Genetic Regulation of the yefM-yoeB Toxin-Antitoxin Locus of Streptococcus pneumoniae

ABSTRACT Type II (proteic) toxin-antitoxin systems (TAS) are ubiquitous among bacteria. In the chromosome of the pathogenic bacterium Streptococcus pneumoniae, there are at least eight putative TAS, one of them being the yefM-yoeBSpn operon studied here. Through footprinting analyses, we showed that purified YefMSpn antitoxin and the YefM-YoeBSpn TA protein complex bind to a palindrome sequence encompassing the −35 region of the main promoter (PyefM2) of the operon. Thus, the locus appeared to be negatively autoregulated with respect to PyefM2, since YefMSpn behaved as a weak repressor with YoeBSpn as a corepressor. Interestingly, a BOX element, composed of a single copy (each) of the boxA and boxC subelements, was found upstream of promoter PyefM2. BOX sequences are pneumococcal, perhaps mobile, genetic elements that have been associated with bacterial processes such as phase variation, virulence regulation, and genetic competence. In the yefM-yoeBSpn locus, the boxAC element provided an additional weak promoter, PyefM1, upstream of PyefM2 which was not regulated by the TA proteins. In addition, transcriptional fusions with a lacZ reporter gene showed that PyefM1 was constitutive albeit weaker than PyefM2. Intriguingly, the coupling of the boxAC element to PyefM1 and yefMSpn in cis (but not in trans) led to transcriptional activation, indicating that the regulation of the yefM-yoeBSpn locus differs somewhat from that of other TA loci and may involve as yet unidentified elements. Conservation of the boxAC sequences in all available sequenced genomes of S. pneumoniae which contained the yefM-yoeBSpn locus suggested that its presence may provide a selective advantage to the bacterium.

Detection of gene variations and regulatory circuits in microbial systems is critical for our knowledge of the evolvability of bacterial species, with one of the driving forces underlying bacterial evolution being stressful environmental conditions (2). Important operative pieces intervening in bacteria that cope with stress are the so-called toxin-antitoxin systems (TAS). Among them, the most studied are those belonging to the type II (proteic) TAS. These usually comprise two cotranscribed genes that encode an unstable antitoxin and a stable toxin. The antitoxin binds to and neutralizes its cognate toxin through protein-protein interactions. Under environmental stress conditions which cause the induction of endogenous proteases, the balance between toxin and antitoxin is shifted and the toxin is released from the TA complex due to proteasemediated degradation of the antitoxin. Thus, the unbound toxin is now free to exert its effect of poisoning the cell machinery so that growth is arrested (8,10,54).
The discovery of the mazEF (or chpAB) chromosomally encoded TAS in Escherichia coli (28) led to the proposal that, when encoded in chromosomes, these loci could function as mediators of programmed cell death (1,9). Unfavorable growth conditions could trigger this pathway, and as a consequence, a subpopulation of bacterial cells would die. Death of these cells would do the following: (i) provide food for the remaining population (altruistic behavior), (ii) serve as a defense mechanism to restrict phage spreading (protection against incoming DNA), and (iii) act as a mechanism to eliminate cells with deleterious mutations (preservation of the gene pool). It would seem that at least in the case of mazEF, cell death is a population-dependent phenomenon requiring a quorum-sensing molecule, termed extracellular death factor, which is a linear pentapeptide (NNWNN) important for mazEF-mediated killing activity (18). Another line of independent investigations led to the different view that TAS would function as modulators of the global levels of translation during environmental stress and that the toxin-mediated inhibition of protein synthesis led to reversible cell stasis rather than cell death (4,38). In the case of the pneumococcal RelBE2 Spn TAS, it was shown that growth cessation could be made irreversible if exposure to the toxin was prolonged for periods longer than 4 h, so that the toxin-exposed cells could not be rescued by triggering antitoxin synthesis (35)(36). Other hypotheses to explain the ubiquity and redundancy of the bacterial TAS (37) include, in addition to the above, a role of TAS in persistence, a process that enables a fraction of a bacterial population to survive prolonged exposure to antibiotics (i.e., antibiotic tolerance) by entering a state of dormancy (15,19,56), and also as a way to protect the cell population against DNA loss (53). Perhaps the functions of TAS cannot be generalized since they may have multiple biological roles that are dependent on the nature of the toxin, their location on the genome, and other yet-to-be-discovered factors.
Concerning regulation of the TAS, it has been shown that in most cases expression of TAS seems to be controlled by selfregulatory transcriptional mechanisms in which the antitoxin would act as a repressor by binding to an operator site that overlaps the promoter. In most instances, the cognate toxin acts as a very efficient corepressor of transcription (10-11, 26, 43, 51). An exception can be found in the case of the plasmid pSM19035 of Streptococus pyogenes, in which the expression of the epsilon-zeta (ε) TA operon is regulated by the global plasmid-encoded Omega () transcriptional repressor (22,55).
Our laboratories have been studying some of the TAS encoded by the chromosome of the Gram-positive bacterium Streptococcus pneumoniae (the pneumococcus), an important human pathogen responsible for over 2,000,000 human deaths per year (25). Sequence analyses performed on the available pneumococcal genomes have shown that the bacterium harbors at least eight TAS homologues: relBE1 Spn , relBE2 Spn , yefM-yoeB Spn , higAB, phd-doc, pezAT, tasAB, and hicAB (36). Of these, three TAS, namely, relBE2 Spn , yefM-yoeB Spn , and pezAT, have been studied and characterized as bona fide TAS, whereas relBE1 Spn was shown to be not functional (16,(34)(35). This number of pneumococcal TAS could be an underestimation, since database mining of sequenced microbial genomes indicates an abundance of TA loci, particularly in free-living microorganisms (37), and the number of reported TAS has been steadily increasing (24). A database containing most known and predicted TAS (48) is available at http://bioinfo -mml.sjtu.edu.cn/TADB/.
Here we present a study on the regulation of the pneumococcal yefM-yoeB Spn locus, which is transcribed as a single unit. Through DNA binding and footprinting assays, in conjunction with transcriptional fusions, we have demonstrated that control of expression of this operon is at three levels. First, as in the case of other TAS, the YefM Spn antitoxin can bind to and repress transcription from its own promoter, P yefM2 . Second, the YoeB Spn toxin exerts further repression of transcription from the same promoter both in cis and in trans by binding to the YefM Spn antitoxin. Third and most distinctly, we have detected the insertion of a pneumococcal BOX element (comprising boxA and boxC subelements) upstream of P yefM2 , which has led to the generation of a new and YefM-YoeB Spn -unregulated promoter, P yefM1 . The BOX was conserved in all sequenced pneumococcal strains but absent in other yefM-yoeB homologues of different bacteria. Thus, BOX seems to have contributed to introducing a new level of complexity to the regulation of the yefM-yoeB Spn locus by increasing the overall basal transcription level. BOX elements are repeated sequences (up to 125 copies per genome) which have been associated with genetic competence, virulence (17), and phenotypic phase variation (45) in S. pneumoniae. Although little is known on their contribution to the genetics of S. pneumoniae (6), it would appear that the hyperrecombinogenic nature of S. pneumoniae (12) may provide the bacterium with selective advantages when subjected to stressful conditions.

MATERIALS AND METHODS
Bacterial strains, transformation, and growth conditions. E. coli Top10 [F Ϫ mcrA ⌬(mrr-hsdRMS-mcrBC) 80lacZ⌬M15 ⌬lacX74 recA1 araD139 ⌬(araAleu)7697 galU galK rpsL (Str r ) endA1 nupG] (Invitrogen) was used as the host for cloning experiments. E. coli BL21-CodonPlus(DE3)-RIL [E. coli B F Ϫ ompT hsdS(rB Ϫ mB Ϫ ) dcm ϩ Tet r gal (DE3) endA Hte [argU ileY leuW Cam r ] (Stratagene) was used as the host for the overexpression of genes cloned into the pET28a expression vectors. E. coli cultures were grown in Luria-Bertani medium (Difco) at 37°C. Liquid cultures were grown in an orbital shaker incubator at 250 rpm. When necessary, growth medium for E. coli was supplemented with antibiotics at the following concentrations: ampicillin, 100 g/ml; chloramphenicol, 34 g/ml; and kanamycin, 50 g/ml. S. pneumoniae R6 cells (21) were grown in AGCH medium (20) supplemented with 1% sucrose and 0.25% yeast extract. All S. pneumoniae cultures were grown at 37°C. DNA manipulations, sequencing, and sequence data analysis. DNA manipulations and other molecular biology techniques were carried out using established protocols (46). Genomic DNA from S. pneumoniae was prepared using the Wizard Genomic DNA purification kit (Promega). Plasmid DNA from recombinant E. coli was isolated using the Wizard SV miniprep kit (Promega). DNA fragments and PCR products were purified using the GFX purification kit from Amersham Biosciences. DNA sequencing was carried out using the BigDye Terminator cycle sequencing ready reaction kit and an ABI Prism 377 DNA sequencer (Applied Biosystems Inc.). DNA sequences were compiled and analyzed using the Lasergene sequence analysis software program (DNAStar). Comparison of nucleotide and amino acid sequence data were carried out using BLAST at the NCBI website (http://www.ncbi.nlm.nih.gov/BLAST). The secondary structures of the boxAC element, the putative transcription terminators of the gene upstream of yefM Spn , and sequences immediately after the yoeB Spn stop codon were predicted using the mfold server (http://mfold.rna.albany.edu/). RNA preparation and RT-PCR. Total RNA samples were prepared from 10 ml S. pneumoniae cultures using an RNeasy Midi kit (Qiagen) according to the manufacturer's procedure. The total RNA concentrations were determined by UV spectrophotometry, and the samples were checked for their integrity by analysis in a 1% agarose gel. Total RNA (400 ng) and 10 M gene-specific primer were mixed and heated for 5 min at 70°C and then placed on ice. This mixture was added to a 20 l reverse transcriptase (RT) reaction mixture containing 4 l of cDNA synthesis buffer, 5 mM dithiothreitol (DTT), 40 U RNase OUT, 1 mM deoxynucleoside triphosphate (dNTP) mix, and 15 U Thermo-Script RT (Invitrogen) and incubated for 60 min at 50°C. Reverse transcription was terminated by incubation of the mixtures for 5 min at 85°C. The RNA template was removed by addition of 2 U of RNase H and incubated at 37°C for 20 min. An aliquot (10%) of the cDNA synthesis reaction was used for PCRs, which were carried out using Phusion DNA polymerase (Finnzymes) and the relevant genespecific primers. PCR conditions were as follows: denaturation, 10 s at 98°C; annealing, 20 s at 53°C with an extension time of 20 s at 72°C, 30 cycles of amplification, and a final extension time of 10 min at 72°C. The following gene-specific primers were used for the RT-PCR analysis: yefMNHis, yefMC, GyoeBN, and GyoeBC (see Table S2 in the supplemental material).
Primer extension analyses. Total RNA was isolated from S. pneumoniae R6 using an Aurum total RNA minikit (Bio-Rad Laboratories). The primer extension assays were performed as described previously (39), by annealing, at 65°C for 5 min, the total RNA with two different specific primers, located 8 nucleotides (nt) (yefM-far) or 90 nt (yefM-near) (see Table S2 in the supplemental material) downstream of the yefM Spn ATG start codon, respectively. A mixture of dATP, dGTP, dTTP (100 M each), and 10 M [␣-32 P]dCTP was then added. The primers were then extended with ThermoScript reverse transcriptase (Invitrogen) at 50°C or at 55°C for 30 min. After that, 80 l of Tris-EDTA (TE) was added, followed by phenol-chloroform extraction and then ethanol precipitation. The pellet was dissolved in 8 l of stop solution (95% formamide, 20 mM EDTA, 0.05% bromophenol blue, and 0.05% xylene cyanol) and loaded onto an 8% urea polyacrylamide gel (7 M) along with the dideoxy sequencing DNA ladder. The dideoxy sequencing DNA ladder was obtained using Sequenase version 2.0 DNA polymerase (United States Biochemical) by annealing the yefM-near or the yefM-far primer to the pMBH13 (34) denatured plasmid DNA, respectively.
Construction of lacZ transcriptional fusions and use of ␤-galactosidase assays to detect promoter activity. The promoter-probe vector pQF52 (30) was used to construct lacZ transcriptional fusions for the detection of promoter activity ( Table 1). The pLNBAD plasmid, which is a derivative of plasmid pLN135 (7), was used to construct recombinant plasmids for overexpression of YeM Spn and YefM-YoeB Spn under the control of the arabinose-inducible P BAD promoter. Details of the construction of all recombinant plasmids as well as the primer sequences are listed in Table 2 and in Table S2 in the supplemental material. VOL. 193, 2011 PNEUMOCOCCAL yefM-yoeB TOXIN-ANTITOXIN REGULATION Recombinant pQF52 and pLNBAD plasmids were transformed into E. coli Top10, and transformants were assayed for ␤-galactosidase activity according to the method of Miller (31) using SDS and chloroform to permeabilize the cells. The ␤-galactosidase assay for each construct was repeated eight times, and the mean value obtained was used for further analysis.
Overexpression and purification of recombinant (His) 6 -YefM Spn and YoeB in E. coli. The YefM Spn antitoxin protein was obtained from pEMH10 (34), whereas the YefM-YoeB Spn TA complex was obtained from the recombinant plasmid constructed in this study (pET28a_HisYefMYoeB). To construct a recombinant plasmid with yefM-yoeB Spn , this operon was amplified using the primer pairs pETyefM-F and pETyoeB-R (Table 2; see also Table S2 in the supplemental material), and the amplified fragment was digested with NdeI and BamHI prior to ligation into the NdeI-BamHI site downstream of the isopropyl-␤-D-thiogalactopyranoside (IPTG)-inducible T7 promoter of the pET28a plasmid (Novagen). Both recombinant plasmids were transformed separately into E. coli BL21-CodonPlus(DE3)-RIL (Stratagene). To obtain adequate amounts of protein for this study, 8 ml of an overnight culture of E. coli BL21-Codon-Plus(DE3)-RIL harboring the pEMH10 or pET28a_HisYefMYoeB recombinants (Table 2), was diluted into 800 ml of fresh Luria-Bertani broth supplemented with 50 g/ml kanamycin and 34 g/ml chloramphenicol and allowed to grow until reaching an optical density at 600 nm (OD 600 ) of ϳ0.4 with shaking at 250 rpm and at 37°C. The cell cultures were cooled to 30°C before IPTG was added to a final concentration of 0.25 mM. Rifampin was added (to a final concentration of 200 g/ml) 15 min after addition of IPTG. After 2 h, the cells were harvested at 6,500 ϫ g at 4°C for 20 min. The cell pellets were then resuspended in 40 ml of 1ϫ His buffer (10 mM Tris, pH 7.6, 1 M NaCl, 5 mM ␤-mercaptoethanol, and 5% glycerol) containing 10 mM imidazole and 10 l/ml protease inhibitor mix (General Electric Healthcare). All purification steps were performed at 4°C. The cell suspension was then subjected to a French press (Constant System) twice, and the supernatant was separated from the cell debris and unbroken cells by centrifugation at 17,000 ϫ g for 20 min. The crude lysate was then loaded into a chromatography column packed with HIS-Select nickel affinity gel (Sigma) with a flow rate of 45 ml/h. HIS-Select nickel affinity gel is an immobilized metal-ion affinity chromatography product which is a quadridentate chelate on 6% beaded agarose charged with nickel and is selective for recombinant proteins with histidine tags and exhibits low nonspecific binding of other proteins. The column was then extensively washed with 100 ml of 1ϫ buffer A (10 mM Tris, pH 7.6, 0.3 M NaCl, 5 mM ␤-mercaptoethanol, and 5% glycerol) containing 10 mM imidazole with a flow rate of 45 ml/h. The histidine-tagged YefM Spn and YefM-YoeB Spn proteins were eluted from the column using 1ϫ His buffer with 250 mM imidazole at a flow rate of 45 ml/h. The protein fractions were collected and analyzed using 16% SDS-PAGE. One band was detected on the SDS-PAGE gel for the pEMH10 construct (data not shown), which corresponded to the expected size of (His) 6 -YefM Spn. On the other hand, two distinct protein bands were observed for the pET28a_HisYefMYoeB construct, which corresponded to the expected sizes of the (His) 6 -YefM Spn fusion protein and YoeB Spn (data not shown), indicating that under native conditions, both YefM Spn and YoeB Spn were copurified. These purified proteins were used for subsequent electrophoretic mobility shift assays (EMSA) and footprinting assays. Determination of the N-terminal sequence of both proteins was done by the Edman degradation method at the Protein Chemistry Facility of the Centro de Investigaciones Biológicas. EMSA, DNase I, and hydroxyl radical footprinting assays. DNA fragments containing the entire upstream sequence were PCR amplified using each of the [␥-32 P]ATP-labeled oligonucleotides and nonlabeled oligonucleotides: labeled PS1-F and non-labeled PS2-R (see Table S2 in the supplemental material), as well as nonlabeled PS1-F and labeled PS2-R. EMSA was carried out in the following mixture to a final volume of 5 l: 1ϫ binding buffer (100 mM Tris, pH 7.6, 5 mM EDTA, 5 mM DTT, and 5% glycerol), 3,000 cpm of labeled DNA, 10 ng/l of heparin, and increasing concentrations of either YefM Spn or YefM-YoeB Spn . The reaction mix was incubated for 20 min at room temperature before being loaded onto a 5% polyacrylamide gel and run with 0.5ϫ Tris-borate-EDTA (TBE) buffer. The gel was then transferred to a film cassette and exposed to an X-ray film overnight at Ϫ70°C. The film was then developed according to the manufacturer's instructions. After the binding of the proteins to the DNA fragments was detected, the DNA fragments were subsequently subjected to a DNase I footprinting assay. Preliminary results showed that the YefM Spn protein or the YefM-YoeB Spn protein complex was bound only to nucleotides within the palindrome sequence (PS). Therefore, another set of primers, PS2-F (see Table  S2) and PS2-R, was used to obtain the DNA fragment to repeat the EMSA and was also subsequently used for the DNase I and hydroxyl radical footprinting assays.
DNase I footprinting assays were carried out using the following reaction mix: 1ϫ binding buffer (400 mM Tris-Cl, pH 8, 60 mM MgCl 2 , 10 mM CaCl 2 , and 5% glycerol), 30,000 cpm labeled DNA, 10 ng/l of heparin, increasing concentrations of YefM Spn or YefM-YoeB Spn , and sterile distilled water to a final volume of 50 l. The reaction mix was incubated for 20 min at room temperature before adding 0.04 U DNase I (Roche) and then was incubated for another 5 min at room temperature. The reaction was stopped by adding 25 l of stop buffer (2 M ammonium acetate, 0.12 M EDTA, 0.8 M NaAc, pH 7.0, and 400 l/ml tRNA) and 187 l absolute ethanol, followed by freezing at Ϫ70°C for 30 min. Samples were centrifuged, and the resulting pellet was then washed with 70% chilled ethanol. The pellet was then air dried and dissolved in 3 l of loading buffer (80% deionized formamide, 10 mM NaOH, 0.1% bromophenol blue, 0.1% xylene cyanol, and 1 mM EDTA, pH 8). The samples were heated at 80°C for 3 min and loaded onto an 8% urea polyacrylamide gel (7 M), which was preheated to 50°C and run at 45 W in 1ϫ TBE buffer at 50°C. The sequencing ladder prepared using a dideoxy sequencing reaction was run along with the reaction samples. The gel was then vacuum dried for 45 min at 80°C and exposed to X-ray film in Ϫ70°C until the desired image intensity was obtained.
The hydroxyl radical footprinting reagent was prepared by mixing equal volumes of 0.6% H 2 O 2 , 20 mM sodium ascorbate, and Fe 2ϩ -EDTA (equal volumes of 0.4 mM Fe 2ϩ in sterile distilled water and 0.8 mM EDTA). First, 30,000 cpm of labeled DNA was incubated with increasing amounts of either YefM Spn or YefM-YoeB Spn in 1ϫ binding buffer, 10 ng/l of heparin, and sterile distilled water (added to a final volume of 50 l). Then, the reaction mix was incubated for 20 min at room temperature to allow the binding of the proteins to the DNA fragments. The reaction was started by adding 9 l of the hydroxyl radical footprinting reagent, and the mixture was incubated at room temperature for another 2 min. The reaction was then stopped by adding 14.7 l of stop solution (0.041 M thiourea, 1.5 M NaAc, pH 6, and 0.68 mg/ml tRNA) and 187 l of absolute ethanol, followed by freezing at Ϫ70°C for 30 min. The mixture was then centrifuged, and the pellet was washed with chilled 70% ethanol. The samples were then treated as described in the DNase I footprinting assays. The sequencing ladder prepared using the Maxam and Gilbert reaction for GϩA (29) was run along with the reaction samples.

Bioinformatics analysis of the S. pneumoniae-encoded yefM-yoeB Spn locus.
To determine whether the presence of the yefM-yoeB Spn locus was universal in the genomes of pneumococcal strains, BLASTP analyses were performed using the 22 annotated pneumococcal genomes in the NCBI database. The re- sults showed that 7 strains that harbor solo YefM Spn antitoxin without its YoeB Spn toxin counterpart were all mutated or truncated at the 3Ј terminus, whereas no solo YoeB Spn homologues were identified. Only 15 strains with both YefM Spn and YoeB Spn intact TA pairs were further analyzed (see Table S1 in the supplemental material). In all of them, the yefM Spn stop codon was separated by 3 nt from the start codon of yoeB Spn (see below). We were also able to resolve a discordant result, since YefM Spn was annotated in the NCBI database to initiate from different start codons among the S. pneumoniae strains. We purified the YefM Spn protein and determined the sequence of the first 10 residues to be MEAVLYSTFR, demonstrating that the protein initiates unambiguously from the ATG codon as annotated in the TIGR4 strain (see Fig. S1). Disregarding the discrepancy in the annotation, the translated amino acid sequences of YefM Spn among the strains were completely identical except for the 77th residue of YefM Spn , which was either a Thr or an Ile (see Fig. S1). All strains shared 100% identity in the yoeB Spn translated amino acid sequences except strain JJA, which showed 96% identity (see Fig. S2). The start codon of yoeB Spn among all the strains was also identical, and the first 10 residues (MLLKFTEDAW) were confirmed by N-terminal sequencing (see Fig. S2). A sequence analysis of the DNA region upstream of the coding sequence of yefM Spn predicted the presence of two putative 70 -dependent promoters 30 nt apart, designated P yefM1 and P yefM2 , which were experimentally shown to exist (see below). In addition, a putative ribosomebinding site (rbs) (5Ј-AcGAGG-3Ј; lowercase letters indicate deviations from the consensus sequence), located 7 nt upstream of the ATG start codon, was observed (Fig. 1A). The sequence of promoter P yefM2 matched with those of other pneumoccocal promoters (44) and showed an almost perfect consensus of the Ϫ35 (5Ј-cTGACA-3Ј) and Ϫ10 (5Ј-TATAA T-3Ј) regions separated by a 17-nt spacer (Fig. 1A). In contrast, the Ϫ35 (5Ј-TTGctc-3Ј) and Ϫ10 (5Ј-TATAAa-3Ј) regions of the P yefM1 promoter were more dissimilar to the consensus, and these promoter elements presented a spacer of 16 nt instead of 17 nt, indicating a suboptimal promoter sequence (Fig. 1A). An imperfect 44-bp palindrome sequence (designated PS) included the Ϫ35 region of P yefM2 and was centered 37 bp upstream of the transcription start site of the promoter P yefM2 (Fig. 1A and B) (see below). While examining the genetic structure around the yefM-yoeB Spn locus, we detected the presence of a BOX element (Fig. 1A) (coordinates 1567052 to 1567160; accession no AE007317) which was placed in the intergenic region between yefM Spn and the upstream gene in the pneumococcal R6 strain. The translated product of this upstream gene (coordinates FIG. 1. The yefM-yoeB Spn locus harbors two promoters. (A) Nucleotide sequence of the pneumococcal yefM upstream region along with the first 71 nt of the yefM coding sequence. The P yefM1 and P yefM2 promoters with their respective Ϫ10 and Ϫ35 sequences are indicated within boxes, as well as the putative rbs. The incomplete palindrome sequence, termed "PS," is also depicted, with the center of the palindrome indicated by the apex of a filled triangle. The boxA subelement is indicated in red, whereas the boxC subelement is indicated in blue. The start codon of yefM Spn is underlined. Asterisks denote the two transcriptional start sites of yefM-yoeB Spn as determined by primer extension. (B) Transcriptional start sites of the yefM-yoeB Spn operon as determined by primer extension. A yefM-near (N) or yefM-far (F) specific primer was used to anneal to the RNA samples prepared from S. pneumoniae R6. The primer was extended, yielding transcripts of 92 nt at 50°C and 115 nt at 55°C. The C, G, T, and A DNA sequencing ladders were obtained using the dideoxy-mediated chain termination sequencing method by annealing the yefM-near or yefM-far primer to pEMBH13 (34).

VOL. 193, 2011
PNEUMOCOCCAL yefM-yoeB TOXIN-ANTITOXIN REGULATION 4617 1567200 to 1568297) showed homology to nucleotidyltransferases, and a possible transcription terminator was located at its 3Ј end. The BOX element was placed immediately downstream of this, followed by the 44-bp PS (see above) (Fig. 1A). BOX was not annotated in the other 21 available strains of S. pneumoniae, but BLASTN alignments showed that its sequence remained highly conserved, along with the putative promoters and rbs, among all the pneumococcal stains studied (see Fig. S3 in the supplemental material). The arrangement of the genes neighboring the yefM-yoeB Spn locus appeared to be similar among all these strains. Furthermore, insertion of BOX apparently generated the P yefM1 promoter, in which the Ϫ35 sequence and 5 nt (5Ј-TATAA-3Ј) of the Ϫ10 sequence were located within the area 3Ј proximal to the BOX element (Fig.  1A). Interestingly, we found no indication of BOX elements upstream of the coding sequences of the other identified pneumococcal TA loci. BOX elements are abundant repeated sequences which are placed apparently at random within intergenic regions of the pneumococcal genome. They are present only in S. pneumoniae and closely related species, such as Streptococcus mitis, and they have been proposed to affect the expression of their neighboring genes (27). Generally, a BOX element is formed by the juxtaposition of three subelements, designated boxA (59 bp), boxB (45 bp), and boxC (50 bp), with boxB placed between boxA and boxC (6,27). In the BOX element that was found upstream of yefM Spn , boxB was absent, and boxA was located immediately upstream of boxC; this BOX element was thus designated boxAC. A closer inspection of these sequences revealed that the sequence 1567053 to 1567161 of the sequence under accession no. AE007317 (which includes 1 nt upstream and excludes 1 nt downstream of the annotated boxAC in the R6 strain) more closely match the BOX element previously described (17,27) in terms of sequence similarity and number of nucleotides compared to the database annotation. Hence, the boxAC element mentioned in this article refers to the above-mentioned annotation. The GϩC content of the boxAC element (40.37%) is similar to that of yefM-yoeB Spn (39.56%) and the chromosome of S. pneumoniae R6 (39.72%), which indicates that boxAC is a native element or that it has been a resident of the genome for a long time. There are 127 copies of BOX elements in the TIGR4 genome altogether (49) and 115 copies in the R6 genome (13). However, their functions are still unclear, since only a few studies have been performed so far, although their presence has been associated with competence and with virulence genes (27) and with pneumococcal phase variation (45).
yefM Spn and yoeB Spn are cotranscribed. The yefM Spn and yoeB Spn reading frames are separated by 3 nt and thus appear to constitute an operon. In the genome of S. pneumoniae R6 and the other strains that contained intact yefM-yoeB Spn in the database (except TIGR4 and CGSP14), the yefM Spn reading frame terminates with two TGA stop codons followed immediately by the ATG start codon of yoeB Spn , which, in turn, would have its rbs (GAGGAG) embedded within the yefM Spn coding sequence ( Fig. 2A). On the other hand, the TIGR4 and CGSP14 strains have 3 nt (GAA) in between the TAA stop codon of yefM Spn and the ATG start codon of yoeB Spn. Experimental validation of the yefM-yoeB Spn genetic organization was obtained by RT-PCR with total RNA isolated from S. pneumoniae R6. Reverse transcriptase was used to synthesize cDNAs that were complementary to the yefM-yoeB Spn mRNA from three independent preparations of pneumococcal RNAs. The resulting cDNAs were PCR amplified using two oligonucleotide primers (yefMNHis and GyoeBC) that would anneal to the 5Ј end of yefM Spn and the 3Ј region of yoeB Spn , respectively ( Fig. 2A). The individual genes were also amplified using the oligonucleotide primer pair yefMNHis and yefMC, as well as the GyoeBN and GyoeBC pair, respectively, as positive controls. No PCR products were evident in the negative controls, in which the same RNA samples were added but the reverse transcriptase extension was not detected, thus ruling out possible DNA contamination (Fig. 2B, lane 6). The resulting DNA products were of the expected sizes for yefM Spn (336 bp; lanes 1 and 2), yoeB Spn (300 bp; lanes 3 and 4), and yefM-yoeB Spn (563 bp; lanes 5 and 7) (Fig. 2B). The latter cDNA corresponded to a single mRNA encompassing both genes, thus demonstrating that in S. pneumoniae, yefM Spn and yoeB Spn are cotranscribed. Similar results were obtained when RT-PCR was conducted on total RNA isolated from E. coli harboring a pGEM-T Easy recombinant plasmid which contained the entire yefM-yoeB Spn reading frames along with 237 bp of sequences upstream of yefM Spn encompassing the putative promoter sequences (not shown). These results are in agreement with those reported for other TA loci, such as relBE2 in S. pneumoniae (35) and the kis-kid genes of the parD maintenance system of plasmid R1 (43), where the antitoxin and toxin genes are cotranscribed, which is also a general characteristic of TAS.
The yefM-yoeB Spn operon is transcribed from two promoters, P yefM1 and P yefM2 . To determine the transcription initiation start site(s) of the yefM-yoeB Spn operon, primer extension analyses were performed. Total RNA was prepared from exponentially growing S. pneumoniae R6 cells, annealed with two labeled primers, termed yefM-near and yefM-far, and extended with reverse transcriptase at either 50°C or at 55°C. Two different extension products, of 92 nt or 115 nt, were observed (Fig. 1B), which indicated that the yefM-yoeB Spn operon was transcribed from two transcriptional start sites -the "A" purine residues located 84 nt and 25 nt upstream of the yefM Spn ATG start codon (see Fig. 1A), respectively. An inspection of the DNA sequence around the transcriptional start sites revealed that the primer extension products would be complementary to mRNAs transcribed from the P yefM1 and P yefM2 promoters, respectively. Our primer extension results showed that the transcriptional start sites for both P yefM1 and P yefM2 are located 5 nt downstream relative to their respective Ϫ10 elements (Fig.  1A). We can conclude that insertion of the boxAC element had indeed led to the creation of a functional P yefM1 promoter.
Determination of promoter activities in vivo. The in vivo functionality and strength of the two promoters were determined by employment of transcriptional fusions with a promoterless lacZ gene as a reporter in plasmid pQF52 in E. coli. The more important results obtained are depicted in Fig. 3 (also, Table 1 shows the various constructs and controls). The functionality of P yefM2 was validated when E. coli cells harboring P yefM2 (pQF_P2) showed a mean ␤-galactosidase activity of 122 Ϯ 36 Miller units (MU) (Fig. 3). The ␤-galactosidase activities of the other constructs were stated as ratios by normalization of their ␤-galactosidase activity levels with that of pQF_P2. An activity ratio of only 0.07 was detected in cells that harbored P yefM1 alone (pQF_P1), suggesting that P yefM1 was a much weaker promoter than P yefM2 (Fig. 3). However, when both the P yefM1 and P yefM2 promoters were included in the construct, resulting in pQF_P1P2, the ratio was reduced to 0.57 compared to that of the P yefM2 promoter alone (pQF_P2) (Fig.  3). From these results, P yefM1 was found to be ϳ15-fold weaker than P yefM2 , and the presence of the additional P yefM1 promoter decreased the overall promoter activity compared to the activity of just P yefM2 alone.
When the YefM Spn antitoxin was expressed in trans (pLN_yM), P yefM2 activity was slightly reduced by 7% compared to results with the uninduced cells (Fig. 3). In addition, when the YefM-YoeB Spn protein complex was coexpressed in trans (pLN_yMyB), the ␤-galactosidase activity was further reduced by 87% (Fig. 3). Similar results were observed when the yefM Spn and yefM-yoeB Spn reading frames were included in cis (Table 1). These results indicated that as in most TAS, the YefM Spn antitoxin by itself acts as a weak repressor and the YoeB Spn toxin acts as a corepressor to further increase transcriptional repression from the P yefM2 promoter.
When P yefM1 and the boxAC element were included in the pQF52 construct along with P yefM2 , expression of YefM Spn in trans (pLN_yM) led to a 12% reduction in ␤-galactosidase activity compared to results with the uninduced cells (Fig. 3). However, no prominent further repression (16% reduction) was observed when YefM-YoeB Spn was expressed in trans (pLN_yMyB) (Fig. 3). Intriguingly, the presence of the yefM Spn reading frame in cis (pQF_P1P2yM) led to an ϳ170% increase in promoter activities compared to the activity in cells that harbored just the entire promoter region (Fig. 3). When both the yefM Spn and yoeB Spn reading frames were included along with the entire promoter region in cis (pQF_P1P2yMyB), lower activity was detected but the activity was still 98% higher than that in cells that harbored just the entire promoter region (pQF_P1P2) (Fig. 3). To investigate whether the YefM Spn antitoxin could serve as an activator to activate the upstream promoters, an amber stop codon was introduced at the 7th codon of yefM Spn to abolish the translation of full-length YefM Spn . High ␤-galactosidase activity (i.e., a ratio of 1.89) was detected in the mutated clone, indicating that the increased activity was still present (Fig. 3). Even when yoeB Spn was included along with this mutated construct, the activation was still evident, with an observed ratio of 2.03 (Fig. 3). No detectable ␤-galactosidase activity was observed in the recombinant clone that harbored just the yefM Spn reading frame, which ruled out the possibility of yefM Spn harboring an internal promoter (Table 1) .
Taken together (including results obtained from the other recombinants in Table 1), these results suggested that the increased ␤-galactosidase activity was observed when the yefM Spn coding sequence was present along with the entire upstream region spanning boxAC, P yefM1 , and P yefM2 in cis. Furthermore, activation was not due to YefM Spn acting as an activator or to the presence of an internal promoter within the yefM Spn reading frame.
Unexpectedly, a functional promoter was detected within the yoeB Spn coding sequence, and this promoter was likely present at the C terminus (the last 159 nt of yoeB Spn in pQF_CyB gave an activity ratio of 1.28) ( Table 1). To verify whether this promoter was indeed functional in its native host, a 159-nt fragment encoding the C terminus of YoeB Spn was cloned into the promoter-probe pAST plasmid, which carries the green fluorescent protein-encoding gene as a reporter (42); transformants were rescued in S. pneumoniae R6. Two clones having the insert in opposite directions were tested for fluorescence, and no promoter activity was evident (not shown). The genome of S. pneumoniae has a high AϩT content (61%) (5); thus, sequences resembling a Ϫ10 region are likely to exist by chance, and they could be recognized by the E. coli 70 RNA polymerase (RNAP) (47).
Determination of DNA binding sites of the YefM Spn protein and the YefM-YoeB Spn protein complex. EMSA and footprint- VOL. 193, 2011 PNEUMOCOCCAL yefM-yoeB TOXIN-ANTITOXIN REGULATION 4619 ing experiments were carried out to determine the DNA binding sites for the YefM Spn antitoxin and the YefM-YoeB Spn TA complex. For EMSA, a 322-bp [␥-32 P]ATP-labeled PCR-amplified DNA fragment encompassing the entire upstream region of yefM Spn , along with the boxAC element and the PS, was incubated with increasing amounts of purified YefM Spn or purified YefM-YoeB Spn proteins. The unbound DNA fragments and the nucleoprotein complexes were observed as wellseparated bands on native 5% polyacrylamide gels, indicating that the antitoxin alone or the protein complex was bound to this DNA fragment. Furthermore, preliminary results indicated that both proteins were bound to a DNA fragment con-taining only PS, and no other binding site was observed (not shown). Consequently, a 284-bp DNA fragment encompassing PS (but not the boxAC element) was used for further assays. This DNA fragment was end labeled with [␥-32 P]ATP on the coding strand. Aliquots (3,000 cpm) were incubated with increasing amounts of either purified YefM Spn protein or purified YefM-YoeB Spn protein complex. EMSA revealed that in the case of YefM Spn , the DNA fragment was retarded as a single complex starting at 3 g of YefM Spn , with most of the labeled DNA fragment shifted when 5 g of YefM Spn was included (Fig. 4A). In the case of the YefM-YoeB Spn protein complex, more than 70% retardation of the labeled DNA frag- FIG. 3. Schematic diagram of the yefM-yoeB Spn -lacZ transcriptional fusion recombinant plasmids and their corresponding ␤-galactosidase activities. The organization of the yefM-yoeB Spn genes, P yefM1 and P yefM2 promoters, the palindrome sequence (PS), and the boxAC element are shown at the top. DNA fragments containing combinations of the yefM-yoeB Spn locus and its upstream region were cloned upstream of the promoterless lacZ gene (depicted as a gray box) in pQF52, resulting in the various pQF52-derived recombinants as shown. yefM Spn or yefM-yoeB Spn was cloned downstream of the P BAD promoter in pLNBAD, resulting in the pLN_yM and pLN_yMyB constructs. In cells that harbored the pLNBAD-derived constructs, ␤-galactosidase activities were measured when cells were induced with 1 mM L-arabinose for the expression of yefM Spn or yefM-yoeB Spn from the P BAD promoter. ␤-Galactosidase activity levels are the average results of eight independent experiments and are presented as relative levels (ratios) by normalization of the ␤-galactosidase activity levels with that of pQF_P2, which is 122 Ϯ 36 MU (a detailed presentation of all the constructs and their resulting ␤-galactosidase activity levels is depicted in Table 1).

4620
CHAN ET AL. J. BACTERIOL. ment was observed when even as little as 0.05 g of the protein complex was added, and most of the DNA fragment was shifted with 0.25 g of the protein complex (Fig. 4B). These results are consistent with a change in the binding affinity of the YefM-YoeB Spn complex for PS compared with that of the YefM Spn protein alone. We next performed a set of DNase I footprinting assays using the same DNA fragment but labeled at either the coding or noncoding strand. The labeled DNA fragments (30,000 cpm) were incubated in a reaction mixture with increasing amounts of either purified YefM Spn protein or purified YefM-YoeB Spn protein complex, followed by DNase I digestion. The footprinting patterns were similar for the antitoxin alone and for the TA complex although weaker with the former protein; thus, only the results obtained with the TA complex are presented (Fig. 5). The protected regions were separated by two unprotected bases in the coding strand only, being located from nucleotides Ϫ14 to Ϫ5 of the left arm and from nucleotide Ϫ2 of the left arm to nucleotide ϩ12 of the right arm of PS ( Fig. 5A and B). The protection sites also overlapped with the entire Ϫ35 element of P yefM2 . Nucleotides Ϫ4 and Ϫ3 of the left arm of PS were not protected ( Fig. 5A and B). For the noncoding strand, only one region was protected, and all nucleotides protected by YefM Spn were also protected by YefM-YoeB Spn . The protected region spanned from nucleotide Ϫ10Ј of the left arm to ϩ13Ј of the right arm of PS ( Fig. 5A and B). Hypersensitive bands were observed from nucleotides ϩ14 to ϩ16 of the right arm of PS in the coding strand but not in the noncoding strand (Fig. 5). This hypersensitivity could be due to deformation of the DNA helix provoked by the binding of the proteins that occurred at this position (32).
Refinement of the contacts between the YefM Spn and YefM-YoeB Spn proteins with their DNA target was achieved by hydroxyl radical footprinting because of the high resolution potential of this technique (52). All of the DNase I footprints were also detectable by hydroxyl radical footprinting, although the former were divided in two, thus defining a region of palindromic symmetry in which two and three protected bases on the coding strand were mirrored by three and two protected bases on the noncoding strand ( Fig. 5 and 6). A very weak footprint was observed when YefM Spn alone was used (not shown), corroborating the low affinity of the antitoxin for its target. Again, only the results obtained with the TA complex are presented, and a summary of the footprints obtained with the YefM Spn antitoxin alone is depicted in Fig. S4 in the supplemental material. The protections by YefM-YoeB Spn in the hydroxyl radical assays (Fig. 5C) were included within those observed for the DNase I footprinting. On the coding strand (Fig. 5C), the two protected regions were separated by 9 nt; these were located at nucleotides Ϫ11 and Ϫ10 of the left arm, as well as from nucleotide ϩ1 to ϩ3 of the right arm of PS, which overlapped with the Ϫ35 element of P yefM2 (Fig. 5C). For the noncoding strand, the protected sites spanned from nucleotides Ϫ3Ј to Ϫ1Ј of the left arm and at nucleotides ϩ9Ј and ϩ10Ј of the right arm of PS (Fig. 5C). An overall picture of the footprinting analyses on this DNA region (Fig. 6A) could be better visualized when the protected regions were depicted on a schematic drawing of a 10.5-bp DNA helix (Fig.  6B). No indication of protein-induced DNA bends was observed; however, three hyperexposed bases were detected on the coding strand at the border of the footprint located 3Ј to the Ϫ35 promoter region (Fig. 6A), indicative of a local helix deformation at these positions.

DISCUSSION
We show here that the pneumococcal yefM-yoeB Spn TAS is constituted as an operon in which two promoters direct the synthesis of the mRNA transcript. The genetic regulation of the pneumococcal yefM-yoeB Spn locus differed from most conventional TA loci due to the presence of the two promoters, P yefM1 and P yefM2 , and the boxAC element. Promoter P yefM2 , which is nearer to the yefM-yoeB Spn locus, is likely the "original" or native promoter of the TA locus, whereas P yefM1 , which is the weaker of the two promoters, came about from the insertion of boxAC upstream of P yefM2 . Thus, the presence of this BOX element provides a unique scenario for regulation of the expression of this particular pneumococcal TAS. If the regulation of the yefM-yoeB Spn locus is considered in terms of the P yefM2 promoter alone, then the locus is negatively autoregulated like other typical TAS. The YefM Spn antitoxin would repress, albeit weakly, transcription from P yefM2 , whereas the YoeB Spn toxin would act as a corepressor to fully regulate transcription from P yefM2 . Both DNase I and hydroxyl radical footprinting assays showed that YoeB Spn increased the affinity of YefM Spn for the operator site. The Ϫ35 region of P yefM2 was not directly protected by the YefM Spn protein (see Fig. S4 in the supplemental material), but protection of this region was very clear in the presence of the YefM-YoeB Spn complex (Fig.  6). Representation of the distribution of bases which were protected by bound YefM-YoeB Spn proteins (Fig. 6) on a DNA molecule with a helical periodicity of 10.5 bp per helix turn (41) clearly showed that the proteins contact the DNA backbone on one face of the helix and that the protections did not extend further than the central region of the PS. The DNase I footprints delineated two central regions of 10 and 19 protected bases (on the coding strand) separated by two unprotected ones; on the noncoding strand, a single footprint could be observed. However, the high-resolution analyses ob-FIG. 5. DNase I and hydroxyl radical footprinting assays for the purified YefM-YoeB Spn proteins on the DNA fragment containing the PS palindrome sequence. (A and B) DNase I footprinting assays. Both coding strands (A) and noncoding strands (B) of [␥-32 P]ATP-labeled DNA fragments containing the PS (30,000 cpm) were incubated with increasing amounts of YefM-YoeB Spn protein complex (0, 1, 2, and 5 g) prior to DNase I digestion. The reaction mixture was then separated on a 8% polyacrylamide gel containing 7 M urea along with the DNA sequencing ladder prepared using the dideoxy sequencing reaction. Nucleotide sequences protected by the YefM-YoeB Spn complex from DNase I digestion (green) and hypersensitivity observed at these nucleotides (brown) are indicated. (C) Hydroxyl radical footprinting assays. Aliquots (30,000 cpm) of both coding (lanes 1 and 2) and noncoding (lanes 3 and 4) strands of a [␥-32 P]ATP-labeled DNA fragment containing the PS were incubated with (lanes 2 and 3) or without (lanes 1 and 4) 5 g of purified YefM-YoeB Spn protein complex prior to hydroxyl radical treatment. The reaction mixture was then separated on a 8% polyacrylamide gel containing 7 M urea along with a DNA sequencing ladder (AϩG) prepared by the Maxam-Gilbert sequencing method. Nucleotide sequences of the coding and noncoding DNA strands of the yefM Spn upstream region containing the PS that were protected from hydroxyl radical attack by the YefM-YoeB Spn proteins are shown in red.
tained with the hydroxyl radical footprinting showed that the DNase I footprints could be further separated into two protected regions on both coding and noncoding strands (Fig. 6).
On the basis of these results, it would appear that the mechanism of transcriptional repression from P yefM2 is likely due to the binding of the antitoxin or the TA complex to the PS site, which overlaps the Ϫ35 region. Occupancy of the TA complex on the PS would lead to interference of the activity of RNAP on P yefM2 . Similar findings have been reported for other proteic TAS (10), such as the higBA locus, where the HigA antitoxin protein acted as a repressor by binding and covering the Ϫ35 and Ϫ10 regions of the P hig promoter (50), thereby preventing RNAP from accessing the promoter. The differential regulation of the pneumococcal yefM-yoeB Spn locus stems from the presence of the upstream boxAC element. BOX elements have been proposed to be acquired through horizontal gene transfer. Single nucleotide duplications, which are common occurrences during transposition (23), have been noted flanking BOX elements (17). Indeed, duplication of an "A" nucleotide was observed for the boxAC element upstream of the yefM-yoeB Spn locus. However, no differences in GϩC content of boxAC and its surrounding DNA were detected, which would exclude recent acquisition of the element from phylogenetically distant bacteria. The insertion of boxAC likely led to the creation of the additional weak promoter, P yefM1 . Nonetheless, its presence along with the boxAC element is able to stimulate the expression of the downstream genes, the yefM-yoeB Spn locus, which is in line with other previous reports (17). Tandem promoters are usually stronger than the individual promoters themselves (3,33). However, with the close proximity (30 nt) of P yefM1 and P yefM2 and their likely position on the same face of the DNA helix, each may influence the other's binding affinity for RNAP and thus reduce the overall promoter activity. Alternatively, the time required to form an open complex and to escape from the promoter may enable one promoter to depress the activity of the other.
Intriguingly, even though the overall promoter activity of both P yefM1 and P yefM2 was lower than that of the P yefM2 promoter alone, the overall promoter activity increased dramatically when the yefM Spn reading frame was provided in cis. However, transcriptional activation by the YefM Spn protein was ruled out, since this activation was not observed when yefM Spn was provided in trans and, more tellingly, stop codons introduced into the yefM Spn reading frame did not abrogate transcriptional activation. No internal promoters were detected within the yefM Spn reading frame. This activation was likely due to cis-acting elements, and perhaps the boxAC element and/or its downstream sequences contain activator binding sites for a host factor(s) that has yet to be elucidated.
BOX elements have been associated with the following: (i) genes involved in the induction of pneumococcal genetic competence (17,27), (ii) virulence-related genes (27), and (iii) expression of phase variation (45). Another scenario was found for the S. pneumoniae maltose regulon, where a full BOX element was detected upstream of the regulatory malAR genes, indicating that the role of these repeated elements may be more complicated than envisaged (40). However, since a large part of the pneumococcal genome is devoted to sugar uptake and metabolism, it seems possible that regulation of these processes (metabolism and response to stress) is important for the bacterial fitness in its natural niche (the human nasopharynx) and may play a role during invasion of other tissues, such as the lungs (14).
In the case of the pneumococcal yefM-yoeB Spn TAS, the presence of the boxAC element also added another level of complexity to the regulation of this operon. The two promoters driving the yefM-yoeB Spn locus may thus enable a more dynamic regulation of the TA operon. Since P yefM1 does not appear to be regulated by either YefM Spn or the YefM- YoeB Spn complex, P yefM1 would provide a higher basal level of transcription for the yefM-yoeB Spn operon, thus enabling cells to have a faster response to any sudden changes in their environment. BOX elements may not serve a specialized role in the host; however, their presence is more likely to affect the regulation or expression of the genes by providing a versatile response to environmental changes. Be that as it may, from an evolutionary point of view (2), the insertion of a translocative piece (the boxAC element) into an operative piece (the yefM-yoeB Spn operon) might have resulted in a novel genetic context, leading to a better fitness of the bacterium within its natural niche. Thus, the presence of the boxAC element upstream of the yefM-yoeB Spn operon opens new avenues to a deeper exploration of the mechanisms governing the bacterial response to stress and their relationships with bacterial evolvability.