of in vitro selection, 14 from both the 6th and 10th rounds, and 11 from the 12th round. All RNAs selected contained between 3 and 6 repeats/RNA (those with 6 repeats involved contributions from fixed flanking sequences). No RNAs were found that spe-cifically bound to TRAP in a tryptophan-dependent manner, which contained less than three G/UAG repeats. The average number of G/UAG repeats per RNA was 3.6 in the 4th round RNAs, which increased to 4.5/RNA in the 6th round and 4.6/ RNA in the 10th and 12th rounds. This finding is consistent with the improved affinity of the RNA pools from the later rounds for TRAP. Since there were no significant differences observed in the composition of the RNAs from the 6th, 10th and 12th rounds, further analyses were done on the combined set of 39 RNAs from these rounds ( Table I). The sequences of 10 representative clones from this pool are shown in Fig. 1.
The ratio of GAG to UAG repeats in the combined set of selected RNAs was 1.8:1 (Table I), which is consistent with previous findings indicating that TRAP binds GAG repeats better than UAGs (7,8). In addition, Clone 87, which contains 6 UAG repeats, bound TRAP with a K obs of 12 nM, whereas a clone containing 4 GAG repeats (Clone 83) bound TRAP with a K obs of 3 nM (Fig. 1), which again shows that GAG repeats bind TRAP better than UAGs.
We also analyzed the length and composition of the spacer nucleotides separating the trinucleotide repeats (Table I). Most (89%) of the G/UAG repeats were separated by 2 nucleotides, which is similar to the spacing seen in the natural TRAP binding sites and is consistent with other studies indicating that 2 nucleotide spacers are optimal (7). Although the exact sequence of the spacers does not appear to be critical, there was a great bias (88%) for pyrimidines in the spacers (Table I).
Based on our model of the TRAP⅐RNA interaction, with each repeat of the RNA interacting with one subunit of TRAP, it could be possible for more than one of these 3-6-trinucleotide repeat-containing RNAs to bind to a single TRAP oligomer. Therefore, we were concerned as to whether each RNA was selected based on individually binding to TRAP or whether there were synergistic effects between RNAs. To address this concern, we tested whether the ability of the weakest binding RNA selected, Clone 3 ( Fig. 1; K obs ϭ 300 nM), to bind TRAP could be influenced by the presence of other TRAP binding RNAs. In the presence of 300 nM TRAP, approximately 50% of Clone 3 RNA bound TRAP in our filter binding assay. Adding increasing amounts of RNA from the 6th round pool (from which Clone 3 derived) did not act synergistically to increase Clone 3 binding to TRAP. Instead, these RNAs only competed for TRAP binding. This result suggests that the RNAs were selected based on individual abilities to bind TRAP.
We also addressed the question of whether more than one RNA can bind to a single TRAP 11-mer using mobility shift gels. Several G/UAG-containing RNAs (such as Clone 3 and Clone 83; Fig. 1) formed single complexes with TRAP that had different, distinct mobilities in native gels (data not shown). When these RNAs were mixed together in various ratios in the presence of tryptophan-activated TRAP, no additional complexes were seen (data not shown). Although not conclusive, these results suggest that only one RNA can bind TRAP at a time, and may mean that there is a mechanism by which after one RNA binds TRAP, a second RNA molecule is excluded.
Temperature-dependent RNA Unstacking-One of the most interesting results from our in vitro selection experiments was the strong selection for pyrimidines in the spacer nucleotides between the G/UAG repeats. A major difference between pyrimidine and purine bases is that pyrimidines do not stack as well as purines (11). In view of the suggestion that upon binding the RNA wraps around the TRAP oligomer, we postulated that this bias for pyrimidine spacer nucleotides reflects a need for flexibility to wrap the RNA around TRAP.
To examine the base stacking interactions present in a natural TRAP binding site, we examined the temperature-dependent hyperchromicity of an RNA containing bases ϩ36 to ϩ92 of the trp leader (RNA 36 -92; Ref. 9). RNA 36 -92 was studied since it contains all 11 G/UAG repeats but is not predicted to form any significant secondary structures (9), which have been shown to interfere with TRAP binding (7). When the UV absorbance of RNA 36 -92 was monitored as a function of temperature, a broad hyperchromatic transition was observed ( Fig.  2A). For comparison, we examined thermal denaturation of poly(A) (Fig. 2A), a process that has been well characterized as indicative of the disruption of base stacking interactions (21). Thermal denaturation of poly(A) also displays a broad noncooperative transition, similar to that observed for RNA 36 -92. Therefore we conclude that the observed hyperchromatic shift of RNA 36 -92 is likely due to disruption of base stacking interactions.
We also examined base stacking in RNA 36 -92 by circular dichroism (CD) spectroscopy (Fig. 2B). The presence of the strong maximum centered around 265 nm in the CD spectrum of RNA 36 -92 at 25°is indicative of base stacking in singlestranded nucleic acids (22). The intensity of this maximum decreased when the spectrum was obtained at 80°C (Fig. 2B), indicating that base stacking interactions are being disrupted. Together with the UV hyperchromicity results described above, these data demonstrate that RNA 36 -92 is stacked in solution and does not appear to contain significant secondary structure.
Role of Base Stacking on the TRAP⅐RNA Interaction-The model in which RNA wraps around TRAP implies significant structural changes in the RNA upon binding to TRAP. To investigate these changes, we examined the CD spectra of RNA 36 -92 before and after binding to TRAP (Fig. 3). When an equimolar amount of tryptophan-activated TRAP was added to RNA 36 -92, the intensity of the maximum at 265 nm decreased by over 60% (Fig. 3), demonstrating the bases in the RNA become unstacked. This effect is dependent upon the RNA binding to TRAP, since addition of neither apoTRAP nor tryptophan alone induced significant changes in the RNA spectrum (data not shown). The absolute amount of this decrease is greater than that seen in thermally denatured RNA (Fig. 2B), suggesting that when bound to TRAP, the bases are separated by a greater distance than when the RNA is thermally dena- a Spacers are defined as the nucleotides between two adjacent trinucleotide repeats. nt, nucleotides. b 39 cDNA clones were analyzed from rounds 6, 10, and 12. These contained a total of 181 repeats consisting of 116 GAG, 64 UAG, and 1 CAG yielding an average of 4.64 trinucleotide repeats per RNA.
c Natural binding sites include the trp leader RNA and the sequences surrounding the trpG ribosome binding site. The trpG binding site also contains 1 AAG. d In calculating the ratio of GAG versus UAG repeats, repeats where the U in the first position came from nonrandom flanking sequence (such as clones 87 and 154) were not considered.
tured. In addition, a hyperchromatic shift in the UV spectrum of RNA 36 -92 occurs upon binding to TRAP (data not shown), and the magnitude of this shift was 3-fold greater than that observed for thermally denatured RNA.
Effects of Temperatures on the TRAP⅐trp Leader RNA Interaction-In light of the indication that base stacking interactions play a role in the TRAP⅐RNA interaction, we examined the affinity of TRAP for RNA 36 -92 at temperatures up to 67°C (Fig. 4). As we have shown previously (9), the affinity of TRAP for RNA 36 -92 increased as the temperature was raised from 5°C to 47°C with K obs decreasing from 3.6 nM to 0.07 nM (Fig. 4). As the temperature was raised above 47°C, the affinity of TRAP for trp leader RNA gradually decreased; however, at 67°C K obs was still 1 nM.
The decrease in affinity of the RNA⅐TRAP interaction at temperatures above 47°C could be due to reduced affinity of TRAP for L-tryptophan at these temperatures, resulting in TRAP not being fully activated to bind RNA. TRAP binds L-tryptophan cooperatively with an S 0.5 of 5 M at 37°C and 100 M at 60°C (data not shown), suggesting that TRAP should be fully activated in our assays containing 1 mM tryptophan. Moreover, TRAP had identical affinity for RNA 36 -92 in the presence of 1 mM or 10 mM L-tryptophan at 52°C and 67°C, confirming that tryptophan is not limiting at these temperatures (data not shown).
A second possible explanation for the decrease in the affinity for RNA 36 -92 at temperatures above 47°C could be that TRAP is denaturing at these temperatures. To test this possibility, we examined the thermostability of TRAP. Heating TRAP at temperatures up to 90°C for 15 min followed by rapid cooling had no effect on its ability to bind RNA (data not shown). These results indicate that TRAP is either stable at 90°C or is undergoing a denaturation event that is fully reversible during the filter binding reaction. We also used circular dichroism spectroscopy (CD) to examine the effect of temperature on the secondary structure of TRAP. Consistent with FIG. 1. Representative RNAs selected based on their ability to bind tryptophan-activated TRAP in vitro. The 25 nucleotides in the random region are capitalized. The round of selection from which the RNA was selected is indicated. Clone 87 has 26 nucleotides in the random region, possibly due to a PCR mutation. The trinucleotide repeats involved in TRAP binding are underlined and bold, and the spacer nucleotides are italicized. The apparent dissociation constant (K obs ) for the binding of three of these RNAs to TRAP is shown. For comparison, K obs for trp leader RNA 36 -92, which contains 11 G/UAG repeats, is 0.25 nM.

FIG. 2. Thermal denaturing of trp leader RNA.
A, the UV absorbance at 260 nm of trp leader RNA 36 -92 (q) or poly(A) (E) was monitored as a function of temperature between 15°C and 75°C. Experiments were done in FBB, and absorbance readings were taken at every degree. The y axis is the absorbance of the RNA normalized to the A 260 at 15°C. All experiments were done in duplicate, and a representative curve is shown. B, near UV circular dichroism spectra of 5 M RNA 36 -92 at 25°C and 80°C. Data were collected at every nanometer between 235 nm and 300 nm. The x axis is wavelength in nanometers, and the Y axis is millidegrees of rotation. the high percentage of ␤-strands (Ͼ60%) seen in the crystal structure (5), the CD spectrum of TRAP displays a strong minimum at 215 nm (Fig. 5). The spectrum remained virtually unchanged at temperatures up to 70°C (Fig. 5), indicating that the secondary structure of TRAP does not change significantly within this temperature range.
To characterize further the effects of temperature on TRAP, we used differential scanning calorimetry (Fig. 6). Between 15°C and 70°C, no significant thermodynamic changes were seen in TRAP either in the absence or presence of 1 mM Ltryptophan. The difference in the absolute heat capacities between the two samples represents the heat capacity of tryptophan. Above 70°C, two major thermal transitions occur. The first has a midpoint of 90°C in both samples and is reversible, as judged by scanning a sample that was previously heated to 95°C (data not shown). The second transition has a midpoint of 103°C in the absence of tryptophan and Ͼ105°C in the presence of tryptophan. TRAP is irreversibly denatured at these temperatures. These experiments suggest there is little change in the structure of TRAP between 15°C and 70°C. Therefore, it does not appear that the decreased affinity of TRAP for trp leader RNA at temperatures between 47°C and 67°C can be attributed to thermal denaturation of TRAP.
Role of the Spacer Nucleotides in TRAP Binding-Previous studies have suggested that the residues separating the trinucleotide repeats in the TRAP binding site do not make specific contacts with the protein in the TRAP⅐RNA complex (4, 7) but instead are important for proper spacing of the GAG and UAG repeats (4,5,8). To test this hypothesis, we generated several RNAs containing artificial TRAP binding sites consisting of 11 GAG repeats separated by various two nucleotide spacers. We characterized the affinities of these RNAs for TRAP at temperatures between 37°C and 62°C. The first RNA (GAGUU) 11 contained UU spacers, the second (GAGAA) 11 had AA spacers, and the third RNA (GAGAU) 11 contained AU spacers. The temperature dependence of TRAP binding to each of these RNAs was found to be nonlinear between 37°C to 62°C (data not shown). Direct comparison of the affinities of these RNAs for TRAP at temperatures below 47°C was not possible since (GAGUU) 11 forms a stable secondary structures at these temperatures, which has been shown to interfere with TRAP bind-ing (7), while neither (GAGAA) 11 nor (GAGAU) 11 appear to form any significant secondary structures at these temperatures (data not shown). However, above 47°C, TRAP binds all three RNAs with nearly equal affinities (data not shown), consistent with the proposal that the spacer nucleotides are not forming specific contacts with TRAP. DISCUSSION Two TRAP binding sites have been characterized in B. subtilis. One site, located within the leader region of the trpED-CFBA operon, contains 11 trinucleotide repeats (7 GAG and 4 UAG), each separated by two or three nucleotide spacers (4,5). The second site, located 5Ј of trpG (4, 6), contains 9 trinucleotide repeats (7 GAG, 1 UAG, and 1 AAG) with spacers up to 8 nucleotides long. Recent footprinting experiments have demonstrated that TRAP contacts all nine of these repeats, including FIG. 4. Temperature dependence of the TRAP⅐trp leader RNA interaction. Apparent association constants for the TRAP⅐RNA 36 -92 interaction were obtained by filter binding at temperatures between 5°C and 67°C and plotted as a van't Hoff plot. The data below 47°C were described previously (9). Each point represents the average of at least two independent experiments. Standard deviations were less than 10% of the mean.
FIG. 5. Thermostability of TRAP measured by circular dichroism spectroscopy. The CD spectrum of 11 M TRAP was obtained at 25°C and 70°C. Data were recorded at every nanometer between 260 nm and 205 nm in 50 potassium phosphate pH 7.4. The x axis is the wavelength in nanometers, and the y axis is millidegrees of rotation.

FIG. 6. Differential scanning calorimetry analysis of TRAP.
The heat capacity traces of 11 M TRAP with and without 1 mM L-tryptophan are shown. Data were collected every 1/10 nm. The x axis is temperature in degrees Celsius, and the y axis is the molar heat capacity in KJ/mol. the AAG repeat. 1 The RNAs that we selected in vitro based on their ability to bind TRAP all contain multiple GAG and/or UAG repeats confirming the importance of these trinucleotides for TRAP binding. Moreover, as the affinity of RNAs from successive rounds of selection improved, the average number of G/UAG repeats/RNA increased. The ratio of GAG to UAG repeats in the in vitro selected RNAs was approximately 2:1 (Table I) suggesting TRAP binds GAGs better that UAGs. This finding is consistent both with the bias seen in the natural TRAP binding sites and with the results of Babitzke et al. (7,8), who showed that RNAs containing GAG repeats bound TRAP significantly better than UAG containing RNAs. Most (89%) of the trinucleotide repeats in the RNAs selected in vitro were separated by two nucleotide spacers (Table I). This finding is also consistent with previous results (8) showing that two nucleotide spacers are optimal for TRAP binding. The natural TRAP binding sites also reflect this bias, with 70% of the spacers containing two nucleotides. However, the trp leader RNA also contains three 3-nucleotide spacers and the trpG binding site contains two rather long spacers of up to 8 nucleotides. The identities of the nucleotides in the spacers of the natural TRAP binding sites are relatively random, though there is a bias toward As in both binding sites and a preference for Us in the trp leader. In total, if the spacers from both natural sites are considered, there are equal numbers of pyrimidines and purines (Table I). In sharp contrast, there was a very strong bias (88%) toward pyrimidines in the spacer nucleotides of the RNAs we selected in vitro with Us favored over Cs; G residues are underrepresented in both in vitro selected RNAs and the natural TRAP binding sites (Table I).
The differences, particularly in the composition of the spacers, seen between our in vitro selected RNAs and the natural TRAP binding sites most likely reflect the regulatory functions of the natural TRAP binding sites in vivo. The sequence of the trp leader RNA must not only allow for TRAP binding but also form the appropriate secondary structures required for proper regulation of the trp operon (1). trpG RNA has been shown to bind TRAP with a lower affinity than trp leader RNA in vitro (4). This reduced affinity is consistent with the dual role of the amidotransferase encoded by trpG in both tryptophan and folic acid biosynthesis (23,24), allowing for an appropriate level of trpG expression in the presence of excess tryptophan. Therefore, in both the trp leader RNA and in trpG RNA, the sequences of the TRAP binding sites appear to reflect the need for both TRAP binding and proper gene regulation.
Footprinting studies have shown that in trp leader RNA, the spacer nucleotides are not protected by TRAP binding and this was interpreted to indicate that the spacer nucleotides do not directly contact TRAP (4). When we performed RNA binding assays with three artificial TRAP binding sites with various spacer nucleotides similar to those previously (7), at temperatures where these RNAs should be unstacked, TRAP binds to all of these RNAs with nearly identical affinities (data not shown). This result also indicates that the spacer nucleotides do not play a direct role in TRAP binding. However, recent results (7) suggest that the composition of the spacers does have an effect on TRAP binding. RNAs containing either As or Us as the spacer nucleotides were found to bind TRAP better than those containing Cs or Gs in the spacers. While these results (7) reflect the bias seen in the natural TRAP binding sites, they differ from our in vitro selection results described above. This discrepancy could be due to the high concentrations of TRAP (5.0 nM) and RNA (2.5 nM) used in the filter binding experiments of Babitzke et al. (7). Under these conditions, relatively large differences in the affinity of these RNAs for TRAP would not be observed. Furthermore, these studies used an RNA containing the sequence 5Ј-GAGCU-3Ј repeated six times and we have found that similar RNAs containing homopyrimidine spacers form stable secondary structures at the temperatures used in these studies (data not shown). Since TRAP binds only single-stranded RNAs (7)(8)(9), the conclusion of Babitzke et al. (7) that Cs in the spacers are inhibitory to TRAP binding may instead reflect the tendency of this RNA to form secondary structures. Therefore, we propose that the spacer nucleotides do not contact TRAP but instead have an indirect role in TRAP binding related to their ability to form base stacking interactions, since we have demonstrated that trp leader RNA becomes unstacked when bound to TRAP (Fig. 3).
The affinity of the TRAP⅐RNA interaction displays a nonlinear temperature dependence between from 5°C to 67°C with a maximum at 47°C (Fig. 4). Results from differential scanning calorimetry and circular dichroism spectroscopy indicate that the decrease in the affinity of the TRAP⅐RNA interaction above 47°C cannot be attributed to thermal denaturation of TRAP. Furthermore, preliminary NMR data, to be published elsewhere, confirms that the structure of TRAP is virtually unchanged between 30°C and 60°C. 3 Several models have been proposed to explain a nonlinear temperature dependence in protein:nucleic acid interactions: 1) the heat capacity model proposed by Record and co-workers (25,26) and 2) the coupled equilibria model proposed by Ferrari and Lohman (27).
Based on studies of the lac repressor⅐DNA interaction, Record and co-workers proposed that a large negative change in the standard molar heat capacity (⌬C p°) is the primary source of the observed nonlinear temperature dependence (25,26). This model predicts that the ⌬H°and ⌬S°of the association process vary in parallel as a function of temperature, resulting in a ⌬G°that is nearly temperature-independent. At the temperature where K obs is maximal ⌬H°ϭ 0 (T H ), and at the temperature where ⌬G°is maximal ⌬S°ϭ 0 (T S ). According to this model, the major source of a ⌬C p°i s removal of nonpolar surfaces from the aqueous environment upon complex formation, otherwise known as the hydrophobic effect. Therefore, they suggest that the major driving force of the lac repressor⅐DNA interaction is the removal of nonpolar surfaces from the aqueous environment with the concomitant release of H 2 O upon complex formation. Recent studies have shown that the interaction of the spliceosomal protein U1A with its RNA stem-loop target also displays a nonlinear temperature dependence (28,29). Moreover, it has also been proposed that a ⌬C p°i s the source of the nonlinear temperature dependence in this interaction.
The second model describes the thermodynamics of the interaction between E. coli single-stranded DNA-binding protein (SSB) and poly(dA) (27). In this system, the temperature dependence of SSB binding to poly(dC) or poly(dT) is linear while binding to poly(dA) displays a nonlinear temperature dependence. To explain these results, the authors proposed that SSB only binds to fully unstacked nucleic acids. Since the unstacking of poly(dA) has a positive change in enthalpy (ϩ⌬H) and the binding of SSB to unstacked poly(dA) has an negative enthalpy change (Ϫ⌬H), the observed temperature dependence represents the thermodynamic coupling of poly(dA) unstacking to SSB⅐poly(dA) binding. At low temperatures, poly(dA) is stacked and binds SSB relatively poorly. As the temperature increases, poly(dA) becomes unstacked and SSB binding improves until a temperature is reached where ⌬H unstacking ϭ ⌬H binding ; this temperature represents the inflection point in the van't Hoff plot. Above this temperature, poly(dA) is fully unstacked and the SSB⅐poly(dA) interaction decreases, reflecting the Ϫ⌬H of this interaction.
Based on the insensitivity of the TRAP⅐RNA interaction to ionic strength, we previously proposed that the association of TRAP with trp leader RNA is driven by the release of ordered H 2 O molecules upon complex formation (9). If this hypothesis is correct, the nonlinear-temperature dependence observed for the TRAP⅐trp leader RNA interaction could be explained by the ⌬C p°m odel proposed by Record and co-workers (25,26). However, this model requires that the structures of the protein and the nucleic acid, both alone and in complex with each other, do not change as a function of temperature (27). While our studies have demonstrated that the structure of TRAP remains virtually unchanged in the temperature range studied, trp leader RNA is clearly undergoing a temperature-dependent structural change (presumably unstacking). Thus, the ⌬C p°m odel may not be adequate to describe the thermodynamics of the TRAP⅐trp leader RNA. In contrast, our data indicate that base stacking interactions play a role in the TRAP⅐trp leader RNA interaction, suggesting that the coupled equilibria model proposed by Ferrari and Lohman (27) may better describe the thermodynamics of the TRAP⅐trp leader RNA interaction.
trp leader RNA, like other single-stranded nucleic acids, is stacked in solution ( Fig. 2; Refs. 21-22 and 30). When bound to TRAP, the structure of this RNA changes dramatically (Fig. 3) with a large increase in the average base to base distance, indicative of the disruption of base stacking interactions (22,31). Unstacking of the bases in the RNA is consistent with the proposal that the RNA wraps around TRAP to form a matching circle with the TRAP oligomer (5). The mechanism of interaction of a number of single-stranded DNA binding proteins with ssDNA, including the E. coli SSB (see above; Refs. 32 and 33), T4 gp32 (31, 32, 34 -37), adenovirus DNA-binding protein (38), and the gene 5 protein from the bacteriophage fd (35,39) have been extensively studied. In all of these systems, the ssDNA becomes markedly unstacked when complexed with the protein, suggesting a mechanistic similarity between the TRAP⅐RNA interaction and the interactions between ssDNA and these proteins. Therefore, we looked for possible structural similarities between TRAP and gp32, whose structure has been solved in the presence of ssDNA (40). In this structure, the phosphate backbone of the DNA lies along a line of positively charged residues with the bases falling into the hydrophobic cleft described above. Results from our laboratory have found a group of positively charged residues (Lys-37, Lys-56, and Arg-58) that when changed to an alanine, severely reduce RNA binding to TRAP. 4 Intriguingly, these three residues fall on a line encircling the outer edge of the TRAP 11-mer. By modeling an RNA with its phosphate backbone interacting with this line of positive residues, we find that it is possible to orient the bases in a position where they could interact with Phe-32, located on the surface of TRAP. However, changing this amino acid to an alanine does not appear to adversely effect RNA binding 4 , and further experiments will be necessary to thoroughly characterize what role, if any, this amino acid is having on RNA binding.