Characterization of mRNA Interferases from Mycobacterium tuberculosis*

mRNA interferases are sequence-specific endoribonucleases encoded by the toxin-antitoxin systems in the bacterial genomes. MazF from Escherichia coli has been shown to be an mRNA interferase that specifically cleaves at ACA sequences in single-stranded RNAs. It has been shown that MazF induction in E. coli effectively inhibits protein synthesis leading to cell growth arrest in the quasidormant state. Here we have demonstrated that Mycobacterium tuberculosis contains at least seven genes encoding MazF homologues (MazF-mt1 to -mt7), four of which (MazF-mt1, -mt3, -mt4, and -mt6) caused cell growth arrest when induced in E. coli. MazF-mt1 and MazF-mt6 were purified and characterized for their mRNA interferase specificities. We showed that MazF-mt1 preferentially cleaves the era mRNA between U and A in UAC triplet sequences, whereas MazF-mt6 preferentially cleaves U-rich regions in the era mRNA both in vivo and in vitro. These results indicate that M. tuberculosis contains sequence-specific mRNA interferases, which may play a role in the persistent dormancy of this devastating pathogen in human tissues.

Most bacteria contain "suicidal" or "toxic" genes whose expression leads to growth arrest and eventual death upon exposure to stress. These toxin genes are usually coexpressed with their cognate antitoxin genes present in the same operon (1,2). The Escherichia coli chromosome contains six such operons called toxin-antitoxin (TA) 2 systems. Of these, the MazE (antitoxin)/MazF (toxin) system is one of the most extensively characterized TA systems, and MazF has been shown to be a sequence-specific endoribonuclease that cleaves at ACA sequences in mRNAs (3,4). Thus, MazF is an mRNA interferase, and its induction causes the effective inhibition of protein synthesis leading to cell growth arrest. However, MazF-induced cells fully retain cellular metabolic activities, including ATP production and amino acid and nucleotide synthesis, as well as RNA and protein synthesis. Therefore, MazF-induced cells are capable of synthesizing a protein, if the gene encoding this protein is devoid of ACA sequences (4). This metabolically active dormant state caused by mRNA interferase induction is called "quasidormancy" (4) and has important implications in the physiology of various pathogenic bacteria, including persistent multidrug resistance.
Notably, Mycobacterium tuberculosis, one of the most devastating human pathogens, contains nearly forty TA systems on its genome, among which nine have been shown to be homologous to E. coli MazF (5). It has been proposed that the TA modules play important roles in bacterial survival and formation of stable persisters in adverse conditions. These phenomena are essential for the persistence of M. tuberculosis. Persisters are the few rare pre-existing bacterial cells in a culture at any growth phase that are not growing and are intrinsically resistant to antibiotics by virtue of their subdued metabolism (6,7).
In the present paper, we have characterized two of the MazF homologues from this pathogen, which showed high toxicity in E. coli when induced. We have demonstrated that these are indeed mRNA interferases that cleave specific RNA sequences. In view of the quasidormancy caused by MazF induction in E. coli, our results suggest that multiple mRNA interferases in M. tuberculosis may play a role in the multidimensional dormancy of this pathogen in human tissues.
Expression and Purification of MazF-mt1 Using Mycobacterium smegmatis MC 2 155 as a Host-The wild-type MazF-mt1 gene was PCR-amplified and cloned into an E. coli-M. smegmatis shuttle vector downstream of an acetamidase gene promoter (cloned from M. smegmatis genomic DNA) and upstream of a hexahistidine tag. This vector, carrying a hygromycin resistance gene, was constructed in our laboratory. M. smegmatis-competent cells carrying the construct were used to inoculate 2 ml of 7H9 medium, and this starter culture was incubated at 37°C with shaking for 2 days and then used to inoculate larger volumes of 7H9 medium supplemented with Middlebrook oleic acid-albumindextrose-catalase enrichment and 2% acetamide (Sigma). The culture was grown for three days and cells were then harvested and resuspended in equilibration buffer (25 mM imidazole, 400 mM NaCl, and 50 mM NaPi, pH 8.0). The cells were then lysed with a French press. Cell debris was removed by one round of centrifugation at 25,000 rpm and at 4°C. The lysate was added to Ni 2ϩ -nitrilotriacetic acid-agarose (Qiagen) for 1 h, and the resin was then transferred to a column and washed with 50 ml of 50 mM NaPi, pH 8.0, 400 mM NaCl, and 20 mM imidazole. The His-tagged proteins were eluted with 50 mM NaPi, pH 8.0, 400 mM NaCl, and 250 mM imidazole. Protein fractions were checked at A 280 and purity was assessed by SDS-PAGE analysis. The highly purified fractions were collected and dialyzed against 10 mM HEPES buffer containing 5% glycerol, 0.1 mM dithiothreitol, and 500 mM NaCl. The protein aliquots were stored at Ϫ80°C.
Primer Extension Analysis in Vivo and in Vitro-For primer extension analysis of mRNA cleavage sites in vivo, pIN-Era plasmid (8) was transformed into E. coli BW25113 cells containing pBAD-MazF-mt1 or MazF-mt6 plasmids. The era mRNA transcription was induced by the addition of 1 mM isopropyl 1-thio-␤-D-galactopyranoside. After a 30-min induction, the toxin protein was induced by the addition of arabinose (a final concentration of 0.2%). Total RNA was isolated at time intervals as indicated in Figs. 3 and 4. Primer extension was carried out using different primers as described previously (3). For in vitro primer extension analysis, the full-length era mRNAs were synthesized in vitro by T7 RNA polymerase from the DNA fragment containing a T7 promoter sequence and the era open reading frame (ORF) using the RiboMAX TM T7 large scale RNA production system (Promega). The  (9). Identical and homologous residues are shown in black and shaded backgrounds. B, alignments of MazF-mt2-mt7 to MazF-mt1. The consensus sequence is shown at the bottom. In MazF-mt2, the 19-residue sequence in blue is immediately downstream of the codon for Thr-88 with a Ϫ1 frameshift. MazF-mt2-mt7 correspond to Rv0456A, Rv1991c, Rv0659c, Rv1942c, Rv1102c, and Rv1495, respectively. C, upstream ORFs for MazF-mt1-mt7. These overlap with downstream MazF ORFs by 14,14,7,14,4,29, and 4 bases for MazF-mt1-mt7, respectively. Total residue numbers and pI values are shown on the right. JULY 7, 2006 • VOLUME 281 • NUMBER 27 era DNA fragments were amplified using the forward primer (5Ј-CCCGCGAAATTAATACGACTCACTATAG-3Ј T7 promoter) and the reverse primers starting from the 3Ј-end of the era ORF. RNA substrates were partially digested with purified toxin protein at 37°C for 15 min. The digestion reaction mixture (10 l) consisted of 1 g of RNA substrate, 0.2 g of E. coli His 6 MazF-mt1 or 1 g of M. smegmatis His 6 MazF-mt1 or E. coli His 6 MazF-mt6 and 0.5 l of RNase inhibitor (Roche Applied Science) in 10 mM Tris-HCl (pH 7.8). Primer extension was carried out at 42°C for 1 h in 20 l of the reaction mixture as described previously (8). The reactions were stopped by adding 12 l of sequence loading buffer (95% formamide, 20 mM EDTA, 0.05% bromphenol blue, and 0.05% xylene cyanol FF (C 25 H 27 N 2 NaO 7 S 2 )). The samples were incubated at 90°C for 5 min prior to electrophoresis on a 6% polyacrylamide and 36% urea gel. The primers E1 (5Ј-CAGT-TCAGCGCCGAGGAAACGCAT-3Ј), E2 (5Ј-GATCCCCACAAT-GCGGTGACGAGT-3Ј), and E4 (5Ј-GCGTTCGTCGTCGGC-CCAACCGGA-3Ј) were used for primer extension analysis of the era mRNA. The primers were 5Ј-labeled with [␥-32 P]ATP using T4 polynucleotide kinase.

Characterization of mRNA Interferases from M. tuberculosis
Cleavage of Synthetic RNA by MazF-mt1-Oligoribonucleotides (11 bases in length; see Fig. 3) were commercially synthesized and 5Ј-labeled with [␥-32 P]ATP using T4 polynucleotide kinase. Endoribonuclease activity was assayed in 10 l of the reaction mixture containing 0.5 l (20 units) of ribonuclease inhibitor (Roche Applied Science), 0.2 g of E. coli MazF-mt1 or 1 g of M. smegmatis MazF-mt1 and 32 P-labeled oligonucleotides in 10 mM Tris-HCl (pH 7.8). Reactions were carried out at 37°C for 30 min and stopped by the addition of the loading buffer as described above. The reaction mixtures were then subjected to 20% sequence gel electrophoresis followed by autoradiography.  that has 32.5% identity and 44% homology to E. coli MazF (Fig. 1A). Using this protein MazF-mt1 for further blast search, six more MazF homologues with Ͼ20% identity to MazF-mt1 (mt2 to mt7) were found (Fig. 1B). Notably, MazF-mt1, -mt2, and -mt3 have the putative 11-residue S1-S2 loop similar to E. coli MazF (9), whereas the others have much shorter S1-S2 loops (Fig. 1, A and B). The putative helix H1 is longer by five residues in MazF-mt1 and -mt2 and by three residues in -mt3, compared with that of E. coli MazF, whereas it is shorter by one residue in MazF-mt4, -mt5, -mt6, and -mt7 as compared with that of E. coli MazF (Fig. 1, A and B). MazF-mt2 lacks the C-terminal region encompassing ␤-strand S7 and helix H3. This seems to be due to a 28-bp deletion after the codon for Thr-88, because the 19-residue sequence highly homologous to helix H3 of MazF-mt1 is found immediately downstream of Thr-88 in a different reading frame (68% identity) (Fig.  1B). Interestingly, all MazF homologues appear to be co-translated with a short upstream ORF, most of which overlaps with the MazF ORFs (Fig.  1C). None of the upstream ORFs, however, show homology to E. coli MazE, antitoxin of MazF. As observed with MazE and MazF, most of the pairs consist of acidic and basic proteins, except for the pair of MazF-mt2 and its upstream ORF, both of which are basic proteins (Fig. 1C). It remains to be elucidated whether these upstream ORFs function as antitoxins for their cognate downstream ORFs.

Four of the Seven MazF Homologues in M. tuberculosis Show Toxicity to E. coli-All
ORFs were tested for their toxicity in E. coli using the pBAD expression plasmid. In the presence of 0.2% arabinose, cells expressing MazF-mt1, -mt3, -mt4, and -mt6 could not grow, whereas cells expressing MazF-mt2, -mt5, and -mt7 proteins were able to grow on an agar plate ( Fig. 2A). Inhibitory effects of MazF-mt1, -mt3, -mt4, and -mt6 were also observed in M9-CAA liquid medium in the presence of 0.2% arabinose (Fig. 2B). In the case of cells expressing MazF-mt4 protein, severe growth inhibition was seen only after one generation of cell growth.
The MazF-mt1 Has Endoribonuclease Activity-We examined whether these MazF homologues have sequence-specific mRNA interferase activity, both in vivo and in vitro. When MazF-mt1 was induced by the addition of 0.2% arabinose, a specific cleavage of the era mRNA . RNA cleavage specificity of MazF-mt1. A, in vivo cleavage of the era mRNA after induction of MazF-mt1 by adding 0.2% arabinose. Total RNA was extracted at each time point after MazF-mt1 induction, and primer extension was carried out using primer E1 as described previously (17). Using the same primer, the DNA ladder was prepared. The cleavage site is indicated by an arrow. B, in vitro cleavage of the era mRNA with His 6 MazF-mt1 purified from E. coli (lane 2) and M. smegmatis (lane 3). Lane 1 represents a control reaction in which no enzymes were added. Cleavage sites are indicated by arrowheads on the RNA sequence determined using the DNA ladder shown on the right. C, cleavage of a synthetic 15-base RNA (AGAUAUACAUAUGAA) labeled at the 5Ј-end with 32 P (lanes 1 and 2). A 19-base DNA with the identical sequence as that of the RNA with extra GG and TC at the 5Ј-and 3Ј-ends, respectively, was also tested (lanes 3 and 4). In lane 5, both RNA and DNA were added. The reaction was carried out 30 min at 37°C as described previously (8). The position of the 5Ј-end product was identical to that obtained with E. coli MazF (6-base RNA). D, cleavage specificity of MazF-mt1 using 11-base RNA substrates. In lanes 1 and 2, AUAUACAUAUG labeled at the 5Ј-end with 32 P was used as the substrate. Judging from the ladder shown on the right side, a 4-base RNA was released by the enzyme, indicating that the enzyme cleaves between U 4 and A 5 . Substrates in which the U 4 , A 5 , C 6 , and A 7 residues were replaced with a G residue were used for lanes 3 and 4, lanes 5 and 6, lanes 7 and 8, and lanes 9 and 10, respectively. between U and A was detected in vivo by primer extension in a timedependent manner (Fig. 3A). The identical cleavage site was detected with use of the era mRNA synthesized with T7 RNA polymerase and purified MazF-mt1 tagged with His 6 at the N-terminal end (Fig. 3B, lane  2). In addition to the same site reported for the in vivo cleavage site (CU2ACC), a weak cleavage site (UU2ACA) was also detected. Using five primers covering almost the entire era mRNA, CU2ACC and UU2ACA were the only major cleavage sites detected, even when era mRNA contained six other UAC sequences (UUACU, AUACU, CUACG, AUACA, GUACU, and UUACG). Importantly, identical sites were cleaved by His 6 MazF-mt1, which was expressed and purified from M. smegmatis (Fig. 3B, lane 3), indicating that the observed cleavage is due to MazF-mt1. We further confirmed that the cleavage activity was not due to contaminating enzymes, because a purified His 6 MazF-mt1 (E19A) mutant protein showed a substantially reduced cleavage activity (not shown). This mutation in E. coli MazF has recently been shown to dramatically reduce the MazF mRNA interferase activity (10).
For further biochemical characterization, purified His 6 MazF-mt1 was used to cleave a synthetic 15-base RNA (AGAUAU2ACAUA-UGAA), which was also shown to be cleaved between U and A of the UAC sequence (Fig. 3C, lane 2 and arrows). A 19-base DNA with a sequence identical to that of the RNA substrate in the center could not be cleaved (lane 4). When the DNA and RNA were mixed and treated with the enzyme, only RNA was cleaved (lane 5), indicating that MazF-mt1 is an endoribonuclease. To further test the cleavage specificity, five 11-base RNA substrates were synthesized, one with the UACA sequence in the center (AUAUACAUAUG) and the others with a G residue in each one of the UACA sequences. The first U and the second A residues could not be replaced with G (Fig. 3D, lanes 3-6). The third C residue also appears to be important, as the cleavage between U and A (lanes 7 and 8) was significantly reduced when this C was replaced with G. On the other hand, the fourth A residue was replaceable with G (lanes 9 and 10), confirming that MazF-mt1 is an endoribonuclease that specifically cleaves UAC sequence.
The MazF-mt6 Is an mRNA Interferase-Because MazF-mt6 showed toxicity in E. coli (Fig. 2), we examined whether it, too, is a sequencespecific RNA interferase. We speculated that other MazF homologues are not toxic in E. coli, possibly because they may have highly specific cleavage activities and thus cleave only a very limited number of sequences in E. coli mRNAs or they are not efficiently expressed in E. coli. We carried out in vivo and in vitro primer extension experiments using the era mRNA and the MazF-mt6 protein. As shown in Fig. 4, a number of in vivo cleavage sites were detected in a time-dependent manner after its induction, all of which were also detected in an in vitro experiment with purified His 6 MazF-mt6 protein. Notably, a new cleavage site was also observed in an in vitro experiment in addition to those found in an in vivo (Fig. 4B, lane 10) experiment. These cleavages preferentially occurred in U-rich regions and after a U residue. Therefore, (U/C)U2(A/U)C(U/C) may be assigned as a consensus cleavage sequence for MazF-mt6 mRNA interferase. The cleavage also occurred after G or A residues in some cases, although all cleavage sites contained UU, UC, or CU residues.

DISCUSSION
In this report, we have demonstrated that M. tuberculosis contains a number of MazF homologues, some of which were toxic to E. coli cells when induced. It has been shown that E. coli MazF causes complete cell growth arrest; however, the cells retain full metabolic activ- ities (4). Therefore, it is tempting to speculate that this novel physiological quasidormant state caused by MazF may be, in some ways, related to the long term dormancy of M. tuberculosis in human tissues. By having a large number of mRNA interferases that may be induced under different environmental conditions, the persistent dormancy of this pathogen in human tissues may be well regulated and may lead to antibiotic resistance.
Microarray studies of the stringent response and the starvation response carried out with M. tuberculosis may provide insights into regulation of expression of the toxin genes in this pathogen. In the M. tuberculosis starvation model by Betts et al. (11), MazF-mt4 (Rv0659c) was found to be down-regulated after 4 h of nutrient starvation. In a microarray study of a strain of H37Rv in which stringent response gene Rel Mtb is deleted, MazF-mt6 (Rv1102c) expression was shown to increase 3-fold after 6 h of starvation in the knock-out strain, whereas expression was unchanged after 6 h of starvation in wild-type Mtb (12). Both of these observations thereby directly implicate a role for Rel Mtb and (p)pppGpp ("the stringent response") in the regulation of these toxin genes, namely down-regulation. However, the net effect on cell metabolism or viability of these changes in toxin expression depends on the respective changes in expression of their antitoxin counterparts. These results may be interpreted in the context of regulation of the mazEF operon in E. coli. When the stringent response is triggered in E. coli, the RelA-dependent increase of (p)ppGpp causes a decrease in activity of the P2 promoter of the mazEF operon (13). Subsequently, MazE is degraded rapidly by ClpPA protease with a half-life of ϳ30 min, whereas MazF is stable over a period of 4 h. Therefore, down-regulation of the mazEF operon results in the persistence of MazF as the amount of its cognate antitoxin MazE decreases. This ultimately leads to cell growth arrest. In M. tuberculosis, the observations so far are promising in terms of mirroring the mazEF regulatory system of E. coli. However, the challenge will be to identify and confirm the cognate antitoxins for the endoribonuclease toxins reported here, as well as to determine the precise genetic organization and regulation of these sets of genes. Furthermore, the reason why MazF-mt1 was expressed well in M. smegmatis is not known at present. It is possible that M. smegmatis contains enough semicognate antitoxin against MazF-mt1 under the growth condition used or MazF-mt1 is highly specific and thus does not show global effect in M. smegmatis.
M. tuberculosis is one of the most devastating human pathogens, as one third of the world's population is infected with this pathogen and eight million people develop active disease each year. Two phenotypic properties of this pathogen account for these statistics: 1) latent infection sometimes referred to as dormancy and 2) persistence of infectious bacteria (14,15). Although the mechanisms of latency and persistence are not well understood, recent studies of the TA systems in this pathogen suggest that these TA systems may be responsible for the dormancy of this pathogen in human tissues (16). Our results demonstrating the existence of multiple mRNA interferases in M. tuberculosis support this notion, and further biochemical and genetic characterization of M. tuberculosis toxins will shed light into our understanding of their roles in the pathogenesis of this persistent human pathogen.