Structure of a novel antibacterial toxin that exploits elongation factor Tu to cleave specific transfer RNAs.

Contact-dependent growth inhibition (CDI) is a mechanism of inter-cellular competition in which Gram-negative bacteria exchange polymorphic toxins using type V secretion systems. Here, we present structures of the CDI toxin from Escherichia coli NC101 in ternary complex with its cognate immunity protein and elongation factor Tu (EF-Tu). The toxin binds exclusively to domain 2 of EF-Tu, partially overlapping the site that interacts with the 3'-end of aminoacyl-tRNA (aa-tRNA). The toxin exerts a unique ribonuclease activity that cleaves the single-stranded 3'-end from tRNAs that contain guanine discriminator nucleotides. EF-Tu is required to support this tRNase activity in vitro, suggesting the toxin specifically cleaves substrate in the context of GTP·EF-Tu·aa-tRNA complexes. However, superimposition of the toxin domain onto previously solved GTP·EF-Tu·aa-tRNA structures reveals potential steric clashes with both aa-tRNA and the switch I region of EF-Tu. Further, the toxin induces conformational changes in EF-Tu, displacing a β-hairpin loop that forms a critical salt-bridge contact with the 3'-terminal adenylate of aa-tRNA. Together, these observations suggest that the toxin remodels GTP·EF-Tu·aa-tRNA complexes to free the 3'-end of aa-tRNA for entry into the nuclease active site.

INTRODUCTION

Bacteria have long been known to antagonize one another with soluble antibiotics and bacteriocins (1,2). These compounds diffuse through the environment and inhibit competitors at a distance. More recently, type IV, V, VI and VII secretion systems have been shown to mediate inter-cellular competition in a contact-dependent manner by delivering toxins directly into neighboring bacteria (3–6). This phenomenon was first described in Escherichia coli isolate EC93, which uses its CdiB/CdiA two-partner secretion proteins to bind and inhibit the growth of other E. coli strains (3). CdiB is an Omp85 transport protein that exports and assembles the CdiA effector onto the cell surface. CdiA is predicted to form a β-helical filament extending several hundred angstroms from the cell surface. CdiA binds to BamA on neighboring E. coli cells, then delivers its C-terminal toxin domain (CdiA-CT) into the target cell to inhibit growth (7,8). Because E. coli EC93 cells exchange toxins with neighboring siblings, they protect themselves from self-inhibition by producing a specific contact-dependent inhibition immunity (CdiI) protein that binds and neutralizes toxin. Since its discovery in E. coli EC93, cdi gene clusters have been identified and characterized in a variety of Gram-negative bacterial pathogens (9–13). Remarkably, CdiA-CT sequences are extraordinarily variable between bacteria, and isolates of the same species commonly deploy different toxins (14,15). Because immunity is conferred through specific protein–protein interactions (16–20), CdiI sequences are also highly variable between bacteria. The diversification of CdiA-CT and CdiI sequences is thought to reflect the selective pressure of inter-bacterial competition, with novel toxins conferring a fitness advantage.

In some instances, CdiA-CT domains require additional factors to promote toxic activity. For example, CdiA-CTEC536 from uropathogenic E. coli 536 is a potent tRNA anticodon nuclease in vivo (9), but the purified toxin has no RNase activity in vitro (21). CdiA-CTEC536 is activated when it binds to CysK, an O-acetylserine sulfhydrylase that catalyzes the final step in L-cysteine biosynthesis. CysK stabilizes the toxin fold and promotes interactions with tRNA substrates (19,21). Another CDI tRNase toxin from E. coli strain EC869 forms a high-affinity complex with elongation factor Tu (EF-Tu) (22). EF-Tu is an essential and highly conserved translation factor that delivers aminoacyl-tRNAs (aa-tRNA) to the ribosome during protein synthesis. CdiA-CTEC869 nuclease activity requires both EF-Tu and GTP, suggesting that GTP·EF-Tu·aa-tRNA ternary complexes are the physiologically relevant substrates (22). Genetic evidence from this latter study indicates that unrelated CDI toxins from E. coli isolates 96.154 and NC101 might also interact with EF-Tu. Here, we present structural and biochemical data that confirm the functional interaction between CdiA-CTNC101 and EF-Tu. Structures of the EF-Tu·CdiA-CT·CdiINC101 ternary complex reveal that the toxin binds EF-Tu at a site that partially overlaps with the aa-tRNA binding site. The toxin domain adopts the barnase/EndoU/colicin E5-D/RelE (BECR) RNase fold (23), and cleaves the single-stranded 3′-end of tRNAs that contain a guanine discriminator nucleotide. Because EF-Tu is required for this unusual nuclease activity, the toxin presumably recognizes substrate in the context of GTP·EF-Tu·aa-tRNA ternary complexes. However, superimposition of the toxin onto GTP·EF-Tu·aa-tRNA indicates that the nuclease active site is ∼12 Å from the scissile phosphodiester bond. Moreover, there are several predicted steric clashes between the 3′-end of aa-tRNA and EF-Tu bound toxin. We propose that CdiA-CTNC101 disrupts contacts between aa-tRNA and EF-Tu, thereby freeing the 3′-end of tRNA for toxin-mediated cleavage.

MATERIALS AND METHODS

Plasmid constructions and site-directed mutagenesis

Plasmids and oligonucleotides are listed in Supplementary Tables S1 and S2, respectively. The coding sequence for residues Val3035—Lys3289 of CdiANC101 (locus tag ECNC101_RS22895) and CdiINC101 (ECNC101_RS22895) from E. coli NC101 was synthesized by Genscript and supplied in plasmid pUC57. For crystallographic studies, the cdiA-CT/cdiINC101 module was amplified with primer pair PSI-1/PSI-2. The forward primer adds a 5′ ligation-independent cloning (LIC) sequence and an ATG start codon to the toxin coding sequence, and the reverse primer adds the 3′ LIC region coding for a short linker and the first two amino acids of the His6-tag to the CdiI coding sequence (24). The polymerase chain reaction (PCR) product was gel-purified, treated with T4 DNA polymerase and dTTP (25) and cloned into expression vector pMCSG58 (24) according to ligation-independent procedures (26,27). The resulting construct pMCSG58-CPX200205 was confirmed by DNA sequencing and transformed into E. coli BL21(DE3)-Gold cells for over-expression.

For biochemical and in vivo activity analyses, the cdiA-CT/cdiINC101 module was amplified with primers CH3880/CH3239 and CH3269/CH3422 and the resulting fragments introduced into pCH10068 and pCH7171 to generate plasmids pCH12641 and pCH12598, respectively. Site-directed mutagenesis of the cdiA-CTNC101 coding sequence was performed by mega-primer PCR or the QuikChange II Site-Directed Mutagenesis Kit (Stratagene) according to manufacturer’s instructions. The cdiINC101 gene was amplified with primers CH3238/CH3422 and ligated to pET21 using KpnI/SpeI restriction sites to generate plasmid pCH12745 for the purification of CdiINC101-His6. The E. coli cca gene encoding tRNA nucleotidyltransferase was amplified with primers CH4062/CH4063 and ligated to pET21 using KpnI/XhoI restrictions sites to generate plasmid pCH13076.

The chimeric cdiAEC93-CTNC101 expression construct (pCH11434) has been described previously (28). Site-directed mutant forms of this plasmid were constructed by allelic exchange of the counter-selectable pheS* marker from plasmid pCH10163 (16). cdiA-CT/cdiINC101 sequences were amplified with primers CH3176/CH3177 and fused to DNA fragments amplified from regions upstream and downstream of the cdiAEC93 gene. The upstream homology fragment was amplified using primers CH4100/CH4101, and the downstream fragment with primers CH4102/CH4103. The three products were then fused to each other by overlap-extension PCR (OE-PCR) using primers CH4100/CH4103. The final DNA product (100 ng) was electroporated together with plasmid pCH10163 (300 ng) into E. coli DY378 cells (29). Recombinant plasmid clones were selected on yeast extract glucose-agar supplemented with 33 μg/ml chloramphenicol and 10 mM d/l-p-chlorophenylalanine. All constructs were confirmed by DNA sequence analysis.

Protein purification and crystallization

The CdiA-CT·CdiINC101-His6 complex was expressed in E. coli BL21(DE3)-Gold cells from plasmid pMCSG58-CPX200205. Cells were cultured in LB medium supplemented with 100 μg/ml ampicillin at 37°C for 6–8 h, then diluted 1:100 into 50 ml of M9 minimal ‘pink’ medium supplemented with 0.5% glycerol, 50 μg/ml ampicillin and trace minerals and vitamins (30). Cells were grown to an optical density at 600 nm (OD600) of ∼0.8, and then the culture was cooled to 18°C. Seleno-methionine (SeMet) was added to a final concentration of 60 μg/ml along with l-isoleucine, l-leucine, l-lysine, l-phenylalanine, l-threonine and l-valine at 100 μg/ml. After 20 min, the culture was adjusted to 0.5 mM isopropyl-D-thiogalactopyranoside (IPTG) and incubated overnight. Cells were harvested by centrifugation, washed and resuspended in 50 mM Tris–HCl (pH 8.0), 500 mM NaCl, 10 mM 2-mercaptoethanol (2-ME), 10% glycerol. Cells were broken using Fast Break reagent (Promega), 10 μg/ml of lysozyme, 500 units of Benzonase Nuclease HC and Complete Protease Inhibitor Cocktail (Roche). Lysates were clarified by centrifugation and passage through a 0.22 μm filter, then loaded onto a Nickel (II) Sepharose HisTrap column equilibrated in resuspension buffer. Proteins were eluted with a linear 20–250 mM gradient of imidazole in resuspension buffer. Fractions containing the CdiA-CT·CdiINC101-His6 complex and co-purified EF-Tu were pooled and loaded onto a Hiload 26/60 Superdex 200 size-exclusion column equilibrated with 20 mM Tris–HCl (pH 7.5), 150 mM NaCl, 2 mM dithiothreitol. Fractions containing the ternary complex were pooled and concentrated to 15 mg/ml using an Amicon Ultracel 10K concentrator. Prior to crystallization, trypsin was added to the protein sample at a final concentration of 40 ng/μl, followed by incubation on ice for 4 h or overnight. Crystallization was performed at 4°C by sitting-drop vapor diffusion in 96-well Crystal Quick plates (Greiner Bio-one). Samples were digested for 4 h (EF-Tutr·CdiA-CT·CdiINC101), then crystallized in 0.05 M KCl, 0.1 M HEPES (pH 7.0), 1.0 M ammonium sulfate [optimized C2 conditions of Protein Complex Suite (Qiagen)]. The overnight digested complex (EF-Tu2,3·CdiA-CT·CdiINC101) was crystallized in 0.1 M Tris–HCl (pH 8.0), 1.5 M ammonium sulfate. These crystals were used as microseeds for further screening, according to protocol from Douglas Instruments Ltd. with the Mosquito nanoliter liquid handler (TTP Labtech). Crystals formed in Hampton Research Index Screen under conditions containing 0.1 M NaCl, 0.1 M Bis-Tris (pH 6.5), 1.5 M ammonium sulfate.

Data collection, structure solution and refinement

Crystals were cryoprotected in their mother liquor supplemented with either 3.0 M ammonium sulfate (EF-Tu2,3·CdiA-CT·CdiINC101, EF-Tutr·CdiA-CT·CdiINC101) or 28% sucrose (EF-Tu2,3·CdiA-CT·CdiINC101) and flash-cooled in liquid nitrogen. Diffraction data were collected at the Structural Biology Center 19-ID beamline at the Advanced Photon Source, Argonne National Laboratory. Single-wavelength anomalous diffraction (SAD) datasets were collected at 100K near the selenium K-absorption edge and diffraction images processed with the HKL3000 suite (31). Intensities were converted to structure factor amplitudes in Ctruncate (32,33) from the CCP4 package (34). Refinement statistics show that the structures are of good quality as presented in Table 1 (35,36). The EF-Tu2,3·CdiA-CT·CdiINC101 structure was solved using the HKL3000 software pipeline (31) by the SAD method with Se peak data merged from the two crystals. Heavy atom positions were determined in SHELXD and initial phases were obtained from SHELXE (37). Initially, 10 Se atoms were identified. After their positions were refined, eight of these sites were used to calculate improved phases through iterations of MLPHARE (38) and DM (39). The initial model was built in Buccaneer (40), and the final model obtained through iterative manual rebuilding in COOT (41) and crystallographic refinement in Refmac (34,42) against the dataset obtained from the crystal cryoprotected in sucrose. The refinement protocol included optimization of translation/libration/screw (TLS) motion parameters with six, five and three groups defined for CdiA-CTNC101 (chain A), CdiINC101 (chain B) and EF-Tu2,3 (chain C), respectively.

Table 1.

Data processing and refinement statistics

Processing
Structure	EF-Tu2,3·CdiA-CT·CdiINC101	EF-Tutr·CdiA-CT·CdiINC101
Wavelength (Å)	0.9793	0.9792
Resolution range (Å)a	30.00–2.35 (2.39–2.35)	30.00–3.30 (3.36–3.30)
Space group	P4212	P21
Unit cell parameters (Å, °)	a = 131.16 c = 63.33	a = 74.40 b = 128.35 c = 100.36 β = 109.6
Unique reflections	23688 (1155)	26374 (1279)
Multiplicity	9.1 (7.3)	3.7 (3.7)
Completeness (%)	100 (100)	99.3 (98.2)
<I >/< σI>	20.0 (2.1)	10.6 (1.5)
Wilson B factor (Å2)	37.2	63.3
R merge b	0.125 (0.953)	0.118 (0.959)
CC1/2c	0.734	0.714
CC*c	0.920	0.913
Refinement
Resolution (Å)	30.00–2.35	30.00–3.30
Reflections work/test set	22536/1125	24766/1008
R work/Rfreed	0.181/0.218	0.243/0.268
Average B factor (Å2) (No of atoms)
macromolecules	54.9 (3061)	116.5 (8885)
ligands	103.8 (6)	145.3 (56)
solvent	45.8 (92)
Rmsd bond lengths (Å)	0.015	0.010
Rmsd bond angles (°)	1.633	1.45
Ramachandran favorede (%)	97.9	96.8
Ramachandran outliers (%)	0	0
Clashscoree	1.64	3.54

aValues in parentheses correspond to the highest resolution shell.

b R merge = ΣhΣj|Ihj–|/ΣhΣjIhj, where Ihj is the intensity of observation j of reflection h.

cAs defined by Karplus (35).

d R = Σh|Fo|–|Fc|/Σh|Fo| for all reflections, where Fo and Fc are observed and calculated structure factors, respectively. Rfree is calculated analogously for the test reflections, randomly selected and excluded from the refinement.

eAs defined by Molprobity (36).

The EF-Tutr·CdiA-CT·CdiINC101 structure was solved by two-step molecular replacement in Phaser-MR (43,44) with the EF-Tu2,3·CdiA-CT·CdiINC101 structure used as a template in the first step and truncated domain 1 of EF-Tu from the 1ERC model (residues 8–41, 68–204) used in a second step. Subsequently, the resulting initial model was morphed into the electron density with the Phenix morph module (44,45). The structure was completed by real-space modeling in COOT and crystallographic refinement in Refmac with TLS option (three groups for EF-Tu (chains CD and GH) and two groups for each toxin domain (chains A and E) and CdiI (chains B and F)). To automatically generated local NCS restraints, the EF-Tu2,3·CdiA-CT·CdiINC101 structure was used to generate external restraints by ProSMART (46). Refinement statistics are shown in Table 1. The atomic coordinates and structure factors have been deposited in the Protein Data Bank under accession codes 5I4Q (EF-Tu2,3·CdiA-CT·CdiINC101) and 5I4R (EF-Tutr·CdiA-CT·CdiINC101).

Analysis of the initial anomalous map from two merged crystals of EF-Tu2,3·CdiA-CT·CdiINC101 and final anomalous map from a single crystal indicates that five of the identified heavy atom sites are contributed by toxin and immunity proteins. Two sites are totally incorrect and one represents a low occupancy Se atom from EF-Tu, with an anomalous signal detectable only in the two-crystal dataset. Thus, Met residues in CdiA-CTNC101 and CdiINC101 have been modeled as SeMet, whereas native Met residues were modeled in EF-Tu, which did not appear to incorporate SeMet. Similarly, only the toxin and immunity proteins appeared to contain SeMet residues in the EF-Tutr·CdiA-CT·CdiINC101 complex.

Protein purification for biochemical analyses

EF-Ts, EF-Tu and CdiA-CT·CdiINC101 were over-produced as fusions to His6-tagged thioredoxin, and tRNA nucleotidyltransferase (CCA-adding enzyme) carried a C-terminal His6 epitope to facilitate purification. Cultures of E. coli CH2016 harboring expression plasmids were grown to OD600 ∼ 0.7, and expression was induced with 1 mM IPTG. After incubation for 2 h, cells were harvested and frozen at −80°C. Cell pellets were re-suspended in lysis buffer [50 mM Tris–HCl (pH 7.5), 150 mM NaCl, 10 mM 2-ME, 0.05% Triton X-100, 20 mM imidazole] and broken by two passages through a French press at 20 000 psi. Cell debris was removed by two rounds of centrifugation at 16 000 ×g at 4°C. His6-tagged proteins were purified by Ni2+-affinity chromatography in lysis buffer. For the purification of CdiA-CTNC101 fusion protein, resins were washed with 20 mM Tris–HCl (pH 7.5), 6 M guanidine-HCl to denature and release CdiINC101 prior to imidazole elution. His6-tagged proteins were eluted with 20 mM Tris–HCl (pH 7.5), 250 mM imidazole and dialyzed against 20 mM sodium phosphate (pH 7.8), 150 mM NaCl, 10 mM 2-ME. Following dialysis, the purified fusion proteins were cleaved with His6-tagged TEV protease, and the protease and His6-TrxA fragment subsequently removed by Ni2+-affinity chromatography. Purified proteins were quantified by absorbance at 280 nm using the following extinction coefficients: EF-Tu, 20 400 cm−1 M−1; EF-Ts, 4470 cm−1 M−1; CdiA-CTNC101, 29 340 cm−1 M−1; CdiINC101, 28 420 cm−1 M−1; and tRNA nucleotidyltransferase, 55 920 cm−1 M−1.

In vivo toxin activity and competition co-cultures

Nuclease activity screens were performed by activating CdiA-CT toxins from E. coli isolates NC101, 3006 and 96.154 inside E. coli X90 cells. CdiI immunity proteins were tagged with a C-terminal ssrA(DAS) degron as described (47,48). Cells were grown in tetracycline-supplemented LB media for 1 h (OD600 ∼ 0.1), then cdiA-CT/cdiI-DAS expression was induced with 0.4% L-arabinose. Induced cultures were incubated at 37°C with shaking for 3 h, then harvested into an equal volume of ice-cold methanol. Cells were collected by centrifugation and frozen at −80°C for subsequent RNA extraction and analyses. This same approach was used to ascertain the toxicity of CdiA-CTNC101 domains harboring site-directed mutations. E. coli X90 cells carrying cdiA-CT/cdiI-DAS expression constructs were seeded at OD600 ∼ 0.05 in LB media supplemented with tetracycline and incubated with shaking at 37°C. After 60 min, the cultures were adjusted to 0.4% L-arabinose and cultured for an additional 4 h. Cell growth was monitored by measuring the OD600 of each culture every 30 min. Culture samples were harvested into an equal volume of ice-cold methanol after 2 h of induction and the cells frozen for subsequent RNA extraction.

Escherichia coli EPI100 inhibitor strains carrying cosmids pCH11434 (wild-type CdiA-CTNC101), pCH12796 (Tyr192Arg), pCH13353 (Gln198Ala), pCH13354 (Arg200Ala), pCH13355 (Glu236Ala), pCH13356 (Gln250Ala) or pCH12749 (His248Ala) were co-cultured at a 1:1 ratio with rifampicin-resistant E. coli MC4100 target cells harboring the empty vector pTrc99A (cdiI) or pCH12042 (cdiINC101) in shaking LB media without antibiotics. Viable target cells were enumerated on selective media as colony forming units (cfu) per ml at the beginning of co-culture and after 2 h. The data from three independent experiments are reported together with the mean ± standard errors. For the analysis of toxic tRNase activity, samples were harvested into an equal volume of ice-cold methanol after 30 min of co-culture. Cells were collected by centrifugation and frozen at −80°C for subsequent RNA extraction.

RNA isolation and analyses

Frozen cell pellets were resuspended in guanidinium isothiocyanate (GITC)-phenol and total RNA extracted as described (49). RNAs (5 μg) were run on 8 M urea 10% polyacrylamide gels buffered with 0.5× Tris-borate ethylenediaminetetraacetic acid (EDTA) and electroblotted to positively charged nylon membranes for Northern blot analysis. Blots were hybridized with [32P]-labeled oligonucleotide probes that are specific for individual E. coli tRNAs (see Supplementary Table S2) (49,50). Blots were visualized on a Bio-Rad phosphorimager using Quantity One software. S1 nuclease protection analysis was performed as described (50) using oligonucleotide CH3916 to map the cleavage site at the 3′-end of tRNAGlu. The S1 probe and marker oligonucleotides were first radiolabeled at their 3′-termini with [α-32P]-cordycepin triphosphate and terminal transferase, then de-salted with a G-25 spin column and 5′-phosphorylated with T4 polynucleotide kinase and unlabeled adenosine triphosphate (ATP). Radiolabeled probe was hybridized with RNA samples for 4 h at 50°C, then digested with S1 nuclease at 37°C for 30 min. Reactions were quenched with sodium acetate (pH 5.0) and precipitated with 90% ethanol. Reaction samples were run on 50% urea/10% polyacrylamide gels buffered with 0.5× Tris-borate-EDTA and visualized by phosphorimaging. In vitro transcripts of tRNAAsp were prepared using phage T7 RNA polymerase as described (22). Oligonucleotide CH4013 was annealed to CH4014 to generate a template for full-length tRNAAsp, and primers CH4013/CH4437 were annealed to generate the template for truncated tRNAAsp lacking the CCA nucleotide sequence. Each oligonucleotide cassette was end-filled with Klenow fragment to produce double-stranded transcription templates.

In vitro enzyme assays

In vitro nuclease assays were performed in reaction buffer [20 mM Tris–HCl (pH 7.5), 100 mM NaCl, 5 mM MgCl2, 10 mM 2-ME, 1 mM GTP and 100 μg/μl bovine serum albumin]. Purified CdiA-CTNC101, EF-Tu and/or EF-Ts were diluted to 1 μM in reaction buffer, and CdiINC101 was used at 3 μM. Protein mixtures were equilibrated for 30 min at room temperature. Reactions were then initiated by addition of E. coli total RNA to final concentration of 800 ng/μl, followed by incubation for 1 h at 37°C. Reactions were quenched with an equal volume of 25 sodium dodecyl sulphate-urea gel loading buffer and run on 50% urea, 7.5% polyacrylamide gels buffered with 0.5× Tris-borate-EDTA. Gels were electroblotted to nylon membranes for hybridization with radiolabeled oligonucleotide probes as described above. Toxin-digested RNA was purified by GITC–phenol extraction, then incubated with 2 μM tRNA nucleotidyltransferase in 50 mM Tris–HCl (pH 8.0), 5 mM MgCl2, 10 mM 2-ME, 1 mM ATP, 1 mM CTP at 37°C for 30 min. Reactions were quenched with gel-loading buffer and analyzed by Northern blot. For GTPase activity assays, purified CdiA-CTNC101, EF-Tu and EF-Ts were used at 0.5 μM final concentration and reactions were supplemented with 15 μM GTP/0.033 μM [α-32P]-GTP. Reactions were initiated by the addition of E. coli total RNA to a final concentration of 0.9 μg/μl and incubated at ambient temperature for 10 min. Positive control reactions were conducted under the same conditions with oligonucleotide CH1737 (5 μM) and polynucleotide kinase. Reactions were resolved by thin-layer chromatography on polyethyleneimine cellulose (Polygram cel 300 PEI/UV254) using 0.3 M sodium phosphate (pH 3.5) as the mobile phase. Chromatograms were visualized by phosphorimaging.

Protein–protein binding was assessed by co-purification during Ni2+-affinity chromatography. Proteins were mixed together at 5 μM final concentrations in binding buffer [20 mM Tris–HCl (pH 7.5), 150 mM NaCl, 20 mM imidazole, 5 mM MgCl2, 1 mM GTP, 10 mM 2-ME, 0.05% Triton X-100] and incubated for 30 min at room temperature. Ni2+-NTA agarose resin was added and incubated for 30 min at room temperature. Resins were washed with binding buffer, and column-bound proteins were eluted in binding buffer supplemented with 500 mM imidazole. Samples of the initial protein mixture, the unbound fraction and the eluted proteins were analyzed by sodium dodecyl sulphate-polyacrylamide gel electrophoresis and Coomassie blue staining.

RESULTS

Crystallization and structure determination

Isolation of the CdiA-CT·CdiINC101 complex by Ni2+-affinity chromatography resulted in the co-purification of EF-Tu (Supplementary Figure S1A), suggesting that the toxin and immunity proteins form a stable ternary complex with this translation factor. EF-Tu is the most abundant cytosolic protein in E. coli, and is apparently present in sufficient quantity to form a stoichiometric complex with over-produced CdiA-CT·CdiINC101. To facilitate crystallization, the EF-Tu·CdiA-CT·CdiINC101 complex was subjected to limited proteolysis with trypsin. This treatment removed the N-terminal ‘translocation’ domain (Val1–Lys163) from CdiA-CTNC101 (28), leaving the ∼9.9 kDa C-terminal domain in the complex. Trypsin also cleaved EF-Tu at Arg45 and Arg59 to yield EF-Tutr, a previously characterized form that lacks the flexible switch I region (51–56). Upon prolonged digestion, the entire N-terminal GTPase domain was removed from EF-Tu to leave an ∼24 kDa fragment composed of β-barrel domains 2 and 3 (EF-Tu2,3). EF-Tu2,3·CdiA-CT·CdiINC101 crystallized in space group P4212 with a single copy of the assembly in the asymmetric unit. The structure was solved by SAD phasing with Se-Met labeled protein to a resolution of 2.35 Å (Table 1 and Supplementary Figure S2A). The resulting model includes residues Gly168–Lys255 of CdiA-CTNC101, Met1–Ala107 of CdiINC101 and Lys209–Gly394 of EF-Tu (Figure 1A). Crystal packing suggests that the ternary complex is a dimer of heterotrimer units related by a crystallographic two-fold axis. This is consistent with size-exclusion chromatography data showing that the mass of the complex is >158 kDa (Supplementary Figure S1B). EF-Tutr·CdiA-CT·CdiINC101 crystallized in space group P21 with two complexes per asymmetric unit, again indicating that the biological assembly is a dimer of heterotrimers. The latter structure was solved by molecular replacement using EF-Tu2,3·CdiA-CT·CdiINC101 and domain 1 of EF-Tu as models and refined to a resolution of 3.30 Å (Table 1 and Supplementary Figure S2B). The final model includes all EF-Tu residues except Met1–Arg8 and Ala46–Arg59. EF-Tutr adopts the GDP-bound conformation and two guanosine diphosphate molecules were identified in the structure (Figure 1B and Supplementary Figure S2C). Heterotrimeric assemblies from the two models superimpose with an average rmsd of 0.4 Å, with the toxin domains exhibiting the best fits and EF-Tu fragments showing the largest deviations between crystal forms (Supplementary Table S3). Because the structures are very similar, the following descriptions focus on the higher resolution EF-Tu2,3·CdiA-CT·CdiINC101 model unless stated otherwise.

Figure 1.

Structure of the EF-Tu·CdiA-CT·CdiINC101 complex. (A) Monomeric EF-Tu2,3·CdiA-CT·CdiINC101 complex in a cartoon representation. EF-Tu secondary structure elements are indicated according to (60). (B) Dimeric EF-Tutr·CdiA-CT·CdiINC101 complex in a cartoon representation. GDP molecules are depicted as spheres with carbon atoms colored pink. (C) The EF-Tu2,3·CdiA-CT·CdiINC101 dimer is rotated 90° with respect to panel B.

Structure of the CdiA-CT·CdiINC101 complex

The C-terminal domain of CdiA-CTNC101 is globular with a central seven-stranded mixed β-sheet decorated by two α-helices (Figure 1A). Helix α1 caps the sheet and helix α2 runs parallel to its curvature. Extended loop L5 connects β4–β5 and forms a large pocket together with helix α1 on the C-terminal face of the toxin domain. DALI server searches revealed that the closest structural homologs of the toxin are BrnT from Brucella abortus (Z = 7.1) and the C-terminal tRNase domains of colicin D (Z = 5.4) and colicin E5 (Z = 4.7) (Supplementary Table S4 and Figure S3). The toxin domain is also similar to a number of mRNA interferases from E. coli, including MqsR (Z = 4.5), YoeB (Z = 4.1) and RelE (Z = 3.3). All of these homologs share the BECR superfamily fold (23).

The CdiINC101 immunity protein consists of four α-helices that form an elongated club-like structure. Helices α2, α3 and α4 form a globular head from which the long N-terminal α1 helix protrudes (Figure 1A). CdiINC101 mediates dimerization of the ternary complex through helices α1 and α2, which form a V-shaped structure that accepts the N-terminus of α1 from the adjacent immunity protomer (Figure 1B and C). The CdiINC101 dimer interface buries ∼1300 Å2, with a predicted interaction free energy of −22.4 kcal/mol as calculated by ePISA (57). The globular head of CdiINC101 fits into the pocket on the C-terminal face of the toxin domain, burying ∼800 Å2 of surface area (ΔG = −2.7 kcal/mol). The toxin-immunity protein interaction is mediated primarily through a network of hydrogen-bonds (H-bonds) and salt bridges (Supplementary Table S5). CdiINC101 residue Asp84 forms a salt bridge with toxin residue Arg200 and makes an additional direct contact with His248 (Figure 2A). The side-chain of CdiINC101 Arg85 forms a prominent salt bridge with Asp220 from loop L5 of the toxin domain. Searches for structural homologs of CdiINC101 identified BAG1 co-chaperones, a translin family nuclease from Nanoarchaeum and a human ADP-ribosylation factor binding protein (Supplementary Table S4). All of these structural homologs are monomeric and have functions that are unrelated to CdiINC101.

Figure 2.

Intermolecular interactions. (A) The CdiA-CT·CdiINC101 interface. (B) Superimposition of EF-Tu2,3 structures from free EF-Tu (PDB ID: 1EFC) and EF-Tu2,3·CdiA-CT·CdiINC101 (PDB ID: 5I4Q). (C) The EF-Tu·CdiA-CTNC101 interface. (D) Interactions between EF-Tu2,3 (blue) and the CdiINC101 dimer (cyan and pink).

CdiA-CT/CdiINC101 interactions with EF-Tu

In complex with CdiA-CT·CdiINC101, EF-Tu adopts the same overall conformation reported for the GDP-bound form (58–61), though the loops connecting strands a2 to b2 and d2 to e2 are displaced ∼6 Å and ∼9 Å (respectively) by the toxin domain (Figure 2B). The N-terminal face of the toxin interacts exclusively with EF-Tu domain 2 and buries ∼1000 Å2 (ΔG = −13.9 kcal/mol). CdiA-CTNC101 helix α2 packs onto the surface of the β-barrel, and strand β1a complements the exposed d2-d2x β-hairpin of EF-Tu to form an antiparallel three-stranded sheet (Figures 1A and 2C). The EF-Tu·CdiA-CTNC101 interface is coordinated by 13 H-bonds and ion-pairs (Supplementary Table S5). Lys171 within toxin strand β1a forms a salt bridge with EF-Tu residue Asp267, and toxin residue Tyr192 forms H-bonds with the side-chain of Glu260 and the main-chain amide of Arg263 in EF-Tu. The interface is further stabilized by hydrophobic interactions, including π–π stacking of the aromatic rings of toxin residue Trp196 and Phe262 of EF-Tu (Figure 2C).

Each EF-Tu molecule also interacts directly with both protomers of the CdiINC101 dimer. Within the heterotrimeric assembly, helices α1 and α2 of CdiINC101 engage EF-Tu domains 3 and 2, respectively (Figure 2D). These interactions are anchored by ion-pairs between Glu6 and Glu69 of the immunity protein and Arg328 and Arg284 of EF-Tu. Helix α1′ from the complementary immunity protomer occupies the groove between the EF-Tu β-barrel domains, and its intermolecular contacts are dominated by the guanidinium of Arg23, which interacts directly with Glu288, Thr336 and Asp337 of EF-Tu (Figure 2D and Supplementary Table S5). Stacking between CdiINC101 Arg15 and EF-Tu Arg289 is also observed, in addition to a few weaker H-bonds and hydrophobic interactions. Overall, the two EF-Tu·CdiINC101 interfaces total ∼630 Å2.

CdiA-CTNC101 is a specific tRNase

The structural similarity between CdiA-CTNC101 and the BECR superfamily indicates that the toxin is likely a ribonuclease. The interaction with EF-Tu further suggests that tRNA is the preferred substrate. To identify substrates, we used controlled proteolysis to degrade CdiINC101 carrying a C-terminal ssrA(DAS) degron, thereby activating the toxin in vivo as described previously (47,48). Northern blot analysis revealed that tRNAAsn, tRNAAsp, tRNAGlu, tRNAUCUArg and tRNASer isoacceptors are cleaved in response to toxin activation (Figure 3A and Supplementary Figure S4). Limited activity was also observed on tRNACUGGln and tRNATrp (Figure 3A and Supplementary Figure S4). We next tested whether these tRNA molecules are also cleaved in target bacteria during co-culture with inhibitor cells that deploy the CdiA-CTNC101 toxin. We mixed inhibitor and target bacteria at a 1:1 ratio in shaking broth for 1 h, then isolated total RNA from the mixed population for Northern blot analysis. Cleaved tRNAs were detected in co-cultures with non-immune target cells (Figure 3B). We note that only half of any given tRNA is cleaved in these samples, because inhibitor cells produce CdiINC101, which neutralizes toxin activity in this population (28). Accordingly, tRNA in target bacteria was protected from degradation when these cells expressed cdiINC101 from a plasmid vector (Figure 3B). These results show that several tRNAs are efficiently cleaved in target bacteria, suggesting that these molecules are physiologically relevant substrates.

Figure 3.

CdiA-CTNC101 is a specific tRNase. (A) CdiA-CTNC101 expression was induced in Escherichia coli and total RNA isolated at intervals for Northern blot analysis of the indicated tRNAs. (B) Inhibitor cells that deploy CdiA-CTNC101 were co-cultured at a 1:1 ratio with sensitive (cdiI) and immune (cdiI) target cells for 30 min, then total RNA was isolated for Northern blot analysis. Anticodon sequences are provided in parentheses to indicate specific tRNA isoacceptors. (C) Inhibitor cells that deploy CdiA-CTNC101 were co-cultured at a 1:1 ratio with target bacteria that over-express tRNAUCUArg, tRNACCUArg or tRNATyr. Total RNA was then isolated for Northern blot analyses. tRNAGlu was monitored as a control to confirm toxin delivery into target cells. The sample in lane 1 (cdiI+) was isolated from a co-culture containing target bacteria that express the cdiINC101 immunity gene. Aminoacyl acceptor stems and discriminator nucleotides (in red) are presented on the right.

The slight increase in electrophoretic mobility observed for toxin-cleaved tRNA indicates that only a few nucleotides are removed. We used S1 nuclease protection to map the cleavage site on tRNAGlu isolated from cells that were intoxicated by internal CdiA-CTNC101 expression (Figure 3A). This analysis revealed that four nucleotides are removed from the 3′-end, truncating the tRNA after nucleotide C72 (Figure 4B and C). The same cleavage site was observed in tRNAGlu isolated from a competition co-culture with non-immune target bacteria (Figures 3B and 4B). We noted that preferred substrates all contain a guanine discriminator nucleotide adjacent to the cleavage site, suggesting this position is a specificity determinant. Therefore, we examined toxin activity on tRNAUCUArg and tRNACCUArg, which contain guanine and adenine discriminators, respectively (Figure 3C). We co-cultured inhibitor cells with target bacteria that over-produce each isoacceptor from a plasmid vector, then monitored tRNA degradation by Northern blot. Because tRNAUCUArg and tRNACCUArg are not abundant in wild-type E. coli cells (62), the hybridization signal is dominated by the over-produced tRNA in the target cell population. tRNAUCUArg was completely cleaved after 1 h of co-culture, whereas tRNACCUArg was unaffected (Figure 3C, lanes 2 and 4). We also probed these samples for tRNAGlu to confirm that toxin was delivered into target bacteria during the co-culture (Figure 3C, lower blot). We then mutated the discriminator nucleotide in each isoacceptor to ascertain its importance in substrate recognition. The G74A substitution protected tRNAUCUArg from degradation, whereas tRNACCUArg carrying the A72G mutation was efficiently cleaved (Figure 3C, lanes 3 and 5). Moreover, tRNATyr was converted into a toxin substrate with a A82G discriminator mutation (Figure 3C, lanes 6 and 7). These results demonstrate that CdiA-CTNC101 possesses a highly specific nuclease activity that cleaves tRNA at the 5′-side of guanine discriminator nucleotides.

Figure 4.

CdiA-CTNC101 cleaves the tRNA acceptor stem. (A) tRNAGlu sequence showing the hybridized S1 probe and oligonucleotide standards used to map the toxin cleavage site. (B) S1 nuclease protection assays. RNA was isolated from competition co-cultures and cells intoxicated by intracellular CdiA-CTNC101 expression. Samples from in vitro tRNase reactions were also analyzed. Where indicated, the neutralizing effect of CdiINC101 immunity protein was examined. RNA samples were incubated with the 3′-radiolabeled S1 probe and treated with S1 nuclease as described in ‘Materials and Methods’ section. A portion of the S1 probe-tRNAGlu heteroduplex sequence is shown to the right of the autoradiogram. (C) The arrow indicates cleavage site within the acceptor stem of tRNAGlu.

CdiA-CTNC101 nuclease active site

Superimposition of CdiA-CTNC101 onto BrnT suggests that the nuclease active site is located on the C-terminal face of the domain (Supplementary Figure S3). We reasoned that toxin residues Arg200, Glu236, His248 and Gln250 could participate directly in catalysis because their side-chains are clustered (Figure 2A), and they are completely conserved in homologous domains (Supplementary Figure S5). We also considered a role for Gln198, which is similarly positioned for catalysis and conserved in several related toxins (Supplementary Figure S5). We mutated each residue to Ala and examined in vivo activities using controlled proteolysis of degron-tagged CdiINC101. The His248Ala and Arg200Ala mutations abrogated growth inhibition activity, but the other mutations had no significant effect on toxicity (Figure 5A). Northern blot analysis confirmed that Arg200 and His248 are critical for nuclease activity, whereas the Gln198Ala, Glu236Ala and Gln250Ala mutations had no discernable effect on tRNA cleavage (Figure 5B). Similar results were obtained when the mutations were introduced into full-length CdiA and tested in CDI competition co-cultures. Cells that deploy the His248Ala and Arg200Ala domains failed to inhibit target bacteria, whereas the Gln198Ala and Gln250Ala toxins were as effective as wild-type toxin (Figure 5C). In contrast to its potent growth inhibition activity when expressed internally, we found that the Glu236Ala domain was attenuated when delivered into target cells during CDI (Figure 5C). To exclude the possibility that mutant CdiA function was abrogated by low expression or defective secretion, we examined the inactive effectors by immunoblot analysis using antibodies against the N-terminal domain of CdiA (63). This analysis showed that each CdiA variant accumulated to approximately the same level as wild-type (Supplementary Figure S6). Moreover, treatment with extracellular proteinase K revealed that each CdiA protein was exported to the cell surface (Supplementary Figure S6). Taken together, these results indicate that Arg200 and His248 are critical for tRNase activity and that Glu236 significantly contributes to toxicity in the context of cell-mediated CDI.

Figure 5.

Mutagenesis of the CdiA-CTNC101 active site. (A) Expression of the indicated toxin variants was induced at 60 min with L-arabinose, and cell growth was monitored by measuring the optical density of the cultures at 600 nm (OD600). (B) Northern blot analysis of RNA isolated at 180 min from the cultures in panel A. (C) Target bacteria were co-cultured at a 1:1 ratio with inhibitor cells that deliver the indicated CdiA-CTNC101 toxin variants. Where indicated (+), target cells carried a plasmid-borne copy of the cdiINC101 immunity gene. Viable target cells were quantified as colony forming units per ml at 0 and 2 h. Data from three independent experiments and the bars indicate the standard error of the mean.

CdiA-CTNC101 tRNase activity is dependent on EF-Tu and EF-Ts

Because CdiA-CTNC101 binds to EF-Tu and has tRNase activity, we reasoned that the toxin probably requires EF-Tu to cleave substrate. Consistent with this hypothesis, CdiA-CTNC101 has no detectable nuclease activity in defined reactions containing purified toxin and total RNA isolated from E. coli (Figure 6A, lane 2). However, inclusion of purified EF-Tu failed to stimulate tRNase activity (Figure 6A, lane 3). We previously reported that mutations in the coiled-coil domain of elongation factor Ts (EF-Ts) provide resistance to CdiA-CTNC101 toxin (22), suggesting that EF-Ts may also be required for nuclease activity. Reactions supplemented with both EF-Tu and EF-Ts supported efficient cleavage of tRNAGluin vitro (Figure 6A, lane 5). This activity was neutralized with purified CdiINC101 immunity protein (Figure 6A, lane 6), indicating that the reaction is catalyzed by CdiA-CTNC101 rather than contaminating nucleases in the various protein preparations. Moreover, this activity was indistinguishable from that detected in vivo, with tRNAGlu cleaved after residue C72 (Figure 4B). We note that S1 nuclease protection also showed intermediate cleavage products, suggesting that CdiA-CTNC101 may act as an exonuclease to progressively remove residues from the 3′-end of the substrate. However, all tRNA molecules carry the same 3′ CCA sequence, so it is unclear why only a subset is cleaved in vivo (Supplementary Figure S4). One possibility is that the toxin cleaves CCA non-specifically, but cannot remove the discriminator nucleotide from non-preferred substrates. In this model, tRNA nucleotidyltransferase would rapidly repair damaged molecules that retain the discriminator nucleotide, which is required for CCA-addition (64). We explored this model by first asking whether other substrates are degraded in vitro, where nucleotidyltransferase is not available for repair. We treated E. coli total RNA with purified toxin, EF-Tu and EF-Ts and analyzed several tRNAs by Northern blot. tRNAAsp and tRNAGlu were cleaved efficiently under these conditions, but none of the other tRNAs were degraded (Figure 6B). Moreover, tRNA nucleotidyltransferase was unable to repair toxin-cleaved tRNAAsp and tRNAGlu (Figure 6C), though it readily extended truncated tRNAAsp transcripts that retain the discriminator nucleotide (Figure 6D). Together, these results suggest that CdiA-CTNC101 is an endonuclease that cleaves the single-stranded 3′-tail from specific tRNAs. Thus, the additional products observed in Figure 4B probably reflect incomplete S1 digestion of the tRNA:probe heteroduplex, rather than toxin degradation intermediates.

Figure 6.

In vitro tRNase and GTPase assays. (A) Purified CdiA-CTNC101, EF-Tu, EF-Ts and CdiINC101 were incubated in GTP-supplemented buffer for 15 min. Total Escherichia coli RNA was then added and the solutions incubated at 37°C for 1 h. Reactions were analyzed by Northern blot for tRNAGlu. (B) In vitro nuclease reactions were conducted as in panel A and reactions analyzed by Northern blot for the indicated tRNA species. Anticodon sequences are provided in parentheses to indicate specific isoacceptors. (C) RNA was isolated from in vitro nuclease reactions, then treated with tRNA nucleotidyltransferase (tRNA NT) in ATP/CTP supplemented buffer for 1 h at 37°C. Reactions were analyzed by Northern blot with probes for tRNAAsp and tRNAGlu. (D) tRNA nucleotidyltransferase is active on truncated tRNA that retains the discriminator nucleotide. tRNAAsp truncated at residue G74 was prepared by in vitro transcription and incubated with tRNA nucleotidyltransferase (tRNA NT) as in panel C. The reaction was run on a denaturing polyacrylamide gel and visualized by ethidium bromide staining. Full-length tRNAAsp transcript was run in the left lane for comparison.

The finding that EF-Ts promotes tRNase activity raises the possibility that CdiA-CTNC101 forms a ternary complex with EF-Tu and EF-Ts. However, we did not detect EF-Ts·EF-Tu·CdiA-CTNC101 complexes by Ni2+-affinity chromatography in the presence of either GDP or GTP (Supplementary Figure S7A). Surprisingly, we also did not observe a binary EF-Tu·CdiA-CTNC101 complex in these latter experiments (Supplementary Figure S7A). However, when purified CdiINC101 was included in the protein mixture, we isolated stable EF-Tu·CdiA-CT·CdiINC101 ternary complexes (Supplementary Figure S7B). Thus, the toxin has lower affinity for EF-Tu in the absence of its immunity protein, suggesting that the EF-Tu·CdiINC101 interfaces contribute significantly to complex stability. We also considered the possibility that EF-Ts promotes toxin activity in its capacity as a guanine nucleotide exchange factor (GEF) for EF-Tu. In this model, the toxin induces EF-Tu GTPase activity, and EF-Ts accelerates the nuclease reaction by reloading EF-Tu with GTP for additional turnover events. Therefore, we tested whether GTPase activity is stimulated during the tRNase reaction. We supplemented in vitro nuclease reactions with radiolabeled [α-32P]-GTP and followed its conversion into [α-32P]-GDP using thin-layer chromatography. However, we did not detect an increase in GDP production under conditions that support efficient tRNase activity (Figure 7, compare lane 4 to lanes 1–3). These data indicate that the role of EF-Ts in toxin-catalyzed tRNA cleavage does not require its intrinsic GEF activity.

Figure 7.

GTPase assays. tRNase reactions were conducted as described in Figure 6, but the buffer was supplemented with radiolabeled [α-32P]-GTP. A polynucleotide kinase (PNK) reaction using [α-32P]-GTP as the phosphoryl donor was included as a positive control. Reactions were analyzed by thin-layer chromatography (upper panel) and Northern blot hybridization (lower panels). The migration positions of GDP and GTP are indicated.

Though the isolated toxin domain has relatively low affinity for EF-Tu (Supplementary Figure S7A), in vitro assays indicate that the translation factor is clearly required for nuclease activity. To confirm this requirement, we introduced a Tyr192Arg mutation into the toxin to disrupt its EF-Tu-binding interface (Figure 2C). The resulting domain retained its affinity for His6-tagged CdiINC101, which facilitated purification of this toxin variant. However, EF-Tu no longer co-purified with the Tyr192Arg toxin-immunity protein complex (data not shown). Furthermore, the EF-Tu·CdiA-CT·CdiINC101 ternary complex could not be reconstituted from purified constituents with the Tyr192Arg toxin domain (Supplementary Figure S7C), demonstrating that the mutation disrupts the interaction with EF-Tu. The purified Tyr192Arg toxin showed no detectable tRNase activity in reactions supplemented with both EF-Tu and EF-Ts, similar to the inactive domain carrying the His248Ala mutation (Figure 6A, lanes 7 and 8). The Tyr192Arg variant also lacked tRNase activity when expressed inside E. coli (Figure 5A and B), and did not inhibit target bacteria in CDI competition co-cultures (Figure 5C). Collectively, these results demonstrate that EF-Tu is required for CdiA-CTNC101 tRNase activity.

DISCUSSION

The CdiANC101 effector carries an RNase domain that interacts with EF-Tu and cleaves the 3′-ends of specific tRNAs. Given that EF-Tu is required for in vitro tRNase activity, these findings suggest that substrate is cleaved only in the context of GTP·EF-Tu·aa-tRNA ternary complexes. In this model, the binding site on EF-Tu provides access to the tRNA acceptor stem. The 3′-end of aa-tRNA binds in the cleft between domains 1 and 2 of EF-Tu (59,60), and superimposition of the toxin onto GTP·EF-Tu·aa-tRNA shows that the nuclease active site is directed toward the acceptor stem (Figure 8A). However, the putative active-site imidazole of His248 is predicted to be ∼16 Å from the 2′-hydroxyl of C72 and ∼12 Å from the scissile phosphodiester bond. Further, the superimposition reveals several intermolecular clashes. Toxin helix α1 interferes with the switch I region in EF-Tu domain 1 (Figure 8B), and the N-terminus of EF-Tu clashes with the inter-molecular β-sheet (Figure 8C). The loop connecting toxin helices α1 and α2 clashes with the 3′-end of aa-tRNA (Figure 8C), and displacement of the d2-d2x β-hairpin should break the ion-pair linking Arg263 of EF-Tu to the 5′-phosphate of the terminal adenosine ribonucleotide (Figure 8D). Thus, the nuclease domain is predicted to disrupt contacts between the 3′-end of aa-tRNA and EF-Tu. Moreover, the clash with switch I suggests that the toxin could induce conformational changes in EF-Tu similar to those that occur during decoding on the ribosome. The β-turn connecting a2 and b2 of EF-Tu is relocated upon codon-anticodon recognition, distorting the interaction between the 3′-end of aa-tRNA and switch I to induce GTP hydrolysis (65,66). Following hydrolysis, domain 1 rotates ∼100° relative to domains 2/3, and EF-Tu adopts an extended conformation with low affinity for aa-tRNA (58,67). Toxin-induced rearrangement of the a2-b2 β-turn is more pronounced than that observed for ribosome-bound EF-Tu, yet the nuclease domain does not stimulate GTP hydrolysis. Together, these observations suggest that the toxin remodels GTP·EF-Tu·aa-tRNA complexes to disrupt interactions between the translation factor and the 3′-end of aa-tRNA. Once liberated from its contacts with EF-Tu, the single-stranded tail of substrate tRNA is presumably free to enter the nuclease active site.

Figure 8.

CdiA-CTNC101 disrupts contacts between EF-Tu and aa-tRNA. (A) Superposition of GDP·EF-Tutr·CdiA-CT·CdiINC101 onto the GDPNP·EF-Tu·aa-tRNA ternary complex from Thermus aquaticus (PDB ID: 1B23). For clarity, only CdiINC101 is shown as a dimer. The expanded region shows CdiA-CTNC101 clashing with domain 1 of EF-Tu and tRNA. Guanine nucleotides are shown as spheres. (B) Potential clashes between CdiA-CTNC101 helix α1 and the switch I region of EF-Tu. (C) Superimposition of EF-Tu in tRNA-bound (TutRNA) and toxin-bound (Tutoxin) states. The predicted clash between CdiA-CTNC101 helix α2 and residues C75 and A76 at the 3′-end of tRNA is illustrated. (D) Rearrangement of the aa-tRNA binding site. EF-Tu from the EF-Tu2,3·CdiA-CT·CdiINC101 structure is superimposed on the kirromycin-stabilized GTP·EF-Tu·aa-tRNA complex from Escherichia coli (PDB ID: 1OB2). The unnumbered Phe is the 3′-aminoacylated residue. EF-Tu numbering differs because 1OB2 does not include the initiating Met residue.

The CDI toxin deployed by E. coli strain EC869 also cleaves tRNA in an EF-Tu-dependent manner (22), but CdiA-CTNC101 and CdiA-CTEC869 differ in several respects. First, the nucleases share no sequence similarity, though given the lack of homology between known BECR toxins and CdiA-CTNC101, the CdiA-CTEC869 domain could still adopt the same fold. Second, the two toxins have distinct substrate specificities, with CdiA-CTEC869 cleaving tRNAGln and tRNAAsn at a different position within the acceptor-stem duplex. Third, CdiA-CTEC869 binds EF-Tu with high affinity even in the absence of its cognate immunity protein (22). Finally, the two CDI toxins appear to have different requirements for EF-Ts. EF-Ts stimulates CdiA-CTEC869 activity, but is not absolutely required for tRNA cleavage (22). By contrast, CdiA-CTNC101 activity is not detected in vitro unless both EF-Ts and EF-Tu are present. Despite this discrepancy, we propose that EF-Ts stimulates the activity of each toxin by increasing steady-state GTP·EF-Tu·aa-tRNA substrate levels as described by Blanchard and colleagues (68,69). Though EF-Tu has nanomolar affinity for aa-tRNA at 0–4°C (70), these interactions are estimated to be 20- to 30-fold weaker at 37°C (71,72), indicating that the dissociation constants approach the micromolar range under the conditions used here to monitor in vitro tRNase activity. Coupled with the low affinity of CdiA-CTNC101 for EF-Tu, in vitro nuclease activity is presumably impeded by the low probability that EF-Tu, aa-tRNA and toxin all converge simultaneously. By contrast, the dynamic nature of EF-Tu·aa-tRNA binding has less of an effect on CdiA-CTEC869 tRNase activity, because this toxin forms a stable, high-affinity binary complex with EF-Tu (22). We note that CdiA-CTNC101 remains a potent nuclease in vivo, because cytoplasmic EF-Tu is present at sufficiently high concentrations (50–60 μM) (73) to drive toxin-binding through mass action. Thus, although relatively few toxin domains are delivered into target cells during CDI (74), the natural abundance of substrate presumably relieves the selective pressure for CdiA-CTNC101 to evolve high-affinity interactions with EF-Tu.

The CdiA-CTNC101 nuclease domain shares the BECR fold and most closely resembles BrnT and the C-terminal tRNase domain of colicin D. Surprisingly, recent computational surveys of prokaryotic toxin domains did not uncover CdiA-CTNC101 (nor any of its homologs) as members of the BECR superfamily (23,75). Structural homology is often presumed to indicate a common evolutionary history; but the colicin D immunity protein does not share any sequence or structural similarity with CdiINC101, suggesting that the corresponding nuclease domains are related through convergent evolution. Further, the structural similarities with RelE and YoeB are intriguing, because these toxins also require a macromolecular scaffold for substrate recognition, cleaving mRNA only when presented in the ribosome A site (76,77). The advantage of scaffold-dependent activity is not immediately apparent. Scaffolds are not necessary for these types of reactions, because several type II toxins degrade mRNA in a ribosome-independent manner (78,79), and other CDI toxins are known to cleave tRNA acceptor stems in the absence of EF-Tu (16,28). Perhaps scaffold-dependent activity is merely an inevitable consequence of toxin evolution in the densely packed bacterial cytoplasm, where EF-Tu and ribosomes are abundant. However, scaffolds provide additional substrate specificity, and therefore should restrict the activity of potential lethal nucleases. This property may be advantageous for type II toxin-antitoxin systems, which are hypothesized to mediate reversible growth stasis and persistence (80,81). Such safe-guards could also protect CDI+ bacteria in the event that their immunity protein levels decrease transiently. Alternatively, it is possible that interactions between CDI toxins and EF-Tu serve other physiological functions. In this model, neutralized EF-Tu·CdiA-CT·CdiI complexes could influence gene expression, and thereby mediate cell-to-cell signaling between siblings as proposed recently for Burkholderia thailandensis (82). We note that CDI toxins from E. coli 96.154 and Klebsiella pneumoniae 342 also interact with EF-Tu, suggesting that this phenomenon may be quite common. It is unclear whether these sequence-diverse toxin domains bind EF-Tu in the same manner, and the consequences of this widespread phenomenon on bacterial physiology remain to be explored.

ACCESSION NUMBERS

PDB IDs: 5I4Q and 5I4R.

Click here for additional data file.