A protein-dependent riboswitch controlling ptsGHI operon expression in
http://www.100md.com
《核酸研究医学期刊》
1 Abteilung für Allgemeine Mikrobiologie, Georg-August-Universit?t G?ttingen, Grisebachstrasse 8, D-37077 G?ttingen, Germany and 2 Lehrstuhl für Mikrobiologie, Institut für Mikrobiologie, Biochemie und Genetik der Friedrich-Alexander-Universit?t Erlangen-Nürnberg, Staudtstrasse 5, D-91058 Erlangen, Germany
*To whom correspondence shold be addressed. Tel: +49 551 393781; Fax: +49 551 393808; Email: jstuelk@gwdg.de
ABSTRACT
The Gram-positive soil bacterium Bacillus subtilis transports glucose by the phosphotransferase system. The genes for this system are encoded in the ptsGHI operon. The expression of this operon is controlled at the level of transcript elongation by a protein-dependent riboswitch. In the absence of glucose a transcriptional terminator prevents elongation into the structural genes. In the presence of glucose, the GlcT protein is activated and binds and stabilizes an alternative RNA structure that overlaps the terminator and prevents termination. In this work, we have studied the structural and sequence requirements for the two mutually exclusive RNA structures, the terminator and the RNA antiterminator (the RAT sequence). In both cases, the structure seems to be more important than the actual sequence. The number of paired and unpaired bases in the RAT sequence is essential for recognition by the antiterminator protein GlcT. In contrast, mutations of individual bases are well tolerated as long as the general structure of the RAT is not impaired. The introduction of one additional base in the RAT changed its structure and resulted in complete loss of interaction with GlcT. In contrast, this mutant RAT was efficiently recognized by a different B.subtilis antitermination protein, LicT.
INTRODUCTION
In bacteria, gene expression is most commonly regulated at the level of transcription initiation. This control is achieved by interactions of specific DNA sequences with regulatory proteins. A second mechanism of control of gene expression is mediated at the RNA level by riboswitches. This type of control may affect either transcription or translation. The formation of riboswitches can be triggered by low molecular weight effectors, tRNAs or by regulatory proteins (for reviews see 1–3).
Regulatory protein–RNA interactions control the expression of genes involved in diverse physiological functions. In Bacillus subtilis and other Gram-positive bacteria, genes encoding enzymes for the biosyntheses of tryptophan and pyrimidines are controlled by a termination/antitermination mechanism (4,5). Genes required for the utilization of several sugars are subject to regulation by antitermination in many bacteria (6). Similarly, the utilization of aliphatic amides in Pseudomonas aeruguniosa and assimilatory nitrate reduction in Klebsiella oxytoca is controlled by RNA-binding antitermination proteins (7,8). Recently, an antitermination mechanism that controls gene expression after a cold shock in Escherichia coli were discovered (9).
As with all other riboswitches, protein-dependent antitermination/termination systems are based on the existence of alternative, mutually exclusive RNA structures. One of these structures is a transcriptional terminator. The alternative structure, also called the RNA antiterminator (RAT) (10), prevents formation of the terminator and allow transcription elongation to proceed. Depending on the nature of the system, one of the structures is energetically favored and can form in the absence of any other factor. In contrast, the less favored structure depends on stabilization, which is caused by binding of the regulatory protein. While the antiterminators are more stable in the anabolic systems resulting in transcription elongation as the default state, formation of the terminators is energetically favored in many catabolic systems. Thus, the respective genes are not expressed unless the inducer is present and binding of the antiterminator protein allows formation of the antitermination structure (1,2,10).
We are interested in the control of glucose utilization in B.subtilis. Glucose is the preferred source of carbon and energy for these bacteria. The sugar is transported by the bacterial sugar:phosphoenolpyruvate phosphotransferase system (PTS) and is subsequently catabolized in the glycolytic and pentose phosphate pathways (for a review see 11). Two operons encoding enzymes of glucose catabolism are inducible by the presence of glucose in the medium, the ptsGHI operon and the gapA operon coding for the components of the PTS and for glycolytic enzymes of triose phosphate interconversions, respectively (12–14). While induction of the gapA operon is governed at the level of transcription initiation by a specific repressor (14–16), the ptsGHI operon is controlled by transcriptional antitermination (12). Transcription of this operon is constitutively initiated but stops at a factor-independent terminator upstream of the ptsG structural gene. In the presence of glucose, an antiterminator protein, GlcT, is activated and is thought to bind a RAT sequence overlapping the transcription terminator (see Fig. 1). GlcT binding to the RAT prevents formation of the terminator and allows transcription elongation (17). The RNA-binding activity of GlcT is controlled by reversible phosphorylation in response to the presence of the inducer, glucose. In the presence of glucose, the glucose-specific enzyme II of the PTS may transfer its phosphate group to the sugar. In contrast, enzyme II is permanently phosphorylated in the absence of glucose and this form may phosphorylate and thereby inactivate the antiterminator protein GlcT (18,19).
Figure 1. Proposed schematic model of the antiterminator and terminator structures of the ptsG leader mRNA. The secondary structure of the antiterminator is based on known RAT structures (28). The RAT sequence is boxed in the terminator structure to highlight the overlap between the RAT and the terminator that constitutes the riboswitch.
GlcT is a member of a family of transcriptional antiterminators, the prototypes of which are BglG from E.coli and SacY and LicT from B.subtilis. These proteins are all composed of an N-terminal RNA-binding domain (17,20) and two reiterated PTS regulation domains (PRDs) which control the RNA binding activity of the antiterminators via phosphorylation-dependent dimerization (21–24). The RNA sequences bound by these antiterminators are all very similar to each other. However, while antiterminators such as SacT and SacY from B.subtilis have a relaxed specificity, i.e. they bind RAT sequences of the bgl type in addition to their cognate targets, binding of the B.subtilis LicT and E.coli BglG antiterminators is restricted to their cognate bgl RAT sequences (20,25). The structures of the RNA-binding domains of SacY and LicT were determined and revealed a novel fold for RNA-binding proteins (20,26,27). Until the recent determination of the solution structure of the LicT–RNA complex, the question of how a symmetric dimeric protein would recognize an apparently asymmetric RNA has remained enigmatic. It now turned out that the bglP RAT sequence adopts a quasi-symmetric structure which allows binding of the protein dimer (28). In contrast, the RAT sequence recognized by GlcT is only distantly related to the bgl- and sac-type RAT sequences and no cross-talk was observed (17). Homologs of GlcT and the respective RAT sequence were identified in Staphylococcus carnosus (29).
In this work, we studied the interaction of the RNA-binding domain of GlcT with its cognate RAT target. Several mutations in the RAT sequence were found to affect the antitermination efficiency in vivo. Mutations that restored expression of the ptsGHI operon in the presence of a non-functional RAT all destroyed the transcriptional terminator in the ptsGHI leader mRNA. The sites of interaction between the RNA-binding domain of GlcT and its RAT were studied by in vitro footprinting and mutagenesis. As predicted by Yang et al. (28), our results suggest that the RAT sequence may form a stem–loop structure with a nearly perfect symmetry, which is the recognition site for GlcT. Moreover, we identified a structural specificity determinant which decides the interaction partner.
MATERIALS AND METHODS
Bacterial strains and growth conditions
The B.subtilis strains used in this study are shown in Table 1. Strains used in the random and site-directed mutagenesis studies are listed in Tables 2 and 3. All B.subtilis strains are derivatives of the wild-type strain 168. Escherichia coli DH5 (30) was used for cloning experiments and for expression of recombinant proteins.
Table 1. B.subtilis strains used in this study
Table 2. Effect of mutations in the terminator on expression of a ptsG–lacZ fusion
Table 3. Effect of mutations in the RAT sequence on expression of a ptsG–lacZ fusion
Bacillus subtilis was grown in SP medium or in CSE minimal medium (31). The media were supplemented with auxotrophic requirements (at 50 mg/l) and carbon sources and inducers as indicated. Escherichia coli was grown in LB medium and transformants were selected on plates containing ampicillin (100 μg/ml). LB and SP plates were prepared by the addition of 17 g/l Bacto agar (Difco) to LB or SP medium, respectively.
DNA manipulation
Transformation of E.coli and plasmid DNA extraction were performed using standard procedures (30). Restriction enzymes, T4 DNA ligase and DNA polymerases were used as recommended by the manufacturers. DNA fragments were purified from agarose gels using a Nucleospin extract kit (Macherey and Nagel). Pfu DNA polymerase was used for PCR, as recommended by the manufacturer. The combined chain reaction was performed with Pfu DNA polymerase and thermostable DNA ligase (Ampligase?; Epicentre, Madison, WI). DNA sequences were determined using the dideoxy chain termination method (30). Chromosomal DNA of B.subtilis was isolated as described (32).
Site-directed mutagenesis of the ptsG terminator and RAT sequences
Translational fusions of mutant variants of the ptsG regulatory region with the lacZ gene were constructed using the vector pAC7 (33) containing the kanamycin resistance gene aphA3. The plasmid harbors a lacZ gene without a promoter located between two fragments of the B.subtilis amyE gene. To study the effect of point mutations in the conditional terminator and the RAT sequence preceding the ptsG structural gene the following strategy was applied. A DNA fragment carrying the mutant form of the terminator or ptsG-RAT was constructed by site-directed mutagenesis using PCR-based approaches as outlined previously (34,35). Plasmid pGP66 (12) containing the ptsG promoter region served as a template. Multiple mutations were inserted using plasmids containing one or two mutations as templates. The mutagenic primers and the resulting plasmids are available upon request. The oligonucleotides JS11 (12) and IL5 (17) were used as outer primers. The final PCR products were purified and cut at the BamHI and MfeI sites introduced by the PCR primers. To introduce the constructed lacZ fusions into the chromosome of B.subtilis, competent cells of the wild-type strain 168 were transformed with plasmids carrying the respective mutations linearized with ScaI.
Transformation and characterization of phenotype
Bacillus subtilis was transformed with plasmid DNA according to the two-step protocol described previously (32). Transformants were selected on SP plates containing chloramphenicol (5 μg/ml) or kanamycin (5 μg/ml). In B.subtilis, amylase activity was detected after growth on SP medium supplemented with 10 g/l hydrolyzed starch (Connaught). Starch degradation was detected by sublimating iodine onto the plates.
Quantitative studies of lacZ expression in B.subtilis in liquid medium were performed as follows. Cells were grown in CSE medium supplemented with the carbon sources indicated. Cells were harvested at OD600 = 0.6–0.8. Cell extracts were obtained by treatment with lysozyme and DNase. ?-Galactosidase activity was determined as previously described using o-nitrophenyl-galactoside as substrate (32). One unit is defined as the amount of enzyme which produces 1 nmol o-nitrophenol per min at 28°C.
Protein purification
To purify the RNA-binding domain of GlcT as a native protein, plasmid pGP230 was constructed as follows. Plasmid pGP114 (17) containing the DNA fragment corresponding to amino acids 1–60 of GlcT fused to an N-terminal hexahistidine sequence was linearized with BamHI. The oligonucleotides MM6 (5'-GATCTCTGGTTCCGCGTGGTTCCATGA) and MM7 (5'-GATCTCATGGAACCACGCGGAACCAGA) carrying a DNA fragment encoding a thrombin cleavage site were hybridized at 80°C and ligated to the linearized plasmid pGP114. Positive candidates were verified by sequencing.
Escherichia coli DH5 was used as host for overexpression of the recombinant proteins. Expression was induced by the addition of IPTG (final concentration 1 mM) to logarithmically growing cultures (OD600 = 0.8). The crude extracts were passed over Ni–NTA Superflow (Qiagen), followed by elution with an imidazole gradient. The Bio-Rad dye-binding assay was used to determine protein concentration. Bovine serum albumin was used as the standard. Purified His::GlcT-RBD protein was concentrated using a centriprep concentrator unit (Millipore). The GlcT-RBD peptidic fragment (7498 Da) was generated by thrombin cleavage (Pharmacia) of the His::GlcT-RBD protein according to the supplier’s instructions (Pharmacia). After thrombin cleavage the protein preparation was loaded onto a Superdex 75 prep grade HR16/60 column (Pharmacia) for size exclusion chromatography in 300 mM NaCl, 50 mM Na2HPO4 (pH 7.8). The purity of the protein was determined on 15% Tris–Tricine gels using Coomassie brilliant blue staining (36).
In vitro transcription
To obtain a template for in vitro synthesis of the wild-type RAT RNA, a 84 bp PCR product was generated using pGP66 (12) as template and primers IL59 (5'-CCAAGTA ATACGACTCACTATAGGACGTGTTACTGATTCG) and IL60 (5'-CAAGAATTGGGACAACTCTTCTTCTCCTTTT TTTTCCTCAATCACTCATGCC). To generate a template for in vitro transcription of mutant RAT RNAs, the primers OS25 (5'-CCAAGTAATACGACTCACTATAGGAATTCA GTTTATCCTTAT) and OS26 (5'-TTGAGGGAAAAAA ACGGGAAGTTC) were used to amplify a 99 bp PCR product. The presence of a T7 RNA polymerase recognition site in primers IL59 and OS25 (underlined) allowed the use of the PCR product as a template for in vitro transcription with T7 RNA polymerase (Roche Diagnostics). As a non-specific RNA, a 350 bp gapA transcript was prepared as described previously. The integrity of the RNA transcripts was analyzed by denaturating agarose gel electrophoresis (14).
Assay of interaction between GlcT-RBD and RAT RNA
Binding of GlcT-RBD to RAT RNA was analyzed by gel retardation experiments. The RAT RNA (in water) was denatured by incubation at 90°C for 2 min and renatured by dilution 1:1 with ice-cold water and subsequent incubation on ice. If required, non-specific RNA was renatured separately and mixed with the RAT RNA prior to protein addition. Purified GlcT-RBD was added to the RAT RNA and the samples were incubated for 10 min at room temperature. After this incubation, glycerol was added to a final concentration of 10% (w/v). The samples were then analyzed on 12% Tris–acetate polyacrylamide gels.
RNase T1 footprinting
The method used for RNase T1 digestion of the RAT RNA was as follows. Radiolabeled RNA was denatured by heating to 90°C and was subsequently allowed to cool down at room temperature. Diluted GlcT-RBD (7 pmol in 1 μl) was added to 5'-radiolabeled RAT RNA (140 000 c.p.m., gel purified) in the presence of 20 mM MgCl2 and 300 mM NaCl (final concentration). The reaction was incubated at 37°C for 10 min, after which 20 μg yeast tRNA (Sigma) and appropriate diluted RNase T1 (1 or 2 U) was added, followed by 15 min incubation at 37°C. The reaction was stopped by adding another 20 μg tRNA. The control reactions were performed in the absence of protein or RNase T1 under corresponding conditions. The RAT RNA was extracted with acid phenol/chloroform, precipitated with ethanol, resuspended and analyzed by electrophoresis.
Screening system to isolate random terminator mutants
Plasmids containing the mutated ptsG-RAT and a promoterless kanamycin resistance cassette were obtained as follows. The mutated ptsG-RAT alleles were amplified via PCR using the plasmids pGP329 and pGP330 as template DNA (obtained by site-directed mutagenesis of the RAT sequence; see above) and the oligonucleotides IL5 (17) and JS11 (12). The PCR products were digested with MfeI and BamHI and the 0.5 kb fragment was cloned into pAC6 (12) cut with the same enzymes. The resulting plasmids were designated pGP337 and pGP338. In a second step a promoterless aphA3 kanamycin resistance cassette (37) was amplified using the primers IL46 (5'-CGGGATCCTAATGTTAGAAAAGAGGAAGGA AATAA) and IL48 (5'-CGGGATCCCTACTAAAACAA TTCATCCAGTAA) and pAC7 as template DNA. The DNA fragment was cut with BamHI and cloned into the target vectors pGP337 and pGP338 linearized with the same enzyme. The resulting plasmids were designated pGP339 and pGP340, respectively. Moreover, by the same strategy we constructed plasmid pGP342 harboring a ptsG–aphA3 lacZ fusion under the control of the wild-type regulatory elements. These plasmids were used to introduce a ptsG-aphA3 lacZ fusion into the chromosome of B.subtilis 168 (see Table 1).
Liquid cultures of GP172 and GP173 containing the artificial ptsG–aphA3 lacZ operon were grown overnight in the absence of kanamycin. Aliquots of 100 μl of independently grown cultures were plated onto SP plates containing kanamycin at a final concentration of 30 and 60 μg/ml, respectively. The plates were incubated for 2 days at 28°C to select for colonies able to grow on kanamycin and express the ptsG–aphA3 lacZ fusion. To verify the localization of the selected mutations in the control region of the artificial operon, B.subtilis 168 was transformed with chromosomal DNA of the mutant strains and transformants were selected for chloramphenicol resistance. Transformants resistant to both chloramphenicol and kanamycin contained the mutation in close proximity to the cat resistance gene, which is located downstream of the lacZ gene. The ptsG promoter regions of these strains were analyzed by sequencing.
RESULTS
Mutational analysis of the ptsG terminator sequence
In a previous work, we demonstrated that a mutation disrupting the presumed stem structure of the transcriptional terminator in the ptsG leader region resulted in constitutive expression of ptsG (17). To ensure that it is the structure of the terminator rather than its sequence we constructed a strain, GP169, carrying a compensatory mutation in the terminator region. In this strain, the potential stem structure of the terminator is restored without correcting the sequence. The ptsG promoter region containing this mutation was inserted into the chromosome of B.subtilis in front of a promoterless lacZ gene. The synthesis of ?-galactosidase after growth of the strains in CSE minimal medium in the absence and presence of glucose was monitored to assay transcription driven by the wild-type and mutant ptsG promoter regions (Table 2). While transcription was inducible by glucose in the wild-type strain QB5448, constitutive expression was observed in the terminator mutant GP151. These results are in good agreement with previous findings (17). The newly constructed compensatory mutation restored the inducibility of ptsG expression and therefore the functionality of the transcriptional terminator. Thus, the structure of the terminator seems to be important rather than its actual sequence. The restoration of the terminator structure results in transcription termination in the absence of the inducer, glucose.
To provide further evidence for the functional role of the terminator, we developped a screening and selection system that allowed us to select for spontaneous mutations resulting in constitutive ptsG expression. Briefly, an artificial aphA3 lacZ operon was created and placed under control of the ptsG expression signal. The two strains used here (GP172 and GP173) carried mutations of the RAT sequence that prevented their recognition by GlcT. Mutant strains allowing expression of the artificial operon were selected as decribed in Materials and Methods. Several mutants exhibiting both resistance to kanamycin and constitutive expression of the lacZ gene (as monitored by the blue color of colonies on X-Gal plates) were obtained and their ptsG promoter regions upstream of the aphA3 lacZ operon were sequenced. The mutations all affected the transcriptional terminator in the ptsG leader region. A total of 12 different spontaneous mutations, six derived from each strain, were identified. Among these mutations were four deletions encompassing the terminator region and eight point mutations. One of these point mutations (G47T) was obtained with both original strains, whereas two different mutants obtained with either GP172 and GP173, affected the same position of the terminator (C75A and C75T, respectively). A summary of the sequence information concerning the mutations is shown in Figure 2. The effect of the mutations was quantified by assaying the ?-galactosidase activity driven by the mutant ptsG control elements. As expected, inducible expression was observed in the wild-type strain GP174, while no expression of the lacZ gene was detected in the mutant strains GP172 and GP173 that were used to select for terminator mutants. The mutations of the terminator all resulted in constitutive expression of the artificial aphA3 lacZ operon. However, the absolute level of expression was different for the individual mutants. The deletions and the mutation of G47 resulted in high constitutive expression (400–1700 U ?-galactosidase/mg protein; see Table 2). A point mutation at position C75 or G90 resulted in intermediate expression (100–200 U ?-galactosidase/mg protein), while the mutations of G38 and G72 resulted in weak constitutive expression (<100 U ?-galactosidase/mg protein; see Table 2).
Figure 2. Sequence of the ptsG promoter region. The RAT sequence is shown in bold and the terminator is boxed. The –10 promoter region and the transcription start site are shown in bold. The identified point mutations are shown in circles below the original nucleotides. Arrows with numbers indicate the deleted regions found in the sequenced mutants. The strain designations, the ‘numbering’ of the positions and the effects of the mutations are given in Table 2.
Taken together, our findings indicate that the terminator is the only negatively acting regulatory element in the ptsG promoter region. Moreover, it is the secondary structure rather than the actual sequence of the terminator which is important for the control of ptsG expression.
Binding of the RNA-binding domain of GlcT to the RAT sequence
Previous genetic and biochemical evidence suggested direct binding of GlcT to the RAT sequence that overlaps the terminator in the ptsG leader mRNA. To substantiate these observations and to study the molecular details of protein–RNA interaction, we first purified the N-terminal domain of GlcT devoid of a His tag. In a previous study, binding of His-tagged GlcT to the RAT RNA was demonstrated (17). However, subsequent gel retardation experiments suggested that the affinity tag might cause some non-specific RNA binding (our unpublished results). We therefore constructed a system that allowed removal of the His tag after initial purification of the protein. The purified protein was efficiently cleaved at the introduced thrombin cleavage site. The resulting protein was purified to apparent homogeneity by size exclusion chromatography as described in Materials and Methods.
The ability of the in vitro generated RNA-binding domain of GlcT to bind its RNA target was tested by electrophoretic mobility shift analysis. As shown in Figure 3, addition of the RNA-binding domain to the RAT RNA resulted in a shift of the apparent size of the RAT fragment. At a high concentration of the RNA-binding domain (3-fold molar excess), the RAT RNA was completely shifted (Fig. 3, lane 7). The shifted band contained a complex of both the RNA and the RNA-binding domain, as suggested by the differential detection of both components with ethidium bromide (detection of RNA) and Coomassie brilliant blue (detection of protein) (data not shown).
Figure 3. Electrophoretic mobility band shift analysis of the interaction between the RAT and the RBD. Lane 1 shows 125 pmol of the free RAT RNA. In lanes 2–7, 250 pmol RAT RNA and 30 pmol non-specific RNA were used. Increasing concentrations of RBD-GlcT were added to the RNA prior to electrophoresis. Aliquots of 250, 500, 750, 1000 and 1500 pmol RBD-GlcT were used. The arrows indicate the different RNA species.
Structural analysis of the GlcT–RAT interaction by RNase T1 protection footprint analysis
Specificity determinants have already been studied for the RAT sequences of the bgl and the sac type and the structure of the complex between the bglP RAT and the LicT antiterminator was determined (10,28). However, there is no cross-talk between the antiterminators of the Bgl and Sac types and the ptsG RAT sequence and GlcT is not capable of causing antitermination at bgl or sac RAT sequences (17). Thus GlcT and the ptsG RAT may be the most distant members of this large family of protein-dependent riboswitches. This prompted us to initiate an analysis of the molecular details of the interaction between the RNA-binding domain of GlcT and the ptsG RAT sequence by RNase T1 footprint analysis.
RNase T1 specifically cleaves adjacent to G residues in single-stranded regions of the RNA. Binding of the G residue to a protein or formation of a double-stranded RNA structure protects the RNA from degradation by RNase T1. As expected, incubation of denatured RAT RNA with RNase T1 resulted in RNA cleavage at each of the G residues in the RAT sequence (see Fig. 4, lane 2). Similarly, all G residues are available for cleavage by RNase T1 in the native form of the RAT RNA (see Fig. 4, lanes 3 and 5). Thus, the RAT does not seem to adopt a secondary structure in the absence of GlcT. In the presence of the RNA-binding domain of GlcT several G residues were protected from nucleolytic degradation, with only one G residue remaining accessible. This residue, G16, is probably located in the top loop of the RAT secondary structure and thus part of a single-stranded region in the absence and in the presence of GlcT. Among the residues protected by GlcT, G22, G26 and G28 were most strongly protected. According to the secondary structure model of the RAT RNA (see Figs 1 and 7), G22 and G26 are located in loop regions of the RAT RNA and may thus be protected by interaction with GlcT. G3, G11 and G21 were weakly protected in the presence of GlcT. This may reflect their location in a double-stranded region rather than interaction with GlcT. Indeed, G11 and G21 are both thought to be located in the third double-stranded region of the RAT structure. Thus, the footprinting data indicate that the RAT RNA exists in the single-stranded form in the absence of GlcT and adopts the double-stranded conformation upon GlcT addition. Moreover, GlcT seems to interact with residues in the two loops of the RAT structure.
Figure 4. RNase T1 footprint of the ptsG RAT region. Two different polyacrylamide concentrations were used to resolve the complete region: (A) 26%; (B) 20%. Weak hydrolysis of the RNA in the absence of the RNase and the antiterminator protein shows every nucleotide of the RAT (lane 1). High concentrations of RNase T1 under denaturing conditions leads to cleavage at every guanine nucleotide (lane 2). In the other assays, the RAT RNA and the RBD were used as indicated. In lanes 5 and 6, twice as much radiolabeled RNA was used as compared to lanes 3 and 4. Arrows indicate the position of the guanine nucleotides in the RAT.
Figure 7. Proposed model of the base pairing in the RAT RNA. The circled bases are proposed to directly interact with GlcT.
Probing of the RAT structure and of protein–RNA interactions by site-directed mutagenesis
Since only the guanine residues can be probed by RNase T1 footprinting, we attempted to study the role of the individual bases of the RAT sequence in more detail by site-directed mutagenesis. Mutations of the RAT sequence were introduced into the ptsG promoter region, located in front of a promoterless lacZ gene. If the mutations in the RAT sequence were expected to affect the structural integrity of the overlapping terminator, we introduced compensatory mutations in the terminator to restore the presumed wild-type secondary structure (see Fig. 1).
To assay the effects of the different mutations, the strains were grown in CSE minimal medium in the absence and presence of glucose and the ?-galactosidase activities were determined as an indicator of termination/antitermination activity. The results are summarized in Table 3. As observed previously, a strong induction of the ptsG–lacZ fusion was detected for the wild-type strain QB5448.
Mutations preventing formation of the first double-stranded region of the RAT RNA resulted in loss of induction (GP367, C2G; GP368, G28C). Interestingly, a strain with an inverse arrangement of the C and G residues (and an additional compensatory mutation in the terminator, GP366; see Table 3) resulted in constitutive expression rather than in restoration of inducibility. Constitutive expression is a strong indication of inactivity of the transcriptional terminator. Thus, the overall structure of the RAT/terminator region may have changed in such a way that transcription termination no longer occurs in the triple mutant GP366. To address this hypothesis, we assayed the activity of the ptsG control region present in GP366 in a glcT background. While no expression of the wild-type ptsG–lacZ fusion was observed in the glcT mutant GP109, constitutive expression was detected for the C2G G28C C85G triple mutant (data not shown). This result confirms that antitermination is not required to express ?-galactosidase in this construct. To test whether GlcT binds the RAT sequence with the inverse C:G base pair, we performed a gel mobility shift assay. As shown in Figure 5, this fragment was as efficiently bound by GlcT as the wild-type RAT fragment. In contrast, a fragment containing only the C2G mutation that interferes with formation of the lowest part of the stem was barely shifted by GlcT. Thus, formation of this stem is important for RNA binding by GlcT, whereas the actual sequence does not seem to be critical.
Figure 5. Electrophoretic mobility band shift analysis of the interaction between different mutant RAT species and the RBD.
Mutations affecting loop 1 of the RAT had different effects: two different mutations affecting the G3 residue resulted in complete loss of induction. This finding is in good agreement with the observed role of this residue in the interaction with GlcT as inferred from the RNase T1 protection experiment. Moreover, a RAT sequence containing the G3C mutation is not bound by GlcT in vitro (see Fig. 5). However, both mutations G3C and G3A may alternatively affect the structure of the RAT RNA, as both introduced bases would find pairing partners on the opposing side of the RAT (U25 or G26). Similarly, mutations affecting the residues U25 and A27 resulted in loss of antitermination. These mutations were not expected to alter the secondary structure of the RAT. However, the compensatory mutation for one of the mutations (U86C in GP381) results in loss of terminator activity and thus in constitutive expression of ?-galactosidase. Finally, the mutation G26U present in GP370 resulted in increased expression of the ptsG–lacZ fusion, both in the absence and presence of glucose. Here, the terminator may be leaky due to the compensatory mutation (A:U instead of G:C base pair) and allow some basal expression even in the absence of glucose. Alternatively, the perfectly symmetrical structure of the RAT generated by the G26U mutation might result in improved binding of GlcT, including some weak antitermination in the absence of glucose. Indeed, the mutated RAT seems to be an efficient target for GlcT, as judged from the clear induction of ?-galactosidase in this strain. In addition, the RAT RNA containing this mutation was efficiently bound by GlcT in vitro (see Fig. 5). Moreover, expression of this ptsG–lacZ fusion was completely lost in a glcT mutant strain (see Table 4). Taken together, our data indicate that loop 1 is of great importance for the proper interaction of the RAT RNA with GlcT.
Table 4. Effect of mutations in the ptsG RAT sequence on induction of a ptsG–lacZ fusion
One mutant affected in the second short double-stranded region of the RAT was studied. In this strain, GP369, the predicted G:C base pair was inverted to be C:G, and efficient induction of ?-galactosidase activity was detected. Additionally, the RAT RNA carrying the inversion interacted with the RNA-binding domain of GlcT in vitro (see Fig. 5). Thus, the structure rather than the actual nucleotide at this position is important for GlcT binding and antitermination. This is in good agreement with the observed weak protection of G5 from RNase T1 digestion in the presence of GlcT.
Several mutations in loop 2 resulted in loss of GlcT binding to the RAT and thus in loss of antitermination, namely those affecting U6 and G22. Results obtained with the ptsG–lacZ fusion (Table 3) and the in vitro RAT–GlcT binding assay (Fig. 5) are in perfect agreement for these mutations. In contrast, replacement of U7 and/or A8 showed no or only slight effects. The importance of G22 for efficient transcriptional antitermination is consistent with the strong RNase T1 protection obtained for this base in the presence of GlcT. However, similarly to loop 1, not all bases seem to play an important role in GlcT recognition.
As we approached the top of the proposed RAT structure, the mutations had lesser effects. The first base pair of the third double-stranded region is important for antitermination, as judged from lack of expression of the ptsG–lacZ fusion in the corresponding mutant strain GP178. An inversion of this base pair along with the required compensatory mutation in the terminator caused constitutive expression (GP372), thus we are unable to decide whether this inversion would affect antitermination. The constitutive expression of ?-galactosidase in this strain was not abolished by deletion of the glcT gene (data not shown), indicating that the mutation inactivated the terminator. An in vitro assay of GlcT binding to the mutant RAT suggested that an interaction occurred, albeit the complex seemed to be less stable (Fig. 5). A mutation of G11 had a mild effect whereas the A12U U13A double mutation present in GP166 did not influence the antitermination activity of GlcT. Similarly, three mutations in the top loop of the RAT did not affect induction by glucose and, therefore, binding of GlcT. These results indicate that the upper part of the RAT may only be required to obtain the correct RAT structure and is not involved in the interaction with GlcT.
Converting the recognition specificity of the RAT sequence to LicT binding
A comparison of RAT sequences of the glucose, sucrose, ?-glucoside and lactose classes revealed that all but the glucose RAT sequences contain 3 bp in the second double-stranded region whereas the ptsG RAT sequences of B.subtilis and S.carnosus contain only 2 bp (see Fig. 6). Moreover, loop 1 of the ptsG RAT is made up of four putatively non-paired bases whereas three non-paired bases form this loop in the other RAT sequences (12,28,29) (see Fig. 6). Our mutagenesis studies suggested an important role of RAT structure in the loop regions for the interaction with GlcT. We therefore asked whether this unique arrangement of the ptsG RAT is important for induction specificity by GlcT.
Figure 6. Proposed secondary structure for different RAT RNAs (28). The fold proposed for the ptsG-Ains RAT is based on similarity to the bglP and sacB RAT RNAs.
To address this question we constructed a B.subtilis strain with a ptsG–lacZ fusion with an additional adenine nucleotide between G3 and U4 of the RAT. This insertion is predicted to result in the formation of an extra A:U base pair in the second paired region. Concomitantly, the mutant loop 1 would contain three bases instead of four. Thus, the structure of this RAT is expected to be similar to those of the sucrose and ?-glucoside classes (see Fig. 6). Binding of antitermination proteins to this mutant RAT was assayed by determining the activity of the ptsG–lacZ fusion under conditions when the different antiterminator proteins are known to be active. In the wild-type strain QB5448, glucose, sucrose and salicin induced expression of the fusion. This finding is in agreement with previous reports and results from the activity of GlcT due to non-specific regulation by glucose permease (compare QB5448 and GP109 in Table 4) (12,17,38). In contrast, the mutant ptsG control region present in GP385 did not allow induction by glucose, suggesting that the altered RAT was not recognized by GlcT. However, significant induction was observed in the presence of sucrose or the ?-glucoside salicin. This induction was independent of a functional glcT gene (compare GP385 and GP386 in Table 4), suggesting that the mutant RAT was bound by a different antiterminator. To test this possibility, we constructed strains carrying the wild-type or mutant RAT sequences and deletions of the genes encoding the other B.subtilis antiterminators of the BglG family, i.e. sacT, sacY and licT. These deletions had no effect on the activity of a ptsG–lacZ fusion in the presence of the wild-type RAT sequence. This observation is in good agreement with the idea that induction of ptsG by sucrose and salicin is mediated by GlcT (17,38; see above). However, the lacZ fusion controlled by the mutant RAT was not induced by sucrose and salicin in the licT mutant strain GP390. Thus, LicT is essential to overcome transcription termination in the presence of the altered RAT sequence. Similarly, induction by sucrose is lost in the sacT sacY double mutant strain GP388. SacT and SacY recognize identical RAT sequences and therefore the effect of these mutations was not assayed individually. Interestingly, induction is still possible in the presence of salicin in GP388, confirming that this induction is caused by LicT.
DISCUSSION
As a target of regulation, RNA differs substantially from DNA in its structural diversity and in its ability to adopt alternative, mutually exclusive structures. Moreover, these structures, the riboswitches, are capable of interacting with a wide variety of regulatory partners, such as metabolites, other RNAs or proteins. Due to this huge variability, the themes in RNA–protein recognition are much more diverse than those identified in DNA–protein interactions.
GlcT and the ptsG riboswitch are members of an expanding family of transcription regulatory systems (22). A recent analysis of the evolution of the regulatory domains of the antitermination proteins of this family revealed that GlcT from B.subtilis and its ortholog from S.carnosus form a distinct subgroup whereas the antiterminators of the sucrose and ?-glucoside classes exhibit a close relationship with each other (39). This is supported by the finding that there is no cross-talk between the regulatory components of the glucose type on the one hand and those of the sucrose and ?-glucoside type on the other (17). In contrast, binding of sucrose antiterminator proteins to RAT sequences of bgl genes was reported (10).
Our studies regarding the overall structure of the ptsG RAT suggest that it folds similarly to the RAT RNAs of the bgl and sac classes, thus confirming a proposal of Yang et al. (28). Interestingly, this structure deviates from that obtained by the Zuker algorithm proposed earlier (17). However, both mutational and footprint analyses are in disagreement with the RAT structure suggested previously, whereas they confirm the model depicted here (see Figs 1, 6 and 7). Mutations in the proposed stem regions had different consequences: single mutations prevented antitermination while compensatory mutations that restored the base pairing restored GlcT binding at the same time. Similar results were obtained with the sacB RAT sequence (10).
While the sequence of the ptsG RAT differs substantially from those of the RAT sequences of the sac and bgl classes, we detected a perfect match of loop 2 of the ptsG, bglP and sacB RAT sequences and the regions surrounding loop 2 (see Fig. 6). It therefore seems safe to conclude that this region of the RAT is structurally identical to that of the bglP RAT as determined by NMR spectroscopy (28). In loop 2, G22 forms contacts with two bases, U6 and A8. Thus, only U7 extrudes from the stem structure (Fig. 7). Analysis of the LicT interaction with the bglP RAT revealed that G22 is not involved in direct interactions with the protein. However, the mutations of G22 and its partners studied in this work abolished antitermination. This confirms the critical importance of these residues for the structure of the RAT. The ptsG RAT is unique since it is the only RAT sequence with a symmetrical arrangement of the two loops. Therefore, it is tempting to speculate that G3 is involved in contacts with U25 and A27, with G26 being exposed. This idea is again supported by the mutational analysis: mutations disrupting the ‘ménage à trois’ in loop 1 strongly interfere with antitermination and are therefore expected to interfere with the RAT structure, as observed for loop 2.
Our studies concerning the interaction of GlcT with the RAT revealed that loops 1 and 2, as well as the stems, are critical for antitermination. This is in good agreement with previous findings for the SacY–sacB RAT and LicT–bglP RAT interactions (10,28). Our mutation and footprint analyses demonstrated that the top loop is important neither for adoption of the proper RAT structure nor for the interaction with GlcT. In contrast, the RNA-binding coat protein of bacteriophage MS2 and the phage N protein requires the top loop of their RNA targets for binding (40,41). As far as the paired regions are concerned, our analysis fails to identify the bases and amino acids in the RAT and GlcT, respectively, that are directly involved in the interaction. However, in our footprint analysis G26 in loop 1 was strongly protected from RNase T1 cleavage in the presence of GlcT, suggesting that it is contacted by the protein. The corresponding base in loop 2 was shown to be involved in interaction of the bglP RAT with several amino acids of LicT (28). We have analyzed mutations of these two exposed nucleotides in loop 1 (G26) and loop 2 (U7). In both cases a strong induction by glucose was observed. We may therefore propose that although these exposed bases are contact partners for GlcT, their actual nature does not seem to be crucial. This fits well with previous observations on the interaction of the bglP RAT and LicT (28).
The idea that the structure of the RAT rather than its sequence is important for recognition by GlcT is strongly supported by the observation that insertion of a single nucleotide that makes the RAT structure more similar to that of the bgl and sac RATs resulted in complete loss of GlcT-dependent induction. In contrast, the novel RAT is recognized by LicT even though the sequence remains quite different from that of the cognate LicT RAT sequences. On the other hand, a small sequence alteration with a probably important structural consequence results in complete loss of antitermination by GlcT. So far, it is known that SacT and SacY bind to RAT sequences of the bgl and sac classes whereas LicT (and its E.coli counterpart BglG) binds only to bgl RAT sequences (10,20). The structures of the RAT sequences of these two classes are very similar (see Fig. 6). Accordingly, the recognition specificity for LicT is determined in the sequence of the RAT, and mutations in loop 1 of the sacB RAT sequence allowed binding of BglG in addition to SacY. The effects of the corresponding mutations were cumulative (10). This is in agreement with the idea that SacY and LicT recognize RAT RNAs with the same structure. In contrast, we report here the first example of a complete change of partner specificity. A single mutation that changes the structure of the RAT RNA is sufficient for altered recognition and results in an all-or-nothing effect. Obviously, a RAT can be a member of the glc or the sac/bgl family, with no intermediate possible. The RNA-binding domains of LicT and GlcT are well conserved (for alignments see 17,28). Five amino acids of LicT were shown to be involved in the direct interaction with the internal loops of the RAT RNA (28). Of these, three (N10, G26 and F31) are conserved in all B.subtilis antiterminator proteins of this family. The residue K5 is conserved in all proteins but SacY. Only R27 is not conserved at all and may thus be involved in providing interaction specificity.
The evolution of regulation by adaptation of DNA-binding proteins to new DNA-binding sites is well documented. In the case of protein-dependent riboswitches, the question arises what was first: does the protein evolve to recognize an altered RNA or does the RNA adapt to protein variations? Our results hint at the second idea: one mutation in the RNA is sufficient to gain a novel regulatory specificity. In contrast, many mutations in the protein would be required to recognize an RNA target with a similar sequence but different structure. It will be interesting to study the molecular details of GlcT–RAT interaction by determining the structure of the complex.
ACKNOWLEDGEMENTS
Wolfgang Hillen is acknowledged for providing a stimulating environment. We are grateful to Matthias G?rlach and Matthias Stoldt for helpful discussions and Steffi Bachem, Shane Hanson and Ingrid Wacker for the help with some experiments. Ulf Gerth and Dominique Le Coq are acknowledged for the gift of strains. I.L. was the recipient of a personal stipend from the Boehringer-Ingelheim Foundation. This work was supported by grants from the Deutsche Forschungsgemeinschaft (through SFB 473) and the Fonds der Chemischen Industrie to J.S.
REFERENCES
Henkin,T.M. (2000) Transcription termination control in bacteria. Curr. Opin. Microbiol., 3, 149–153.
Stülke,J. (2002) Control of transcription termination in bacteria by RNA-binding proteins that modulate RNA structures. Arch. Microbiol., 177, 433–440.
Winkler,W.C. and Breaker,R.R. (2003) Genetic control by metabolite-binding riboswitches. Chem. Biochem., 4, 1024–1032.
Babitzke,P. and Gollnick,P. (2001) Posttranscription initiation control of tryptophan metabolism in Bacillus subtilis by the trp RNA-binding attenuation protein (TRAP), anti-TRAP and RNA structure. J. Bacteriol., 183, 5795–5802.
Switzer,R.L., Turner,R.J. and Lu,Y. (1999) Regulation of the Bacillus subtilis pyrimidine biosynthetic operon by transcriptional attenuation: control of gene expression by an mRNA-binding protein. Prog. Nucleic Acid Res. Mol. Biol., 62, 329–367.
Rutberg,B. (1997) Antitermination of transcription of catabolic operons. Mol. Microbiol., 23, 413–421.
Wilson,S.A., Wachira,S.J.M., Norman,R.A., Pearl,L.H. and Drew,R.E. (1996) Transcription antitermination regulation of the Pseudomonas aeruginosa amidase operon. EMBO J., 15, 5907–5916.
Chai,W. and Stewart,V. (1998) NasR, a novel RNA-binding protein, mediates nitrate-responsive transcription antitermination of the Klebsiella oxytoca M5al nasF operon leader in vitro. J. Mol. Biol., 283, 339–351.
Bae,W., Xia,B., Inouye,M. and Severinov,K. (2001) Escherichia coli CspA-family RNA chaperones are transcription antiterminators. Proc. Natl Acad. Sci. USA, 97, 7784–7789
Aymerich,S. and Steinmetz,M. (1992) Specificity determinants and structural features in the RNA target of the bacterial antiterminator proteins of the BglG/SacY family. Proc. Natl Acad. Sci. USA, 89, 10410–10414.
Stülke,J. and Hillen,W. (2000) Regulation of carbon catabolism in Bacillus species. Annu. Rev. Microbiol., 54, 849–880.
Stülke,J., Martin-Verstraete,I., Zagorec,M., Rose,M., Klier,A. and Rapoport,G. (1997) Induction of the Bacillus subtilis ptsGHI operon by glucose is controlled by a novel antiterminator, GlcT. Mol. Microbiol., 25, 65–78.
Tobisch,S., Zühlke,D., Bernhardt,J., Stülke,J. and Hecker,M. (1999) Role of CcpA in regulation of the central pathways of carbon catabolism in Bacillus subtilis. J. Bacteriol., 181, 6996–7004.
Ludwig,H., Homuth,G., Schmalisch,M., Dyka,F.M., Hecker,M. and Stülke,J. (2001) Transcription of glycolytic genes and operons in Bacillus subtilis: evidence for the presence of multiple levels of control of the gapA operon. Mol. Microbiol., 41, 409–422.
Fillinger,S., Boschi-Muller,S., Azza,S., Dervyn,E., Branlant,G. and Aymerich,S. (2000) Two glyceraldehyde-3-phosphate dehydrogenases with opposite physiological roles in a nonphotosynthetic bacterium. J. Biol. Chem., 275, 14031–14037.
Doan,T. and Aymerich,S. (2003) Regulation of the central glycolytic genes in Bacillus subtilis: binding of the repressor CggR to its single DNA target sequence is modulated by fructose-1,6-bisphosphate. Mol. Microbiol., 47, 1709–1721.
Langbein,I., Bachem,S. and Stülke,J. (1999) Specific interaction of the RNA-binding domain of the Bacillus subtilis transcriptional antiterminator GlcT with its RNA target, RAT. J. Mol. Biol., 293, 795–805.
Bachem,S. and Stülke,J. (1998) Regulation of the Bacillus subtilis GlcT antiterminator protein by components of the phosphotransferase system. J. Bacteriol., 180, 5319–5326.
Schmalisch,M., Bachem,S. and Stülke,J. (2003) Control of the Bacillus subtilis antiterminator protein GlcT by phosphorylation: elucidation of the phosphorylation chain leading to inactivation of GlcT. J. Biol. Chem., 278, 51108–51115.
Manival,X., Yang,Y., Strub,M.P., Kochoyan,M., Steinmetz,M. and Aymerich,S. (1997) From genetic to structural characterization of a new class of RNA-binding domain within the SacY/BglG family of antiterminator proteins. EMBO J., 16, 5019–5029.
Tortosa,P., Aymerich,S., Lindner,C., Saier,M.H.,Jr, Reizer,J. and Le Coq,D. (1997) Multiple phosphorylation of SacY, a Bacillus subtilis antiterminator negatively controlled by the phosphotransferase system. J. Biol. Chem., 272, 17230–17237.
Stülke,J., Arnaud,M., Rapoport,G. and Martin-Verstraete,I. (1998) PRD—a protein domain involved in PTS-dependent induction and carbon catabolite repression of catabolic operons in bacteria. Mol. Microbiol., 28, 865–874.
Amster-Choder,O. and Wright,A. (1992) Modulation of the dimerization of a transcriptional antiterminator protein by phosphorylation. Science, 257, 1395–1398.
Fux,L., Nussbaum-Schochat,A. and Amster-Choder,O. (2003) Interactions between the PTS regulation domains of the BglG transcriptional antiterminator from Escherichia coli. J. Biol. Chem., 278, 46203–46209.
Schnetz,K., Stülke,J., Gertz,S., Krüger,S., Krieg,M., Hecker,M. and Rak,B. (1996) LicT, a Bacillus subtilis transcriptional antiterminator of the BglG family. J. Bacteriol., 178, 1971–1979.
van Tilbeurgh,H., Manival,X., Aymerich,S., Lhoste,J.M., Dumas,C. and Kochoyan,M. (1997) Crystal structure of a new RNA-binding domain from the antiterminator protein SacY of Bacillus subtilis. EMBO J., 16, 5030–5036.
Declerck,N., Vincent,F., Hoh,F., Aymerich,S. and van Tilbeurgh,H. (1999) RNA recognition by transcriptional antiterminators of the BglG/SacY family: functional and structural comparison of the CAT domains from SacY and LicT. J. Mol. Biol., 294, 389–402.
Yang,Y., Declerck,N., Manival,X., Aymerich,S. and Kochoyan,M. (2002) Solution structure of the LicT-RNA antitermination complex: CAT clamping RAT. EMBO J., 21, 1987–1997.
Knezevic,I., Bachem,S., Sickmann,A., Meyer,H., Stülke,J. and Hengstenberg,W. (2000) Regulation of the glucose specific phosphotransferase system (PTS) of Staphylococcus carnosus by the antiterminator protein GlcT. Microbiology, 146, 2333–2342.
Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning: A Laboratory Manual, 2nd Edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Faires,N., Tobisch,S., Bachem,S., Martin-Verstraete,I., Hecker,M. and Stülke,J. (1999) The catabolite control protein CcpA controls ammonium assimilation in Bacillus subtilis. J. Mol. Microbiol. Biotechnol., 1, 141–148.
Kunst,F. and Rapoport,G. (1995) Salt stress is an environmental signal affecting degradative enzyme synthesis in Bacillus subtilis. J. Bacteriol., 177, 2403–2407.
Weinrauch,Y., Msadek,T., Kunst,F. and Dubnau,D. (1991) Sequence and properties of comQ, a new competence regulatory gene of Bacillus subtilis. J. Bacteriol., 173, 5685–5693.
Bachem,S., Faires,N. and Stülke,J. (1997) Characterization of the presumptive phosphorylation sites of the Bacillus subtilis glucose permease by site-directed mutagenesis: implication in glucose transport and catabolite repression. FEMS Microbiol. Lett., 156, 233–238.
Bi,W. and Stambrook,P.J. (1998) Site-directed mutagenesis by combined chain reaction. Anal. Biochem., 256, 137–140.
Sch?gger,H. and von Jagow,G. (1987) Tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis for the separation of proteins in the range from 1 to 100 kDa. Anal. Biochem., 166, 368–379.
Trieu-Cuot,P. and Courvalin,P. (1983) Nucleotide sequence of the Streptococcus faecalis plasmid gene encoding the 3'-5'-aminoglycoside phosphotransferase type III. Gene, 23, 331–341.
G?rke,B., Fraysse,L. and Galinier,A. (2004) Drastic differences in Crh and HPr synthesis reflect their different impacts on catabolite repression in Bacillus subtilis. J. Bacteriol., 186, 2992–2995.
Greenberg,D.B., Stülke,J. and Saier,M.H.,Jr (2002) Domain analysis of transcriptional regulators bearing PTS regulatory domains. Res. Microbiol., 153, 519–526.
Grahn,E., Stonehouse,N.J., Murray,J.B., van den Worm,S., Valeg?rd,K., Fridborg,K., Stockley,P.G. and Liljas,L. (1999) Crystallographic studies of RNA hairpins in complexes with recombinant MS2 capsids: implications for binding requirements. RNA, 5, 131–138.
Legault,P., Li,J., Mogridge,J., Lewis,K.E. and Greenblatt,J. (1998) NMR structure of the bacteriophage N peptide/boxB RNA complex: recognition of a GNRA fold by an arginine-rich motif. Cell, 93, 289–299.
Krüger,S. and Hecker,M. (1995) Regulation of the putative bglPH operon for aryl-?-glycoside utilization in Bacillus subtilis. J. Bacteriol., 177, 5590–5597.
LeCoq,D., Lindner,C., Krüger,S., Steinmetz,M. and Stülke,J. (1995) New ?-glucoside (bgl) genes in Bacillus subtilis: the bglP gene product has both transport and regulatory functions similar to those of BglF, its Escherichia coli homolog. J. Bacteriol., 177, 1527–1535.(Oliver Schilling1,2, Ines Langbein2, Mic)
*To whom correspondence shold be addressed. Tel: +49 551 393781; Fax: +49 551 393808; Email: jstuelk@gwdg.de
ABSTRACT
The Gram-positive soil bacterium Bacillus subtilis transports glucose by the phosphotransferase system. The genes for this system are encoded in the ptsGHI operon. The expression of this operon is controlled at the level of transcript elongation by a protein-dependent riboswitch. In the absence of glucose a transcriptional terminator prevents elongation into the structural genes. In the presence of glucose, the GlcT protein is activated and binds and stabilizes an alternative RNA structure that overlaps the terminator and prevents termination. In this work, we have studied the structural and sequence requirements for the two mutually exclusive RNA structures, the terminator and the RNA antiterminator (the RAT sequence). In both cases, the structure seems to be more important than the actual sequence. The number of paired and unpaired bases in the RAT sequence is essential for recognition by the antiterminator protein GlcT. In contrast, mutations of individual bases are well tolerated as long as the general structure of the RAT is not impaired. The introduction of one additional base in the RAT changed its structure and resulted in complete loss of interaction with GlcT. In contrast, this mutant RAT was efficiently recognized by a different B.subtilis antitermination protein, LicT.
INTRODUCTION
In bacteria, gene expression is most commonly regulated at the level of transcription initiation. This control is achieved by interactions of specific DNA sequences with regulatory proteins. A second mechanism of control of gene expression is mediated at the RNA level by riboswitches. This type of control may affect either transcription or translation. The formation of riboswitches can be triggered by low molecular weight effectors, tRNAs or by regulatory proteins (for reviews see 1–3).
Regulatory protein–RNA interactions control the expression of genes involved in diverse physiological functions. In Bacillus subtilis and other Gram-positive bacteria, genes encoding enzymes for the biosyntheses of tryptophan and pyrimidines are controlled by a termination/antitermination mechanism (4,5). Genes required for the utilization of several sugars are subject to regulation by antitermination in many bacteria (6). Similarly, the utilization of aliphatic amides in Pseudomonas aeruguniosa and assimilatory nitrate reduction in Klebsiella oxytoca is controlled by RNA-binding antitermination proteins (7,8). Recently, an antitermination mechanism that controls gene expression after a cold shock in Escherichia coli were discovered (9).
As with all other riboswitches, protein-dependent antitermination/termination systems are based on the existence of alternative, mutually exclusive RNA structures. One of these structures is a transcriptional terminator. The alternative structure, also called the RNA antiterminator (RAT) (10), prevents formation of the terminator and allow transcription elongation to proceed. Depending on the nature of the system, one of the structures is energetically favored and can form in the absence of any other factor. In contrast, the less favored structure depends on stabilization, which is caused by binding of the regulatory protein. While the antiterminators are more stable in the anabolic systems resulting in transcription elongation as the default state, formation of the terminators is energetically favored in many catabolic systems. Thus, the respective genes are not expressed unless the inducer is present and binding of the antiterminator protein allows formation of the antitermination structure (1,2,10).
We are interested in the control of glucose utilization in B.subtilis. Glucose is the preferred source of carbon and energy for these bacteria. The sugar is transported by the bacterial sugar:phosphoenolpyruvate phosphotransferase system (PTS) and is subsequently catabolized in the glycolytic and pentose phosphate pathways (for a review see 11). Two operons encoding enzymes of glucose catabolism are inducible by the presence of glucose in the medium, the ptsGHI operon and the gapA operon coding for the components of the PTS and for glycolytic enzymes of triose phosphate interconversions, respectively (12–14). While induction of the gapA operon is governed at the level of transcription initiation by a specific repressor (14–16), the ptsGHI operon is controlled by transcriptional antitermination (12). Transcription of this operon is constitutively initiated but stops at a factor-independent terminator upstream of the ptsG structural gene. In the presence of glucose, an antiterminator protein, GlcT, is activated and is thought to bind a RAT sequence overlapping the transcription terminator (see Fig. 1). GlcT binding to the RAT prevents formation of the terminator and allows transcription elongation (17). The RNA-binding activity of GlcT is controlled by reversible phosphorylation in response to the presence of the inducer, glucose. In the presence of glucose, the glucose-specific enzyme II of the PTS may transfer its phosphate group to the sugar. In contrast, enzyme II is permanently phosphorylated in the absence of glucose and this form may phosphorylate and thereby inactivate the antiterminator protein GlcT (18,19).
Figure 1. Proposed schematic model of the antiterminator and terminator structures of the ptsG leader mRNA. The secondary structure of the antiterminator is based on known RAT structures (28). The RAT sequence is boxed in the terminator structure to highlight the overlap between the RAT and the terminator that constitutes the riboswitch.
GlcT is a member of a family of transcriptional antiterminators, the prototypes of which are BglG from E.coli and SacY and LicT from B.subtilis. These proteins are all composed of an N-terminal RNA-binding domain (17,20) and two reiterated PTS regulation domains (PRDs) which control the RNA binding activity of the antiterminators via phosphorylation-dependent dimerization (21–24). The RNA sequences bound by these antiterminators are all very similar to each other. However, while antiterminators such as SacT and SacY from B.subtilis have a relaxed specificity, i.e. they bind RAT sequences of the bgl type in addition to their cognate targets, binding of the B.subtilis LicT and E.coli BglG antiterminators is restricted to their cognate bgl RAT sequences (20,25). The structures of the RNA-binding domains of SacY and LicT were determined and revealed a novel fold for RNA-binding proteins (20,26,27). Until the recent determination of the solution structure of the LicT–RNA complex, the question of how a symmetric dimeric protein would recognize an apparently asymmetric RNA has remained enigmatic. It now turned out that the bglP RAT sequence adopts a quasi-symmetric structure which allows binding of the protein dimer (28). In contrast, the RAT sequence recognized by GlcT is only distantly related to the bgl- and sac-type RAT sequences and no cross-talk was observed (17). Homologs of GlcT and the respective RAT sequence were identified in Staphylococcus carnosus (29).
In this work, we studied the interaction of the RNA-binding domain of GlcT with its cognate RAT target. Several mutations in the RAT sequence were found to affect the antitermination efficiency in vivo. Mutations that restored expression of the ptsGHI operon in the presence of a non-functional RAT all destroyed the transcriptional terminator in the ptsGHI leader mRNA. The sites of interaction between the RNA-binding domain of GlcT and its RAT were studied by in vitro footprinting and mutagenesis. As predicted by Yang et al. (28), our results suggest that the RAT sequence may form a stem–loop structure with a nearly perfect symmetry, which is the recognition site for GlcT. Moreover, we identified a structural specificity determinant which decides the interaction partner.
MATERIALS AND METHODS
Bacterial strains and growth conditions
The B.subtilis strains used in this study are shown in Table 1. Strains used in the random and site-directed mutagenesis studies are listed in Tables 2 and 3. All B.subtilis strains are derivatives of the wild-type strain 168. Escherichia coli DH5 (30) was used for cloning experiments and for expression of recombinant proteins.
Table 1. B.subtilis strains used in this study
Table 2. Effect of mutations in the terminator on expression of a ptsG–lacZ fusion
Table 3. Effect of mutations in the RAT sequence on expression of a ptsG–lacZ fusion
Bacillus subtilis was grown in SP medium or in CSE minimal medium (31). The media were supplemented with auxotrophic requirements (at 50 mg/l) and carbon sources and inducers as indicated. Escherichia coli was grown in LB medium and transformants were selected on plates containing ampicillin (100 μg/ml). LB and SP plates were prepared by the addition of 17 g/l Bacto agar (Difco) to LB or SP medium, respectively.
DNA manipulation
Transformation of E.coli and plasmid DNA extraction were performed using standard procedures (30). Restriction enzymes, T4 DNA ligase and DNA polymerases were used as recommended by the manufacturers. DNA fragments were purified from agarose gels using a Nucleospin extract kit (Macherey and Nagel). Pfu DNA polymerase was used for PCR, as recommended by the manufacturer. The combined chain reaction was performed with Pfu DNA polymerase and thermostable DNA ligase (Ampligase?; Epicentre, Madison, WI). DNA sequences were determined using the dideoxy chain termination method (30). Chromosomal DNA of B.subtilis was isolated as described (32).
Site-directed mutagenesis of the ptsG terminator and RAT sequences
Translational fusions of mutant variants of the ptsG regulatory region with the lacZ gene were constructed using the vector pAC7 (33) containing the kanamycin resistance gene aphA3. The plasmid harbors a lacZ gene without a promoter located between two fragments of the B.subtilis amyE gene. To study the effect of point mutations in the conditional terminator and the RAT sequence preceding the ptsG structural gene the following strategy was applied. A DNA fragment carrying the mutant form of the terminator or ptsG-RAT was constructed by site-directed mutagenesis using PCR-based approaches as outlined previously (34,35). Plasmid pGP66 (12) containing the ptsG promoter region served as a template. Multiple mutations were inserted using plasmids containing one or two mutations as templates. The mutagenic primers and the resulting plasmids are available upon request. The oligonucleotides JS11 (12) and IL5 (17) were used as outer primers. The final PCR products were purified and cut at the BamHI and MfeI sites introduced by the PCR primers. To introduce the constructed lacZ fusions into the chromosome of B.subtilis, competent cells of the wild-type strain 168 were transformed with plasmids carrying the respective mutations linearized with ScaI.
Transformation and characterization of phenotype
Bacillus subtilis was transformed with plasmid DNA according to the two-step protocol described previously (32). Transformants were selected on SP plates containing chloramphenicol (5 μg/ml) or kanamycin (5 μg/ml). In B.subtilis, amylase activity was detected after growth on SP medium supplemented with 10 g/l hydrolyzed starch (Connaught). Starch degradation was detected by sublimating iodine onto the plates.
Quantitative studies of lacZ expression in B.subtilis in liquid medium were performed as follows. Cells were grown in CSE medium supplemented with the carbon sources indicated. Cells were harvested at OD600 = 0.6–0.8. Cell extracts were obtained by treatment with lysozyme and DNase. ?-Galactosidase activity was determined as previously described using o-nitrophenyl-galactoside as substrate (32). One unit is defined as the amount of enzyme which produces 1 nmol o-nitrophenol per min at 28°C.
Protein purification
To purify the RNA-binding domain of GlcT as a native protein, plasmid pGP230 was constructed as follows. Plasmid pGP114 (17) containing the DNA fragment corresponding to amino acids 1–60 of GlcT fused to an N-terminal hexahistidine sequence was linearized with BamHI. The oligonucleotides MM6 (5'-GATCTCTGGTTCCGCGTGGTTCCATGA) and MM7 (5'-GATCTCATGGAACCACGCGGAACCAGA) carrying a DNA fragment encoding a thrombin cleavage site were hybridized at 80°C and ligated to the linearized plasmid pGP114. Positive candidates were verified by sequencing.
Escherichia coli DH5 was used as host for overexpression of the recombinant proteins. Expression was induced by the addition of IPTG (final concentration 1 mM) to logarithmically growing cultures (OD600 = 0.8). The crude extracts were passed over Ni–NTA Superflow (Qiagen), followed by elution with an imidazole gradient. The Bio-Rad dye-binding assay was used to determine protein concentration. Bovine serum albumin was used as the standard. Purified His::GlcT-RBD protein was concentrated using a centriprep concentrator unit (Millipore). The GlcT-RBD peptidic fragment (7498 Da) was generated by thrombin cleavage (Pharmacia) of the His::GlcT-RBD protein according to the supplier’s instructions (Pharmacia). After thrombin cleavage the protein preparation was loaded onto a Superdex 75 prep grade HR16/60 column (Pharmacia) for size exclusion chromatography in 300 mM NaCl, 50 mM Na2HPO4 (pH 7.8). The purity of the protein was determined on 15% Tris–Tricine gels using Coomassie brilliant blue staining (36).
In vitro transcription
To obtain a template for in vitro synthesis of the wild-type RAT RNA, a 84 bp PCR product was generated using pGP66 (12) as template and primers IL59 (5'-CCAAGTA ATACGACTCACTATAGGACGTGTTACTGATTCG) and IL60 (5'-CAAGAATTGGGACAACTCTTCTTCTCCTTTT TTTTCCTCAATCACTCATGCC). To generate a template for in vitro transcription of mutant RAT RNAs, the primers OS25 (5'-CCAAGTAATACGACTCACTATAGGAATTCA GTTTATCCTTAT) and OS26 (5'-TTGAGGGAAAAAA ACGGGAAGTTC) were used to amplify a 99 bp PCR product. The presence of a T7 RNA polymerase recognition site in primers IL59 and OS25 (underlined) allowed the use of the PCR product as a template for in vitro transcription with T7 RNA polymerase (Roche Diagnostics). As a non-specific RNA, a 350 bp gapA transcript was prepared as described previously. The integrity of the RNA transcripts was analyzed by denaturating agarose gel electrophoresis (14).
Assay of interaction between GlcT-RBD and RAT RNA
Binding of GlcT-RBD to RAT RNA was analyzed by gel retardation experiments. The RAT RNA (in water) was denatured by incubation at 90°C for 2 min and renatured by dilution 1:1 with ice-cold water and subsequent incubation on ice. If required, non-specific RNA was renatured separately and mixed with the RAT RNA prior to protein addition. Purified GlcT-RBD was added to the RAT RNA and the samples were incubated for 10 min at room temperature. After this incubation, glycerol was added to a final concentration of 10% (w/v). The samples were then analyzed on 12% Tris–acetate polyacrylamide gels.
RNase T1 footprinting
The method used for RNase T1 digestion of the RAT RNA was as follows. Radiolabeled RNA was denatured by heating to 90°C and was subsequently allowed to cool down at room temperature. Diluted GlcT-RBD (7 pmol in 1 μl) was added to 5'-radiolabeled RAT RNA (140 000 c.p.m., gel purified) in the presence of 20 mM MgCl2 and 300 mM NaCl (final concentration). The reaction was incubated at 37°C for 10 min, after which 20 μg yeast tRNA (Sigma) and appropriate diluted RNase T1 (1 or 2 U) was added, followed by 15 min incubation at 37°C. The reaction was stopped by adding another 20 μg tRNA. The control reactions were performed in the absence of protein or RNase T1 under corresponding conditions. The RAT RNA was extracted with acid phenol/chloroform, precipitated with ethanol, resuspended and analyzed by electrophoresis.
Screening system to isolate random terminator mutants
Plasmids containing the mutated ptsG-RAT and a promoterless kanamycin resistance cassette were obtained as follows. The mutated ptsG-RAT alleles were amplified via PCR using the plasmids pGP329 and pGP330 as template DNA (obtained by site-directed mutagenesis of the RAT sequence; see above) and the oligonucleotides IL5 (17) and JS11 (12). The PCR products were digested with MfeI and BamHI and the 0.5 kb fragment was cloned into pAC6 (12) cut with the same enzymes. The resulting plasmids were designated pGP337 and pGP338. In a second step a promoterless aphA3 kanamycin resistance cassette (37) was amplified using the primers IL46 (5'-CGGGATCCTAATGTTAGAAAAGAGGAAGGA AATAA) and IL48 (5'-CGGGATCCCTACTAAAACAA TTCATCCAGTAA) and pAC7 as template DNA. The DNA fragment was cut with BamHI and cloned into the target vectors pGP337 and pGP338 linearized with the same enzyme. The resulting plasmids were designated pGP339 and pGP340, respectively. Moreover, by the same strategy we constructed plasmid pGP342 harboring a ptsG–aphA3 lacZ fusion under the control of the wild-type regulatory elements. These plasmids were used to introduce a ptsG-aphA3 lacZ fusion into the chromosome of B.subtilis 168 (see Table 1).
Liquid cultures of GP172 and GP173 containing the artificial ptsG–aphA3 lacZ operon were grown overnight in the absence of kanamycin. Aliquots of 100 μl of independently grown cultures were plated onto SP plates containing kanamycin at a final concentration of 30 and 60 μg/ml, respectively. The plates were incubated for 2 days at 28°C to select for colonies able to grow on kanamycin and express the ptsG–aphA3 lacZ fusion. To verify the localization of the selected mutations in the control region of the artificial operon, B.subtilis 168 was transformed with chromosomal DNA of the mutant strains and transformants were selected for chloramphenicol resistance. Transformants resistant to both chloramphenicol and kanamycin contained the mutation in close proximity to the cat resistance gene, which is located downstream of the lacZ gene. The ptsG promoter regions of these strains were analyzed by sequencing.
RESULTS
Mutational analysis of the ptsG terminator sequence
In a previous work, we demonstrated that a mutation disrupting the presumed stem structure of the transcriptional terminator in the ptsG leader region resulted in constitutive expression of ptsG (17). To ensure that it is the structure of the terminator rather than its sequence we constructed a strain, GP169, carrying a compensatory mutation in the terminator region. In this strain, the potential stem structure of the terminator is restored without correcting the sequence. The ptsG promoter region containing this mutation was inserted into the chromosome of B.subtilis in front of a promoterless lacZ gene. The synthesis of ?-galactosidase after growth of the strains in CSE minimal medium in the absence and presence of glucose was monitored to assay transcription driven by the wild-type and mutant ptsG promoter regions (Table 2). While transcription was inducible by glucose in the wild-type strain QB5448, constitutive expression was observed in the terminator mutant GP151. These results are in good agreement with previous findings (17). The newly constructed compensatory mutation restored the inducibility of ptsG expression and therefore the functionality of the transcriptional terminator. Thus, the structure of the terminator seems to be important rather than its actual sequence. The restoration of the terminator structure results in transcription termination in the absence of the inducer, glucose.
To provide further evidence for the functional role of the terminator, we developped a screening and selection system that allowed us to select for spontaneous mutations resulting in constitutive ptsG expression. Briefly, an artificial aphA3 lacZ operon was created and placed under control of the ptsG expression signal. The two strains used here (GP172 and GP173) carried mutations of the RAT sequence that prevented their recognition by GlcT. Mutant strains allowing expression of the artificial operon were selected as decribed in Materials and Methods. Several mutants exhibiting both resistance to kanamycin and constitutive expression of the lacZ gene (as monitored by the blue color of colonies on X-Gal plates) were obtained and their ptsG promoter regions upstream of the aphA3 lacZ operon were sequenced. The mutations all affected the transcriptional terminator in the ptsG leader region. A total of 12 different spontaneous mutations, six derived from each strain, were identified. Among these mutations were four deletions encompassing the terminator region and eight point mutations. One of these point mutations (G47T) was obtained with both original strains, whereas two different mutants obtained with either GP172 and GP173, affected the same position of the terminator (C75A and C75T, respectively). A summary of the sequence information concerning the mutations is shown in Figure 2. The effect of the mutations was quantified by assaying the ?-galactosidase activity driven by the mutant ptsG control elements. As expected, inducible expression was observed in the wild-type strain GP174, while no expression of the lacZ gene was detected in the mutant strains GP172 and GP173 that were used to select for terminator mutants. The mutations of the terminator all resulted in constitutive expression of the artificial aphA3 lacZ operon. However, the absolute level of expression was different for the individual mutants. The deletions and the mutation of G47 resulted in high constitutive expression (400–1700 U ?-galactosidase/mg protein; see Table 2). A point mutation at position C75 or G90 resulted in intermediate expression (100–200 U ?-galactosidase/mg protein), while the mutations of G38 and G72 resulted in weak constitutive expression (<100 U ?-galactosidase/mg protein; see Table 2).
Figure 2. Sequence of the ptsG promoter region. The RAT sequence is shown in bold and the terminator is boxed. The –10 promoter region and the transcription start site are shown in bold. The identified point mutations are shown in circles below the original nucleotides. Arrows with numbers indicate the deleted regions found in the sequenced mutants. The strain designations, the ‘numbering’ of the positions and the effects of the mutations are given in Table 2.
Taken together, our findings indicate that the terminator is the only negatively acting regulatory element in the ptsG promoter region. Moreover, it is the secondary structure rather than the actual sequence of the terminator which is important for the control of ptsG expression.
Binding of the RNA-binding domain of GlcT to the RAT sequence
Previous genetic and biochemical evidence suggested direct binding of GlcT to the RAT sequence that overlaps the terminator in the ptsG leader mRNA. To substantiate these observations and to study the molecular details of protein–RNA interaction, we first purified the N-terminal domain of GlcT devoid of a His tag. In a previous study, binding of His-tagged GlcT to the RAT RNA was demonstrated (17). However, subsequent gel retardation experiments suggested that the affinity tag might cause some non-specific RNA binding (our unpublished results). We therefore constructed a system that allowed removal of the His tag after initial purification of the protein. The purified protein was efficiently cleaved at the introduced thrombin cleavage site. The resulting protein was purified to apparent homogeneity by size exclusion chromatography as described in Materials and Methods.
The ability of the in vitro generated RNA-binding domain of GlcT to bind its RNA target was tested by electrophoretic mobility shift analysis. As shown in Figure 3, addition of the RNA-binding domain to the RAT RNA resulted in a shift of the apparent size of the RAT fragment. At a high concentration of the RNA-binding domain (3-fold molar excess), the RAT RNA was completely shifted (Fig. 3, lane 7). The shifted band contained a complex of both the RNA and the RNA-binding domain, as suggested by the differential detection of both components with ethidium bromide (detection of RNA) and Coomassie brilliant blue (detection of protein) (data not shown).
Figure 3. Electrophoretic mobility band shift analysis of the interaction between the RAT and the RBD. Lane 1 shows 125 pmol of the free RAT RNA. In lanes 2–7, 250 pmol RAT RNA and 30 pmol non-specific RNA were used. Increasing concentrations of RBD-GlcT were added to the RNA prior to electrophoresis. Aliquots of 250, 500, 750, 1000 and 1500 pmol RBD-GlcT were used. The arrows indicate the different RNA species.
Structural analysis of the GlcT–RAT interaction by RNase T1 protection footprint analysis
Specificity determinants have already been studied for the RAT sequences of the bgl and the sac type and the structure of the complex between the bglP RAT and the LicT antiterminator was determined (10,28). However, there is no cross-talk between the antiterminators of the Bgl and Sac types and the ptsG RAT sequence and GlcT is not capable of causing antitermination at bgl or sac RAT sequences (17). Thus GlcT and the ptsG RAT may be the most distant members of this large family of protein-dependent riboswitches. This prompted us to initiate an analysis of the molecular details of the interaction between the RNA-binding domain of GlcT and the ptsG RAT sequence by RNase T1 footprint analysis.
RNase T1 specifically cleaves adjacent to G residues in single-stranded regions of the RNA. Binding of the G residue to a protein or formation of a double-stranded RNA structure protects the RNA from degradation by RNase T1. As expected, incubation of denatured RAT RNA with RNase T1 resulted in RNA cleavage at each of the G residues in the RAT sequence (see Fig. 4, lane 2). Similarly, all G residues are available for cleavage by RNase T1 in the native form of the RAT RNA (see Fig. 4, lanes 3 and 5). Thus, the RAT does not seem to adopt a secondary structure in the absence of GlcT. In the presence of the RNA-binding domain of GlcT several G residues were protected from nucleolytic degradation, with only one G residue remaining accessible. This residue, G16, is probably located in the top loop of the RAT secondary structure and thus part of a single-stranded region in the absence and in the presence of GlcT. Among the residues protected by GlcT, G22, G26 and G28 were most strongly protected. According to the secondary structure model of the RAT RNA (see Figs 1 and 7), G22 and G26 are located in loop regions of the RAT RNA and may thus be protected by interaction with GlcT. G3, G11 and G21 were weakly protected in the presence of GlcT. This may reflect their location in a double-stranded region rather than interaction with GlcT. Indeed, G11 and G21 are both thought to be located in the third double-stranded region of the RAT structure. Thus, the footprinting data indicate that the RAT RNA exists in the single-stranded form in the absence of GlcT and adopts the double-stranded conformation upon GlcT addition. Moreover, GlcT seems to interact with residues in the two loops of the RAT structure.
Figure 4. RNase T1 footprint of the ptsG RAT region. Two different polyacrylamide concentrations were used to resolve the complete region: (A) 26%; (B) 20%. Weak hydrolysis of the RNA in the absence of the RNase and the antiterminator protein shows every nucleotide of the RAT (lane 1). High concentrations of RNase T1 under denaturing conditions leads to cleavage at every guanine nucleotide (lane 2). In the other assays, the RAT RNA and the RBD were used as indicated. In lanes 5 and 6, twice as much radiolabeled RNA was used as compared to lanes 3 and 4. Arrows indicate the position of the guanine nucleotides in the RAT.
Figure 7. Proposed model of the base pairing in the RAT RNA. The circled bases are proposed to directly interact with GlcT.
Probing of the RAT structure and of protein–RNA interactions by site-directed mutagenesis
Since only the guanine residues can be probed by RNase T1 footprinting, we attempted to study the role of the individual bases of the RAT sequence in more detail by site-directed mutagenesis. Mutations of the RAT sequence were introduced into the ptsG promoter region, located in front of a promoterless lacZ gene. If the mutations in the RAT sequence were expected to affect the structural integrity of the overlapping terminator, we introduced compensatory mutations in the terminator to restore the presumed wild-type secondary structure (see Fig. 1).
To assay the effects of the different mutations, the strains were grown in CSE minimal medium in the absence and presence of glucose and the ?-galactosidase activities were determined as an indicator of termination/antitermination activity. The results are summarized in Table 3. As observed previously, a strong induction of the ptsG–lacZ fusion was detected for the wild-type strain QB5448.
Mutations preventing formation of the first double-stranded region of the RAT RNA resulted in loss of induction (GP367, C2G; GP368, G28C). Interestingly, a strain with an inverse arrangement of the C and G residues (and an additional compensatory mutation in the terminator, GP366; see Table 3) resulted in constitutive expression rather than in restoration of inducibility. Constitutive expression is a strong indication of inactivity of the transcriptional terminator. Thus, the overall structure of the RAT/terminator region may have changed in such a way that transcription termination no longer occurs in the triple mutant GP366. To address this hypothesis, we assayed the activity of the ptsG control region present in GP366 in a glcT background. While no expression of the wild-type ptsG–lacZ fusion was observed in the glcT mutant GP109, constitutive expression was detected for the C2G G28C C85G triple mutant (data not shown). This result confirms that antitermination is not required to express ?-galactosidase in this construct. To test whether GlcT binds the RAT sequence with the inverse C:G base pair, we performed a gel mobility shift assay. As shown in Figure 5, this fragment was as efficiently bound by GlcT as the wild-type RAT fragment. In contrast, a fragment containing only the C2G mutation that interferes with formation of the lowest part of the stem was barely shifted by GlcT. Thus, formation of this stem is important for RNA binding by GlcT, whereas the actual sequence does not seem to be critical.
Figure 5. Electrophoretic mobility band shift analysis of the interaction between different mutant RAT species and the RBD.
Mutations affecting loop 1 of the RAT had different effects: two different mutations affecting the G3 residue resulted in complete loss of induction. This finding is in good agreement with the observed role of this residue in the interaction with GlcT as inferred from the RNase T1 protection experiment. Moreover, a RAT sequence containing the G3C mutation is not bound by GlcT in vitro (see Fig. 5). However, both mutations G3C and G3A may alternatively affect the structure of the RAT RNA, as both introduced bases would find pairing partners on the opposing side of the RAT (U25 or G26). Similarly, mutations affecting the residues U25 and A27 resulted in loss of antitermination. These mutations were not expected to alter the secondary structure of the RAT. However, the compensatory mutation for one of the mutations (U86C in GP381) results in loss of terminator activity and thus in constitutive expression of ?-galactosidase. Finally, the mutation G26U present in GP370 resulted in increased expression of the ptsG–lacZ fusion, both in the absence and presence of glucose. Here, the terminator may be leaky due to the compensatory mutation (A:U instead of G:C base pair) and allow some basal expression even in the absence of glucose. Alternatively, the perfectly symmetrical structure of the RAT generated by the G26U mutation might result in improved binding of GlcT, including some weak antitermination in the absence of glucose. Indeed, the mutated RAT seems to be an efficient target for GlcT, as judged from the clear induction of ?-galactosidase in this strain. In addition, the RAT RNA containing this mutation was efficiently bound by GlcT in vitro (see Fig. 5). Moreover, expression of this ptsG–lacZ fusion was completely lost in a glcT mutant strain (see Table 4). Taken together, our data indicate that loop 1 is of great importance for the proper interaction of the RAT RNA with GlcT.
Table 4. Effect of mutations in the ptsG RAT sequence on induction of a ptsG–lacZ fusion
One mutant affected in the second short double-stranded region of the RAT was studied. In this strain, GP369, the predicted G:C base pair was inverted to be C:G, and efficient induction of ?-galactosidase activity was detected. Additionally, the RAT RNA carrying the inversion interacted with the RNA-binding domain of GlcT in vitro (see Fig. 5). Thus, the structure rather than the actual nucleotide at this position is important for GlcT binding and antitermination. This is in good agreement with the observed weak protection of G5 from RNase T1 digestion in the presence of GlcT.
Several mutations in loop 2 resulted in loss of GlcT binding to the RAT and thus in loss of antitermination, namely those affecting U6 and G22. Results obtained with the ptsG–lacZ fusion (Table 3) and the in vitro RAT–GlcT binding assay (Fig. 5) are in perfect agreement for these mutations. In contrast, replacement of U7 and/or A8 showed no or only slight effects. The importance of G22 for efficient transcriptional antitermination is consistent with the strong RNase T1 protection obtained for this base in the presence of GlcT. However, similarly to loop 1, not all bases seem to play an important role in GlcT recognition.
As we approached the top of the proposed RAT structure, the mutations had lesser effects. The first base pair of the third double-stranded region is important for antitermination, as judged from lack of expression of the ptsG–lacZ fusion in the corresponding mutant strain GP178. An inversion of this base pair along with the required compensatory mutation in the terminator caused constitutive expression (GP372), thus we are unable to decide whether this inversion would affect antitermination. The constitutive expression of ?-galactosidase in this strain was not abolished by deletion of the glcT gene (data not shown), indicating that the mutation inactivated the terminator. An in vitro assay of GlcT binding to the mutant RAT suggested that an interaction occurred, albeit the complex seemed to be less stable (Fig. 5). A mutation of G11 had a mild effect whereas the A12U U13A double mutation present in GP166 did not influence the antitermination activity of GlcT. Similarly, three mutations in the top loop of the RAT did not affect induction by glucose and, therefore, binding of GlcT. These results indicate that the upper part of the RAT may only be required to obtain the correct RAT structure and is not involved in the interaction with GlcT.
Converting the recognition specificity of the RAT sequence to LicT binding
A comparison of RAT sequences of the glucose, sucrose, ?-glucoside and lactose classes revealed that all but the glucose RAT sequences contain 3 bp in the second double-stranded region whereas the ptsG RAT sequences of B.subtilis and S.carnosus contain only 2 bp (see Fig. 6). Moreover, loop 1 of the ptsG RAT is made up of four putatively non-paired bases whereas three non-paired bases form this loop in the other RAT sequences (12,28,29) (see Fig. 6). Our mutagenesis studies suggested an important role of RAT structure in the loop regions for the interaction with GlcT. We therefore asked whether this unique arrangement of the ptsG RAT is important for induction specificity by GlcT.
Figure 6. Proposed secondary structure for different RAT RNAs (28). The fold proposed for the ptsG-Ains RAT is based on similarity to the bglP and sacB RAT RNAs.
To address this question we constructed a B.subtilis strain with a ptsG–lacZ fusion with an additional adenine nucleotide between G3 and U4 of the RAT. This insertion is predicted to result in the formation of an extra A:U base pair in the second paired region. Concomitantly, the mutant loop 1 would contain three bases instead of four. Thus, the structure of this RAT is expected to be similar to those of the sucrose and ?-glucoside classes (see Fig. 6). Binding of antitermination proteins to this mutant RAT was assayed by determining the activity of the ptsG–lacZ fusion under conditions when the different antiterminator proteins are known to be active. In the wild-type strain QB5448, glucose, sucrose and salicin induced expression of the fusion. This finding is in agreement with previous reports and results from the activity of GlcT due to non-specific regulation by glucose permease (compare QB5448 and GP109 in Table 4) (12,17,38). In contrast, the mutant ptsG control region present in GP385 did not allow induction by glucose, suggesting that the altered RAT was not recognized by GlcT. However, significant induction was observed in the presence of sucrose or the ?-glucoside salicin. This induction was independent of a functional glcT gene (compare GP385 and GP386 in Table 4), suggesting that the mutant RAT was bound by a different antiterminator. To test this possibility, we constructed strains carrying the wild-type or mutant RAT sequences and deletions of the genes encoding the other B.subtilis antiterminators of the BglG family, i.e. sacT, sacY and licT. These deletions had no effect on the activity of a ptsG–lacZ fusion in the presence of the wild-type RAT sequence. This observation is in good agreement with the idea that induction of ptsG by sucrose and salicin is mediated by GlcT (17,38; see above). However, the lacZ fusion controlled by the mutant RAT was not induced by sucrose and salicin in the licT mutant strain GP390. Thus, LicT is essential to overcome transcription termination in the presence of the altered RAT sequence. Similarly, induction by sucrose is lost in the sacT sacY double mutant strain GP388. SacT and SacY recognize identical RAT sequences and therefore the effect of these mutations was not assayed individually. Interestingly, induction is still possible in the presence of salicin in GP388, confirming that this induction is caused by LicT.
DISCUSSION
As a target of regulation, RNA differs substantially from DNA in its structural diversity and in its ability to adopt alternative, mutually exclusive structures. Moreover, these structures, the riboswitches, are capable of interacting with a wide variety of regulatory partners, such as metabolites, other RNAs or proteins. Due to this huge variability, the themes in RNA–protein recognition are much more diverse than those identified in DNA–protein interactions.
GlcT and the ptsG riboswitch are members of an expanding family of transcription regulatory systems (22). A recent analysis of the evolution of the regulatory domains of the antitermination proteins of this family revealed that GlcT from B.subtilis and its ortholog from S.carnosus form a distinct subgroup whereas the antiterminators of the sucrose and ?-glucoside classes exhibit a close relationship with each other (39). This is supported by the finding that there is no cross-talk between the regulatory components of the glucose type on the one hand and those of the sucrose and ?-glucoside type on the other (17). In contrast, binding of sucrose antiterminator proteins to RAT sequences of bgl genes was reported (10).
Our studies regarding the overall structure of the ptsG RAT suggest that it folds similarly to the RAT RNAs of the bgl and sac classes, thus confirming a proposal of Yang et al. (28). Interestingly, this structure deviates from that obtained by the Zuker algorithm proposed earlier (17). However, both mutational and footprint analyses are in disagreement with the RAT structure suggested previously, whereas they confirm the model depicted here (see Figs 1, 6 and 7). Mutations in the proposed stem regions had different consequences: single mutations prevented antitermination while compensatory mutations that restored the base pairing restored GlcT binding at the same time. Similar results were obtained with the sacB RAT sequence (10).
While the sequence of the ptsG RAT differs substantially from those of the RAT sequences of the sac and bgl classes, we detected a perfect match of loop 2 of the ptsG, bglP and sacB RAT sequences and the regions surrounding loop 2 (see Fig. 6). It therefore seems safe to conclude that this region of the RAT is structurally identical to that of the bglP RAT as determined by NMR spectroscopy (28). In loop 2, G22 forms contacts with two bases, U6 and A8. Thus, only U7 extrudes from the stem structure (Fig. 7). Analysis of the LicT interaction with the bglP RAT revealed that G22 is not involved in direct interactions with the protein. However, the mutations of G22 and its partners studied in this work abolished antitermination. This confirms the critical importance of these residues for the structure of the RAT. The ptsG RAT is unique since it is the only RAT sequence with a symmetrical arrangement of the two loops. Therefore, it is tempting to speculate that G3 is involved in contacts with U25 and A27, with G26 being exposed. This idea is again supported by the mutational analysis: mutations disrupting the ‘ménage à trois’ in loop 1 strongly interfere with antitermination and are therefore expected to interfere with the RAT structure, as observed for loop 2.
Our studies concerning the interaction of GlcT with the RAT revealed that loops 1 and 2, as well as the stems, are critical for antitermination. This is in good agreement with previous findings for the SacY–sacB RAT and LicT–bglP RAT interactions (10,28). Our mutation and footprint analyses demonstrated that the top loop is important neither for adoption of the proper RAT structure nor for the interaction with GlcT. In contrast, the RNA-binding coat protein of bacteriophage MS2 and the phage N protein requires the top loop of their RNA targets for binding (40,41). As far as the paired regions are concerned, our analysis fails to identify the bases and amino acids in the RAT and GlcT, respectively, that are directly involved in the interaction. However, in our footprint analysis G26 in loop 1 was strongly protected from RNase T1 cleavage in the presence of GlcT, suggesting that it is contacted by the protein. The corresponding base in loop 2 was shown to be involved in interaction of the bglP RAT with several amino acids of LicT (28). We have analyzed mutations of these two exposed nucleotides in loop 1 (G26) and loop 2 (U7). In both cases a strong induction by glucose was observed. We may therefore propose that although these exposed bases are contact partners for GlcT, their actual nature does not seem to be crucial. This fits well with previous observations on the interaction of the bglP RAT and LicT (28).
The idea that the structure of the RAT rather than its sequence is important for recognition by GlcT is strongly supported by the observation that insertion of a single nucleotide that makes the RAT structure more similar to that of the bgl and sac RATs resulted in complete loss of GlcT-dependent induction. In contrast, the novel RAT is recognized by LicT even though the sequence remains quite different from that of the cognate LicT RAT sequences. On the other hand, a small sequence alteration with a probably important structural consequence results in complete loss of antitermination by GlcT. So far, it is known that SacT and SacY bind to RAT sequences of the bgl and sac classes whereas LicT (and its E.coli counterpart BglG) binds only to bgl RAT sequences (10,20). The structures of the RAT sequences of these two classes are very similar (see Fig. 6). Accordingly, the recognition specificity for LicT is determined in the sequence of the RAT, and mutations in loop 1 of the sacB RAT sequence allowed binding of BglG in addition to SacY. The effects of the corresponding mutations were cumulative (10). This is in agreement with the idea that SacY and LicT recognize RAT RNAs with the same structure. In contrast, we report here the first example of a complete change of partner specificity. A single mutation that changes the structure of the RAT RNA is sufficient for altered recognition and results in an all-or-nothing effect. Obviously, a RAT can be a member of the glc or the sac/bgl family, with no intermediate possible. The RNA-binding domains of LicT and GlcT are well conserved (for alignments see 17,28). Five amino acids of LicT were shown to be involved in the direct interaction with the internal loops of the RAT RNA (28). Of these, three (N10, G26 and F31) are conserved in all B.subtilis antiterminator proteins of this family. The residue K5 is conserved in all proteins but SacY. Only R27 is not conserved at all and may thus be involved in providing interaction specificity.
The evolution of regulation by adaptation of DNA-binding proteins to new DNA-binding sites is well documented. In the case of protein-dependent riboswitches, the question arises what was first: does the protein evolve to recognize an altered RNA or does the RNA adapt to protein variations? Our results hint at the second idea: one mutation in the RNA is sufficient to gain a novel regulatory specificity. In contrast, many mutations in the protein would be required to recognize an RNA target with a similar sequence but different structure. It will be interesting to study the molecular details of GlcT–RAT interaction by determining the structure of the complex.
ACKNOWLEDGEMENTS
Wolfgang Hillen is acknowledged for providing a stimulating environment. We are grateful to Matthias G?rlach and Matthias Stoldt for helpful discussions and Steffi Bachem, Shane Hanson and Ingrid Wacker for the help with some experiments. Ulf Gerth and Dominique Le Coq are acknowledged for the gift of strains. I.L. was the recipient of a personal stipend from the Boehringer-Ingelheim Foundation. This work was supported by grants from the Deutsche Forschungsgemeinschaft (through SFB 473) and the Fonds der Chemischen Industrie to J.S.
REFERENCES
Henkin,T.M. (2000) Transcription termination control in bacteria. Curr. Opin. Microbiol., 3, 149–153.
Stülke,J. (2002) Control of transcription termination in bacteria by RNA-binding proteins that modulate RNA structures. Arch. Microbiol., 177, 433–440.
Winkler,W.C. and Breaker,R.R. (2003) Genetic control by metabolite-binding riboswitches. Chem. Biochem., 4, 1024–1032.
Babitzke,P. and Gollnick,P. (2001) Posttranscription initiation control of tryptophan metabolism in Bacillus subtilis by the trp RNA-binding attenuation protein (TRAP), anti-TRAP and RNA structure. J. Bacteriol., 183, 5795–5802.
Switzer,R.L., Turner,R.J. and Lu,Y. (1999) Regulation of the Bacillus subtilis pyrimidine biosynthetic operon by transcriptional attenuation: control of gene expression by an mRNA-binding protein. Prog. Nucleic Acid Res. Mol. Biol., 62, 329–367.
Rutberg,B. (1997) Antitermination of transcription of catabolic operons. Mol. Microbiol., 23, 413–421.
Wilson,S.A., Wachira,S.J.M., Norman,R.A., Pearl,L.H. and Drew,R.E. (1996) Transcription antitermination regulation of the Pseudomonas aeruginosa amidase operon. EMBO J., 15, 5907–5916.
Chai,W. and Stewart,V. (1998) NasR, a novel RNA-binding protein, mediates nitrate-responsive transcription antitermination of the Klebsiella oxytoca M5al nasF operon leader in vitro. J. Mol. Biol., 283, 339–351.
Bae,W., Xia,B., Inouye,M. and Severinov,K. (2001) Escherichia coli CspA-family RNA chaperones are transcription antiterminators. Proc. Natl Acad. Sci. USA, 97, 7784–7789
Aymerich,S. and Steinmetz,M. (1992) Specificity determinants and structural features in the RNA target of the bacterial antiterminator proteins of the BglG/SacY family. Proc. Natl Acad. Sci. USA, 89, 10410–10414.
Stülke,J. and Hillen,W. (2000) Regulation of carbon catabolism in Bacillus species. Annu. Rev. Microbiol., 54, 849–880.
Stülke,J., Martin-Verstraete,I., Zagorec,M., Rose,M., Klier,A. and Rapoport,G. (1997) Induction of the Bacillus subtilis ptsGHI operon by glucose is controlled by a novel antiterminator, GlcT. Mol. Microbiol., 25, 65–78.
Tobisch,S., Zühlke,D., Bernhardt,J., Stülke,J. and Hecker,M. (1999) Role of CcpA in regulation of the central pathways of carbon catabolism in Bacillus subtilis. J. Bacteriol., 181, 6996–7004.
Ludwig,H., Homuth,G., Schmalisch,M., Dyka,F.M., Hecker,M. and Stülke,J. (2001) Transcription of glycolytic genes and operons in Bacillus subtilis: evidence for the presence of multiple levels of control of the gapA operon. Mol. Microbiol., 41, 409–422.
Fillinger,S., Boschi-Muller,S., Azza,S., Dervyn,E., Branlant,G. and Aymerich,S. (2000) Two glyceraldehyde-3-phosphate dehydrogenases with opposite physiological roles in a nonphotosynthetic bacterium. J. Biol. Chem., 275, 14031–14037.
Doan,T. and Aymerich,S. (2003) Regulation of the central glycolytic genes in Bacillus subtilis: binding of the repressor CggR to its single DNA target sequence is modulated by fructose-1,6-bisphosphate. Mol. Microbiol., 47, 1709–1721.
Langbein,I., Bachem,S. and Stülke,J. (1999) Specific interaction of the RNA-binding domain of the Bacillus subtilis transcriptional antiterminator GlcT with its RNA target, RAT. J. Mol. Biol., 293, 795–805.
Bachem,S. and Stülke,J. (1998) Regulation of the Bacillus subtilis GlcT antiterminator protein by components of the phosphotransferase system. J. Bacteriol., 180, 5319–5326.
Schmalisch,M., Bachem,S. and Stülke,J. (2003) Control of the Bacillus subtilis antiterminator protein GlcT by phosphorylation: elucidation of the phosphorylation chain leading to inactivation of GlcT. J. Biol. Chem., 278, 51108–51115.
Manival,X., Yang,Y., Strub,M.P., Kochoyan,M., Steinmetz,M. and Aymerich,S. (1997) From genetic to structural characterization of a new class of RNA-binding domain within the SacY/BglG family of antiterminator proteins. EMBO J., 16, 5019–5029.
Tortosa,P., Aymerich,S., Lindner,C., Saier,M.H.,Jr, Reizer,J. and Le Coq,D. (1997) Multiple phosphorylation of SacY, a Bacillus subtilis antiterminator negatively controlled by the phosphotransferase system. J. Biol. Chem., 272, 17230–17237.
Stülke,J., Arnaud,M., Rapoport,G. and Martin-Verstraete,I. (1998) PRD—a protein domain involved in PTS-dependent induction and carbon catabolite repression of catabolic operons in bacteria. Mol. Microbiol., 28, 865–874.
Amster-Choder,O. and Wright,A. (1992) Modulation of the dimerization of a transcriptional antiterminator protein by phosphorylation. Science, 257, 1395–1398.
Fux,L., Nussbaum-Schochat,A. and Amster-Choder,O. (2003) Interactions between the PTS regulation domains of the BglG transcriptional antiterminator from Escherichia coli. J. Biol. Chem., 278, 46203–46209.
Schnetz,K., Stülke,J., Gertz,S., Krüger,S., Krieg,M., Hecker,M. and Rak,B. (1996) LicT, a Bacillus subtilis transcriptional antiterminator of the BglG family. J. Bacteriol., 178, 1971–1979.
van Tilbeurgh,H., Manival,X., Aymerich,S., Lhoste,J.M., Dumas,C. and Kochoyan,M. (1997) Crystal structure of a new RNA-binding domain from the antiterminator protein SacY of Bacillus subtilis. EMBO J., 16, 5030–5036.
Declerck,N., Vincent,F., Hoh,F., Aymerich,S. and van Tilbeurgh,H. (1999) RNA recognition by transcriptional antiterminators of the BglG/SacY family: functional and structural comparison of the CAT domains from SacY and LicT. J. Mol. Biol., 294, 389–402.
Yang,Y., Declerck,N., Manival,X., Aymerich,S. and Kochoyan,M. (2002) Solution structure of the LicT-RNA antitermination complex: CAT clamping RAT. EMBO J., 21, 1987–1997.
Knezevic,I., Bachem,S., Sickmann,A., Meyer,H., Stülke,J. and Hengstenberg,W. (2000) Regulation of the glucose specific phosphotransferase system (PTS) of Staphylococcus carnosus by the antiterminator protein GlcT. Microbiology, 146, 2333–2342.
Sambrook,J., Fritsch,E.F. and Maniatis,T. (1989) Molecular Cloning: A Laboratory Manual, 2nd Edn. Cold Spring Harbor Laboratory Press, Cold Spring Harbor, NY.
Faires,N., Tobisch,S., Bachem,S., Martin-Verstraete,I., Hecker,M. and Stülke,J. (1999) The catabolite control protein CcpA controls ammonium assimilation in Bacillus subtilis. J. Mol. Microbiol. Biotechnol., 1, 141–148.
Kunst,F. and Rapoport,G. (1995) Salt stress is an environmental signal affecting degradative enzyme synthesis in Bacillus subtilis. J. Bacteriol., 177, 2403–2407.
Weinrauch,Y., Msadek,T., Kunst,F. and Dubnau,D. (1991) Sequence and properties of comQ, a new competence regulatory gene of Bacillus subtilis. J. Bacteriol., 173, 5685–5693.
Bachem,S., Faires,N. and Stülke,J. (1997) Characterization of the presumptive phosphorylation sites of the Bacillus subtilis glucose permease by site-directed mutagenesis: implication in glucose transport and catabolite repression. FEMS Microbiol. Lett., 156, 233–238.
Bi,W. and Stambrook,P.J. (1998) Site-directed mutagenesis by combined chain reaction. Anal. Biochem., 256, 137–140.
Sch?gger,H. and von Jagow,G. (1987) Tricine-sodium dodecyl sulfate-polyacrylamide gel electrophoresis for the separation of proteins in the range from 1 to 100 kDa. Anal. Biochem., 166, 368–379.
Trieu-Cuot,P. and Courvalin,P. (1983) Nucleotide sequence of the Streptococcus faecalis plasmid gene encoding the 3'-5'-aminoglycoside phosphotransferase type III. Gene, 23, 331–341.
G?rke,B., Fraysse,L. and Galinier,A. (2004) Drastic differences in Crh and HPr synthesis reflect their different impacts on catabolite repression in Bacillus subtilis. J. Bacteriol., 186, 2992–2995.
Greenberg,D.B., Stülke,J. and Saier,M.H.,Jr (2002) Domain analysis of transcriptional regulators bearing PTS regulatory domains. Res. Microbiol., 153, 519–526.
Grahn,E., Stonehouse,N.J., Murray,J.B., van den Worm,S., Valeg?rd,K., Fridborg,K., Stockley,P.G. and Liljas,L. (1999) Crystallographic studies of RNA hairpins in complexes with recombinant MS2 capsids: implications for binding requirements. RNA, 5, 131–138.
Legault,P., Li,J., Mogridge,J., Lewis,K.E. and Greenblatt,J. (1998) NMR structure of the bacteriophage N peptide/boxB RNA complex: recognition of a GNRA fold by an arginine-rich motif. Cell, 93, 289–299.
Krüger,S. and Hecker,M. (1995) Regulation of the putative bglPH operon for aryl-?-glycoside utilization in Bacillus subtilis. J. Bacteriol., 177, 5590–5597.
LeCoq,D., Lindner,C., Krüger,S., Steinmetz,M. and Stülke,J. (1995) New ?-glucoside (bgl) genes in Bacillus subtilis: the bglP gene product has both transport and regulatory functions similar to those of BglF, its Escherichia coli homolog. J. Bacteriol., 177, 1527–1535.(Oliver Schilling1,2, Ines Langbein2, Mic)