Correspondence: Masayori Inouye, Department of Biochemistry, Center for Advanced Biotechnology and Medicine, Robert Wood Johnson Medical School, Piscataway, NJ 08854, USA. Tel: +1 732 235 4115; fax: +732 235 4559/4783; e-mail: firstname.lastname@example.org
Nearly all free-living bacteria carry toxin–antitoxin (TA) systems on their genomes, through which cell growth and death are regulated. Toxins target a variety of essential cellular functions, including DNA replication, translation, and cell division. Here, we identified a novel toxin, YgfX, on the Escherichia coli genome. The toxin, consisting of 135 residues, is composed of the N-terminal membrane domain, which encompasses two transmembrane segments, and the C-terminal cytoplasmic domain. Upon YgfX expression, the cells were initially elongated and then the middle portion of the cells became inflated to form a lemon shape. YgfX was found to interact with MreB and FtsZ, two essential cytoskeletal proteins in E. coli. The cytoplasmic domain [YgfX(C)] was found to be responsible for the YgfX toxicity, as purified YgfX(C) was found to block the polymerization of FtsZ and MreBin vitro. YgfY, located immediately upstream of YgfX, was shown to be the cognate antitoxin; notably, YgfX is the first membrane-associating toxin in bacterial TA systems. We propose to rename the toxin and the antitoxin as CptA and CptB (for Cytoskeleton Polymerization inhibiting Toxin), respectively.
Nearly all free-living bacteria contain toxin–antitoxin (TA) systems on their genomes (Pandey & Gerdes, 2005). The sets of toxin and antitoxin proteins are most often encoded from a single operon. In all known cases, in normally growing cells, toxins form a stable complex with their cognate antitoxins that blocks the toxin activity. Antitoxin also functions as a repressor for individual TA operons (Gerdes et al., 2005). Under stress conditions, intrinsically unstable antitoxin is lost from the cells, releasing toxin freely and inhibiting various essential cellular functions, such as DNA replication, mRNA stability, protein synthesis, and cell division (Jiang et al., 2002; Zhang et al., 2003; Tan et al., 2011; Zhang & Inouye, 2011). This leads to a reversible cell growth arrest, which is implicated in the persister phenotype. The TA system is also shown to be associated with pathogenicity, programmed cell death, and biofilm formation (Pandey & Gerdes, 2005; Nariya & Inouye, 2008; Wang & Wood, 2011).
Escherichia coli have two essential bacterial cytoskeletal proteins, FtsZ and MreB. FtsZ is a highly conserved GTPase and is homologous to eukaryotic cytoskeleton protein, tubulin (Mukherjee et al., 1998). It forms a ring structure at the mid-cell and functions as a scaffold for divisome, a multiprotein complex responsible for cell division. MreB is an actin-like ATPase, essential for maintaining the typical rod shape and cell polarity in E. coli (Osborn & Rothfield, 2007). MreB is also implicated in chromosome segregation, localization of membranous organelles, and coordinating cell division with cell biosynthesis (Kruse et al., 2005; Komeili et al., 2006; Madabhushi & Marians, 2009; Domínguez-Escobar et al., 2011; Garner et al., 2011). Because both FtsZ and MreB are involved in a number of essential cellular functions, the inhibition of their functions is detrimental to the cells. For example, the inhibition of FtsZ polymerization by SulA or MinCD results in blocking the septum formation, causing the formation of filamentous cells (Mukherjee et al., 1998; Pichoff & Lutkenhaus, 2001). The inhibition of MreB by A22 [S-(3,4-dichlorobenzyl) isothiourea] leads to the loss of its rod shape and eventual cell lysis (Karczmarek et al., 2007; Bean et al., 2009).
Here, we have identified a novel TA system in E. coli genome using RASTA (Sevin & Barloy-Hubler, 2007). The putative toxin, YgfX, inhibits the cell growth and causes significant changes in the cellular morphology of E. coli. Upon induction of YgfX, the cells were first elongated and then subsequently became inflated in the middle. The YgfX toxicity was neutralized by the co-expression of YgfY, indicating that YgfY is an antitoxin of YgfX. YgfX is the first toxin of E. coli TA systems shown to be associated with membrane. We further demonstrated that YgfX physically interacts with FtsZ and MreB and inhibits their polymerization in vitro and that the C-terminal soluble domain of the YgfX is responsible for the inhibition. On the basis of these results, we propose to rename YgfX and YgfY as CptA and CptB (for Cytoskeleton Polymerization inhibiting Toxins A and B), respectively.
Materials and methods
Strains, plasmids, and growth conditions
Escherichia coli BW25113 (ΔaraBD) (Datsenko & Wanner, 2000) and BL21 (DE3) were grown in M9 medium supplemented with 0.2% casamino acids and 0.5% glycerol at 37 °C. The primers used in this study are summarized in Table 1. The coding sequences of ygfX alone or ygfYX were PCR-amplified using primers YGFX-F and YGFX-R1, or YGFY-F and YGFX-R1, respectively. The fragments were cloned into pBAD24 vector (Guzman et al., 1995) and designated as pBAD24-ygfX and pBAD24-ygfYX, respectively. The coding sequence of YgfX in a fusion with His6-tag at the C-terminal (YgfX−His) was also cloned into pBAD24 using YGFX-F and YGFX-R2. A truncated protein of YgfX (YgfX(C); cloned from V49 to Z135) was cloned into pCold-Km (unpublished results, Inouye laboratory) using YGFXs-F and YGFX-R1. His6-tagged FtsZ and MreB were constructed previously (Tan et al., 2011). FLAG-tagged FtsZ and MreB were also previously constructed in pET17b, having a tag at the C-terminal end (H. Masuda and M. Inouye, unpublished results). For examining the growth rate, 0.2% arabinose was added to the cultures during the early exponential phase.
Protein expression, purification, and pulldown assay
His6-tagged YgfX(C), FtsZ, and MreB were expressed in E. coli BL21(DE3). Protein expression was induced for 2 h by adding 1 mM IPTG when the OD600 nm reached 0.8. The cells were collected by brief centrifugation at 8000 g and lysed by French pressure press (Thermo Fisher Scientific, MA). FtsZ and MreB were purified as described before (Tan et al., 2011). YgfX(C)−HIS was purified from the insoluble materials after being dissolved in 8 M urea (pH 8.0). Proteins were purified using Ni-NTA agarose according to the manufacturer's instructions (Qiagen, CA).
Inner and outer membrane proteins were isolated following the method described previously (Hobb et al., 2009). Briefly, the total membrane proteins were collected from the lysate by ultracentrifugation at 100 000 g for 1 h. The pellet was washed, then resuspended in 1% (w/v) N-lauroylsarcosine in 10 mM HEPES, pH 7.4, and incubated at 25 °C for 30 min with gentle agitation. The inner and outer membrane fractions were further separated by ultracentrifugation.
His6-tag pulldown assays were carried out by incubating the cell lysate containing YgfX−HIS and the cell lysate containing FsZ−FLAG or MreB−FLAG (lysis buffer: 50 mM HEPES-KOH, pH 7.5, 10 mM MgCl2, 200 mM KCl, 0.1 mM EDTA, and 10% glycerol) overnight at 4 °C. Ni-NTA agarose (0.5 mL) was added to the lysate, and the mixture was incubated at room temperature for 1 h. The beads were washed three times with 20 mL of the same lysis buffer containing 20 mM imidazole. Protein complexes were then separated by 17.5% SDS-PAGE and visualized by Western blot using monoclonal anti-FLAG antibody conjugated with horseradish peroxidase (Sigma-Aldrich, MO).
The effect of YgfX on FtsZ and MreB polymerization was determined by a sedimentation method as previously described (Anand et al., 2004) with a few modifications. Purified FtsZ−HIS or MreB−HIS was mixed with different amounts of purified YgfX(C)−HIS in polymerization buffer P (5 mM MgCl2; 50 mM NaH2PO4; 100 mM NaCl; 20% glycerol, pH 6.5) and pre-incubated at room temperature for 30 min before 1 mM GTP or ATP was added to initiate the polymerization. The polymerization reaction was carried out at room temperature for 30 min. FtsZ or MreB polymers were precipitated by centrifugation at 100 000 g for 20 min, and the pellets were suspended in 50 μL of buffer P. Both the supernatant and pellet fractions were separated by a 17.5% SDS-PAGE, followed by Coomassie blue staining.
Cell morphology was observed using an Olympus BX40 microscope.
YgfX is localized in the inner membrane
YgfX contains a long hydrophobic segment at the N-terminal region from W16 to V54 (Fig. 1a). There are two Pro residues (P33 and P35) in the middle of the hydrophobic region, and thus, this protein likely forms a hydrophobic hair-pin structure with two transmembrane (TM) domains: TM1 from W16 to M32 and TM2 from L36 to V54. The presence of positively charged residues on either side of the putative TM segments suggests that N-terminal and C-terminal soluble domains of YgfX reside in cytosol (Fig. 1b). In order to experimentally determine the localization of YgfX, the full-size YgfX was expressed from arabinose inducible vector, pBAD24 (pBAD24-ygfX). After YgfX expression was induced by the addition of 0.2% arabinose for 2 h, the total membrane proteins were collected from the cellular lysate by ultracentrifugation. YgfX was found exclusively localized in the membrane fraction (lane 4, Fig. 2). Total membrane proteins were further separated into the inner and outer membrane fractions based on the solubility in 1% N-lauroylsarcosine (Hobb et al., 2009). As predicted, YgfX was shown to be localized in the inner membrane (lane 6, Fig. 2).
YgfX and YgfY show toxin and antitoxin activity, respectively
Intriguingly, the overexpression of YgfX caused growth arrest starting at 5 h postinduction (Fig. 3a). The growth arrest was accompanied by morphological change (Fig. 3b). After 1-h induction of YgfX expression from pBAD24-ygfX, some cells started to elongate. After 5 h, elongated cells were divided into smaller cells and simultaneously, cells became inflated in the middle or at the poles of cells. After overnight induction, cells became lemon shaped. We then examined whether YgfY can neutralize the toxicity caused by YgfX. First, the coding sequences of both ygfY and ygfX were cloned together in pBAD24. This construct did not show any growth inhibition at least for 48 h. The morphological change was also not observed. This result was confirmed by the expression of YgfX and YgfY separately from two independent plasmids. For this purpose, YgfY was cloned in a derivative of pCold vector (pCold-Km) and shown to be highly expressed (data not shown). Consistent with above experiments, cells expressing both YgfY and YgfX did not show any growth defect and alteration in morphology at least for 18 h, confirming that YgfY functions as an antitoxin for YgfX.
YgfX interacts with FtsZ and MreB
The morphological changes, observed after YgfX induction, were similar to what was observed in cells overexpressing YeeV (Tan et al., 2011). YeeV inhibits the cell division by blocking the polymerization of FtsZ and MreB. We thus examined whether YgfX also interferes with FtsZ and MreB functions. In order to assess the physical interaction between the YgfX and FtsZ or MreB, pulldown experiments were performed using the full-length YgfX, which was fused to a His6-tag (YgfX−HIS). The cell lysate of E. coli BL21 cells expressing YgfX−HIS was mixed with the cell lysate containing FtsZ−FLAG or MreB−FLAG. Protein complexes were purified with affinity chromatography, using Ni-NTA beads. Eluted proteins were analyzed by SDS-PAGE, and FLAG-tagged proteins were detected by Western blotting, with the use of the anti-FLAG antibody (Sigma-Aldrich). As a control, a lysate containing FtsZ−FLAG or MreB−FLAG was incubated with Ni-NTA beads without YgfX−HIS. As shown in Fig. 4a, FtsZ−FLAG or MreB−FLAG was detected in the elution fractions only when it was mixed with YgfX−HIS, indicating that YgfX interacts with FtsZ and MreB.
The interaction between FtsZ and YgfX was confirmed by yeast two-hybrid (Y2H) assay (James et al., 1996). The full-length and various truncated mutants of FtsZ were fused to the activation domain (AD) of pGAD-C1, while YgfX was fused to the binding domain (BD) of pGBD-C1. The interaction was assessed by monitoring the growth on selective media (SD-trp, -leu, -his supplemented with 25 mM 3-aminotriazole). The growth was observed when pGBD-ygfX was cotransformed with pGAD plasmid containing the full-length FtsZ as well as truncated variants of FtsZ, ΔC(−191), ΔC(−287), ΔN(−32), each lacking C-terminal 191, C-terminal 287, and N-terminal 31 residues, respectively (Fig. 4b). The interaction was lost when N-terminal 49 residues of FtsZ were deleted (ΔN(−49)). These results suggest that residues 33–96 of FtsZ are essential for the interaction with YgfX and that the majority of C-terminal residues and the first 31 N-terminal residues are dispensable for the interaction with YgfX.
YgfX(C) inhibits FtsZ and MreB polymerization
To directly assess the biological role of the interactions between YgfX and the cytoskeletal proteins, the effect of YgfX on in vitro polymerization of FtsZ and MreB was analyzed. To avoid the use of detergent to solubilize TM-containing full-length YgfX for polymerization assay, the soluble C-terminal 87-residue fragment (from V49 to R135) was cloned into pCold-Km. The clone was designed to express the truncated YgfX (YgfX(C)) in fusion with His6 tag at its N-terminal (YgfX(C)−HIS). YgfX(C)−HIS was produced at very high level in the cell; however, it was entirely localized in the inclusion bodies. In order to purify YgfX(C)−HIS, the insoluble fraction was collected by centrifugation and solubilized by 8 M urea. Solubilized YgfX(C)−HIS was then purified using Ni-NTA (Qiagen), which led to a high degree of purification (Fig. 5a).
The GTP-dependent polymerization of FtsZ was assessed by the sedimentation assay as described previously (Anand et al., 2004). FtsZ polymer was collected in the pellet fraction by ultracentrifugation (Fig. 5b). In the absence of YgfX, almost all FtsZ was polymerized and collected in the pellet fraction. However, when YgfX(C)−HIS was added to the reaction mixture, FtsZ polymer formation was decreased reciprocally to the amounts of YgfX(C)−HIS added. The polymerization of FtsZ was almost completely inhibited when YgfX(C)−HIS was added to FtsZ in the 1 : 1 molar ratio.
In a similar manner, the effect of YgfX on the ATP-dependent polymerization of MreB was analyzed. Addition of equimolar YgfX(C)−HIS almost completely inhibited MreB polymerization (Fig. 5c). These results clearly demonstrated that YgfX inhibits the GTP-dependent FtsZ polymerization, as well as ATP-dependent MreB polymerization, and that the C-terminal 87-residue cytoplasmic domain of YgfX is responsible for the inhibition of cytoskeletal polymerization.
YgfX is an inhibitor of FtsZ and MreB
Here, we identified a novel TA system, YgfY–YgfX, on the E. coli chromosome. The toxin, YgfX, was shown to inhibit cell division by interfering with the polymerization of essential bacterial cytoskeletal proteins, FtsZ and MreB. Unlike another recently identified soluble E. coli toxin, YeeV, which also interacts with FtsZ and MreB, YgfX is an inner membrane protein having two TM domains. This is consistent with the previous microscopic observation of GFP-YgfX, showing that YgfX is associated with the membrane (Kitagawa et al., 2005). In this study, we also demonstrated that YgfX inhibited FtsZ and MreB polymerization through its soluble C-terminal domain. The role of the TM domains of YgfX still has to be elucidated. The localization in the inner membrane may spatially limit the YgfX activity only near the membrane. For instance, Z-ring is known to be anchored to the inner membrane by ZipA (RayChaudhuri, 1999). A number of cell division proteins such as FtsW, FtsQ, FtsN, FtsL, FtsK, and FtsB also contain a TM domain(s) (Barondess et al., 1991; Dai et al., 1996; RayChaudhuri, 1999; Buddelmeijer & Beckwith, 2002; Bigot et al., 2004). Interestingly, spatially regulated inhibition of FtsZ polymerization by inner membrane–associated MinC is responsible for the localization of Z-ring at mid-cell (Bi & Lutkenhaus, 1993). YgfX may play a similar role in temporal and spatial control of FtsZ and MreB polymerization, thus regulating cell division events in vivo.
The interaction between FtsZ and YgfX was confirmed by Y2H assay. Furthermore, using Y2H assay, the region of FtsZ that is essential for the interaction with YgfX was analyzed. N-terminal 31 residues of FtsZ were not required for the interaction with YgfX. In contrast, N-terminal 31 residues are essential for the interaction with YeeV (Tan et al., 2011). This suggests that although both YeeV and YgfX target the same proteins (FtsZ and MreB) and cause equivalent morphological change, they bind distinct sites of FtsZ. The predicted secondary structures differ significantly between YeeV and YgfX. This raises the possibility that a number of different protein families can bind and modulate the activity of FtsZ and/or MreB.
The interaction between YgfX and MreB, however, could not be detected by Y2H in this study. It is likely because of the presence of large activating or BD, fused to N-terminal of YgfX and MreB, respectively. It is equally possible that the lack of the interaction is because of the low expression of YgfX in yeast. It was previously shown that the apparent interaction between YeeV and MreB was 10-fold less than the interaction between YeeV and FtsZ (Tan et al., 2011). In the case of YgfX, even the interaction with FtsZ, measured by β-galactosidase assay, was not as strong as the interaction between YeeV and FtsZ (data not shown). This apparent weaker interaction is unlikely due to a weak physical binding of YgfX with target proteins in E. coli, as the rate at which YgfX and YeeV cause morphological defects in E. coli was approximately the same.
YgfY neutralizes YgfX toxicity
Commonly, the regulation of the toxin activity occurs in two different ways: one through physical sequestration of toxin by antitoxin and the other by the autoregulatory mechanism of the toxin gene by the TA complex (Zhang et al., 2003; Makarova et al., 2006; Motiejūnaite et al., 2007). Although the toxicity of YgfX was neutralized by the co-expression of YgfY, the mechanism of how YgfY neutralizes the YgfX toxicity remains unknown. Interestingly, we could not detect the physical interaction between YgfX and YgfY, suggesting that YgfY may exert its antitoxin function at the level of transcription or by an unknown mechanism; notably, the X-ray structure of YgfY has been determined (Lim et al., 2005), predicting that YgfY is a DNA-binding protein. These observations are also similar to what was observed for yeeUV; YeeU and YeeV do not physically interact. The mode of neutralization of YeeV toxicity by YeeU is also predicted to involve the regulation at the level of transcription (Brown & Shaw, 2003).
Intriguingly, despite the lack of sequence similarity, YgfX and YeeV show the same mode of toxicity, and YgfY and YeeU share a similar mode of antitoxin mechanism. Interestingly, however, YeeV is a soluble protein, while YgfX is an inner membrane protein. Based on this different localization pattern, it is possible that YgfX may be able to exert its toxic function in a more specified manner than YeeV, as discussed above. Further study is necessary to characterize the physiological role of ygfYX. So far, no phenotype has been shown to be associated with the deletion of ygfYX. We speculate that this TA system may be involved in cell growth regulation under stress conditions, as in other TA systems. For instance, the expression of YgfYX is affected by norfloxacin, an inhibitor of DNA gyrase (Jeong et al., 2006). It is interesting to further investigate the importance of YgfYX under such conditions.
The authors thank Dr Peter Tupa for critical reading of the manuscript. This work was supported by grants from the National Institute of Health (RO1GM081567). H.M. was supported by NIH postdoctoral fellowship F32 GM095200.