Structure–function analyses reveal the molecular architecture and neutralization mechanism of a bacterial HEPN–MNT toxin–antitoxin system

Abstract

Toxin–antitoxin (TA) loci in bacteria are small genetic modules that regulate various cellular activities, including cell growth and death. The two-gene module encoding a HEPN (higher eukaryotes and prokaryotes nucleotide-binding) domain and a cognate MNT (minimal nucleotidyltransferase) domain have been predicted to represent a novel type II TA system prevalent in archaea and bacteria. However, the neutralization mechanism and cellular targets of the TA family remain unclear. The toxin SO_3166 having a HEPN domain and its cognate antitoxin SO_3165 with an MNT domain constitute a typical type II TA system that regulates cell motility and confers plasmid stability in the bacterium Shewanella oneidensis. Here, we report the crystal structure and solution conformation of the SO_3166–SO_3165 pair, representing the first complex structures in this TA family. The structures revealed that SO_3165 and SO_3166 form a tight heterooctamer (at a 2:6 ratio), an organization that is very rare in other TA systems. We also observed that SO_3166 dimerization enables the formation of a deep cleft at the HEPN-domain interface harboring a composite RX4–6H active site that functions as an RNA-cleaving RNase. SO_3165 bound SO_3166 mainly through its two α-helices (α2 and α4), functioning as molecular recognition elements. Moreover, their insertion into the SO_3166 cleft sterically blocked the RX4–6H site or narrowed the cleft to inhibit RNA substrate binding. Structure-based mutagenesis confirmed the important roles of these α-helices in SO_3166 binding and inhibition. Our structure–function analysis provides first insights into the neutralization mechanism of the HEPN–MNT TA family.

Introduction

Toxin–antitoxin (TA)⁴ loci are small genetic modules that are widespread in bacterial plasmids and chromosomes and target various cellular functions to regulate cell growth and death (1, 2). Six different types of TA systems (types I–VI) have been characterized so far based on the interaction mode of the TA and the molecular nature of the antitoxin (3). In the most well-characterized and abundant type II TA system, an antitoxin can bind directly to a cognate toxin to form a tight protein–protein complex for inactivating toxicity during normal growth. When the expression from TA loci is impaired by stresses, specific ATP-dependent proteases (such as Lon and ClpXP) will be activated to destroy the labile antitoxin by proteolytic cleavage (4, 5). The subsequently released toxin is activated to regulate cell mortality by blocking DNA replication or translation, or by facilitating mRNA degradation. These key cellular processes are associated with many roles in cell physiology, such as biofilm formation and virulence in bacteria (1, 2, 6, 7).

The HEPN (higher eukaryotes and prokaryotes nucleotide-binding) superfamily members adopt an all-helical fold and are distributed in the proteins associated with bacterial drug resistance and human neurodegeneration (8). The MNT (minimal nucleotidyltransferase) domains have been identified as the minimal units of DNA polymerase protein superfamily that are responsible for transferring nucleic acids to an acceptor hydroxyl group in prokaryotes and animal Sacsin proteins (9). The two-gene modules encoding a MNT domain and an accompanying HEPN domain have been proposed to represent a novel, nonconventional type II TA system widely distributed in both archaea and bacteria (10). The MNT- and an accompanying HEPN-containing protein have been predicted as the toxin and the cognate antitoxin, respectively. However, a genome-wide screen for toxins has identified the HEPN domain function as the toxin in the HEPN–MNT module belonging to a TA system (11). A subsequent bioinformatics analysis further predicted that the HEPN-domain functions as toxins that are essential components of numerous TA and abortive infection systems in prokaryotes, and are also tightly associated with many restriction–modification (R–M) and CRISPR–Cas systems (12). Moreover, most HEPN domains contain a conserved RX4–6H motif (where X is any amino acid and the residue immediately after the conserved R is typically a polar amino acid, and 4–6 indicates the number of amino acids between R and H). As a conserved feature of the HEPN domain, this motif may function as a novel RNase active site, but it is usually lost when fused with an MNT domain. Until now, the neutralization mechanism and cellular targets of HEPN–MNT TA family members remain unknown.

Recently, we experimentally characterized SO_3166 and SO_3165 from Shewanella oneidensis as a novel type II TA pair with a critical role in regulating cell motility and conferring plasmid stability (13). SO_3166 is composed of a single HEPN domain and a conserved RX4–6H catalytic motif. It can function as a toxin with strong inhibition on cell growth in S. oneidensis and Escherichia coli. The toxicity can be neutralized by its cognate antitoxin SO_3165 with a MNT domain, which binds to the promoter of the TA operon and repressed its activity (13). However, the inhibition mechanism of SO_3166 toxicity by SO_3165 remains unclear. To this end, we performed structure–function studies on the SO_3166–SO_3165 system and demonstrate how the HEPN toxin (SO_3166) is recognized and inhibited by the MNT antitoxin (SO_3165). The unique structure and the following mutagenesis study revealed one SO_3165 can recognize and inhibit the toxicity of three SO_3166 by sterically blocking its RX4–6H catalytic domain. These findings may provide novel insights into the neutralization mechanism of HEPN–MNT TA family members, and are useful to further understand the function of HEPNs and MNTs in prokaryotes.

Results

HEPN toxin SO_3166 is a RNase that can cleave mRNA

To evaluate whether toxin SO_3166 can function as a RNase similar to many type II toxins, its RNA cleavage activity against different types of RNAs (mRNA, tRNA, and rRNA) was studied. The abundant ompA mRNA in E. coli was synthesized and used for the in vitro RNA cleavage assay. The ompA mRNA (306 nt, 0.5 μg) substrate can be cleaved into smaller fragments by purified SO_3166 (Fig. S1) after 20 min (∼16.6 μm, Fig. 1A). However, its cleavage capacity is completely inhibited by SO_3165 in this complex, consistent with the previous observation that SO_3166 toxicity could be neutralized by SO_3165 (13). Meanwhile, no RNA cleavage fragments were detected against E. coli total tRNAs (Fig. 1B), or S. oneidensis rRNA (pHGE-SO_3166 overexpressing in S. oneidensis, Fig. 1C). DNA cleavage activities of SO_3166 against different types of DNAs (circular dsDNA, linear dsDNA, and circular ssDNA) were also tested, but no activity was detected in the in vitro DNA cleavage assay (Fig. S2). These in vitro and in vivo results suggest that SO_3166 can function as a RNase by cleaving mRNA rather than tRNA or rRNA.

Figure 1.

RNA cleavage activity of HEPN toxin SO_3166. A, the ompA (1–306 nt) mRNA cleavage activity of SO_3166 and the SO_3166–SO_3165 complex at different time points (in vitro). B, E. coli total tRNAs treated with SO_3166 at different time points (in vitro). C, total RNAs isolated from MR-1/pHGE and MR-1/pHGE-SO_3166 after a 40-min induction with 0.5 mm IPTG added at a turbidity of 0.7 at 600 nm (in vivo). HI indicates heat-inactivated SO_3166 or SO_3166–SO_3165 complex. M indicates ssRNA ladder in A and B, and DNA ladder in C.

Antitoxin SO_3165 and toxin SO_3166 can form a heterooctamer structure

The crystal structure of the SO_3166–SO_3165 complex was solved by the single-wavelength anomalous dispersion method from synchrotron data using selenomethionine (Se-Met)-labeled protein in space group P2₁2₁2 and was refined to a final R/R_free factor of 0.26/0.29 at 3.0-Å resolution (Table 1). There are four molecules composed of one SO_3165 binding three SO_3166 in an asymmetric unit (ASU) to form an heterotetramer (Fig. 2A), with overall dimensions of ∼57 × 223 × 54 Å. The residues Met-1–Asn-5 and Asn-128–Ser-139 in SO_3165 (139 residues at full-length), were not observed in the electron density map and not included in the current model, whereas all the residues in the three SO_3166 molecules (133 residues at full-length) could be built into the model (except for the absence of selenomethionine-substituted-1 (Se-Met-1) in SO_3166_C). The three SO_3166 molecules are highly similar with a root mean square deviation (r.m.s. deviation) of 0.70–0.72 Å for 131 aligned Cα atoms. The notable differences among them are the variable conformations of the α1–α2 and α4–α5 loops (harboring part of the catalytic RX4–6H motif, discussed below) (Fig. S3). Meanwhile, a tight heterooctamer can be generated by crystal packing with symmetry-related molecules with the organization (SO_3166)₂–SO_3165-(SO_3166)₂–SO_3165-(SO_3166)₂ (Fig. 2B).

Table 1.

Data collection and refinement statistics

Data collection
Wavelength (Å)	0.9788
Space group	P21212
Unit-cell parameters	a = 56.6 Å, b = 224.4 Å, c = 53.3 Å, α = β = γ = 90°
Resolution (Å)	3.00 (3.08–3.00)a
Number of unique reflections	26,218 (1,904)
Completeness (%)	99.9 (100)
Redundancy	7.6 (7.8)
Mean I/σ(I)	13.93 (3.51)
Molecules in asymmetric unit	4
Rmerge (%)	6.5 (64.8)
CC1/2	99.9 (94.9)
Structure refinement
Reflections used in refinement	14,273
Resolution range (Å)	48.15–3.00
Rwork/Rfree (%)	26.4/29.2
Macromolecules	4,145
Protein residues	517
Waters	0
Average B factor ( Å2)
Main chain	72.95
Side chain	77.56
Clash score	8.06
Ramachandran plot (%)
Most favored	95.6
Allowed	3.8
Disallowed	0.6
R.m.s. deviations
Bond lengths (Å)	0.003
Bond angles (°)	0.55

^a The values in parentheses indicate the highest resolution shell.

Figure 2.

Overview of HEPN-MNT SO_3166–SO_3165 complex structure. A, overall crystal structure of the SO_3166–SO_3165 complex shown as a schematic. The heterotetramer structure is composed of one SO_3165 (in green and labeled using the subscript A) binding SO_3166 molecules (in cyan/pink/yellow and labeled using the subscripts A–C, respectively) simultaneously. B, heterooctamer structure of the SO_3166–SO_3165 complex generated by crystallographic symmetry. It is composed of two SO_3165 (in green/dark green and labeled using the subscripts A and A′, respectively) binding six SO_3166 molecules (in cyan/yellow/pink/blue/magenta/wheat and labeled using subscripts A–C, A′–C′, respectively). C, purified SO_3165 (blue) and SO_3166–SO_3165 complex (red) eluted from gel filtration chromatogram (Superdex^TM 200 10/300 GL) at 16.3 and 12.9 ml, respectively. D, solution conformation of the SO_3166–SO_3165 complex by SAXS analysis. Left: Curve 1, experimental data; Curve 2, scattering patterns computed from the DAMMIN model; Curve 3, scattering patterns computed from the GASBOR model; insertion in right above, P(r) function. Right, DAMMIN models overlap with the tight heterooctamer crystal structure. The experimental data compare well with the theoretical curves of the heterooctamer crystal structure of the complex.

The SO_3166–SO_3165 complex migrated on size exclusion chromatography with a molecular mass of ∼140 kDa compared with its calculated heterotetramer molecular mass of ∼60 kDa (Figs. 2C and Fig. S5). The oligomeric state of the complex in solution was further studied by small-angle X-ray scattering (SAXS). General structural parameters calculated from the SAXS profile using Guinier approximation and indirect Fourier transformation (program GNOM) are shown in Table S3. Experimental SAXS curves after standard preliminary processing are presented in Fig. 2D (curve 1). Independent confirmation of the results above is the ab initio reconstruction of the complex shape in solution by the programs DAMMIN (χ² = 1.272) and GASBOR (χ² = 1.446) (Table S3). The results show that the experimental SAXS curves are in agreement with the heterooctamer theoretical curve (curves 2 and 3), and the available heterooctamer crystal structure can be fit into the respective SAXS-derived low resolution envelopes. Moreover, comparison of the theoretical scattering patterns from the heterotetramer crystal structure in the ASU with the experimental SAXS profile showed that the model curve differs considerably, whereas the crystallographic heterooctamer yields good fit to it (χ = 2.6) (Fig. S4). The results suggest the active form of the complex is a heterooctamer that is composed of six SO_3166 molecules binding two SO_3165 molecules in solution.

Such antitoxin to toxin oligomerization was mediated entirely by toxin molecules, and the unique organization (toxin–antitoxin = 6:2) has not been previously reported in a TA system. Moreover, further analysis of the recombinant complex at different concentrations (the same as those used in the SAXS experiment) by size exclusion chromatography showed they have almost identical retention time (Fig. S5). Therefore, we can conclude that the SO_3166–SO_3165 complex always exists as a heterooctamer and this form is not concentration dependent in vitro. Considering that the SO_3166–SO_3165 concentration used in the in vitro studies (at the micromolar level) is considerably higher than that in cells, additional studies may be required to confirm the unusual stoichiometry of the complex in vivo.

Toxin SO_3166 possesses a typical HEPN domain fused with an RX4–6H motif

SO_3166 is composed of a five-helix bundle, as found in HEPN superfamily members that all adopt the all α-helical-fold (12). Notably, the fused RX4–6H motif (the residues Arg-97–His-102) is located at the end of the α4 helix and in the beginning of the α4–α5 loop (Fig. 3, A and B). The conserved fused RX4–6H motif has been suggested as the primary determinant of a novel RNase active site in the HEPN superfamily (13). This motif is exposed to the solvent (before the antitoxin binding) and may be used to accommodate substrate RNA to trigger the catalysis of RNA cleavage. Moreover, the conformation of the region containing this motif is flexible (Fig. S3), which may be required for substrate-binding or effective catalysis.

Figure 3.

Structural characteristics of HEPN toxin SO_3166. A, structure-based sequence alignment of SO_3166 with its representative homologs performed using Clustal X (version 1.81) and ESPript 3. They include SO_3166 from S. oneidensis and the homologs from Pectobacterium carotovorum (P. carotovorum), Oleibacter marinus (O. marinus), Marinomonas aquimarina (M. aquimarina), Clostridium butyricum (C. butyricum), and Desulfotomaculum gibsoniae (D. gibsoniae). The residues that were previously identified important for SO_3166 toxicity are highlighted using black arrows. The conserved RX4–6H motif are labeled using a red box. B, homodimer structure of SO_3166 shown as a schematic from the side view. The two subunits SO_3166_A and SO_3166_B are shown in cyan and pink, respectively. The RX4–6H motifs from both subunits are highlighted in red. C, the molecular surface representation of the SO_3166 dimer from the same view as B (blue, +7.1KT; red, −7.1KT), colored by its local electrostatic potential. The deep cleft formed by HEPN-domain dimerization of both subunits is highlighted using an ellipse. D, surface representation of the SO_3166 dimer from the top view. The key residues (except Cys-15 and Leu-107) for SO_3166 toxicity are highlighted in red. E, structural superimposition of SO_3166_A (dark gray) and the homolog HI0074 (light gray, PDB ID 1JOG). The regions containing RX4–6H motif in SO_3166 (Val⁹⁴–Asp¹⁰³) and HI0074 (Asp¹⁰⁴–Tyr¹¹³) are highlighted in cyan and orange, respectively. The conformations of the regions (especially the catalytic histidine) are remarkably different.

The DALI search (http://ekhidna.biocenter.helsinki.fi/dali_server)⁵ revealed that SO_3166 has remarkable similarities with several uncharacterized proteins with a single HEPN-fold composed of five helices, as well as other proteins containing HEPN domains, such as the aminoglycoside NTs. One of the closest structural homologs is HI0074 (PDB ID 1JOG), with a typical HEPN-fold, from Haemophilus influenza, with a DALI Z-score of 12.0 and a r.m.s. deviation of 2.8 Å for 119 Cα atoms (Fig. 3E). Similarly, HI0074 with a single HEPN-domain is also a triangular homodimer harboring an RX4–6H motif (PDB ID 1JOG) (Fig. S6). However, the conformations of the regions containing the RX4–6H motif are remarkably different in the HI0074 and SO_3166 structures (especially the catalytic histidine), indicating they have different substrate-binding specificities (Fig. 3E). Moreover, the large positively charged cavity at the dimer interface formed by two HEPN domains has also been observed in Sacsin (PDB ID 3O10) and kanamycin nucleotidyltransferase (KNTase, PDB ID 1KNY) for the binding of GTP and other nucleotides (14, 15) (Fig. S6). All these structures suggest that dimerization of the HEPN-domain is conserved in its evolution.

Dimerization of toxin SO_3166 may be associated with its catalytic activity

Two SO_3166 protomers (SO_3166_A–SO_3166_B) in the ASU assemble as a tight homodimer and the dimerization occurs mainly through the interactions of the α2 helix in both monomers (Fig. 3B). These extensive direct interactions are composed of dozens of hydrogen bonds (H-bonds) and salt bridges between the two subunits (Fig. S6). Another SO_3166 dimer (SO_3166_C–SO_3166_C′) can be generated by crystallographic symmetry (Fig. 2B). Structural comparisons of these two dimers showed that their overall conformations are highly similar, with a r.m.s. deviation of 1.3 Å for 252 Cα atoms (Fig. S7, A and B). Moreover, they have similar buried dimer surface areas at the dimer interface (with the ration of buried surface area/total surface area of 861 Ų/7,628 Ų and 1,062 Ų/7,701 Ų, respectively), as well as very similar interacting residues (Fig. S7, C and D). Therefore, the SO_3166_A–SO_3166_B dimer (also SO_3166_B and SO_3165_A in the ASU) was used for detailed analysis below, unless otherwise stated.

The contacting residues at the dimer interface are highly conserved across the HEPN toxin family (Fig. 3A), supporting the importance of the dimerization of SO_3166 from an evolutionary point of view. Dimerization of SO_3166 enables the formation of a deep cleft mainly composed of the α2, α4, and α3 helices between the two subunits (Fig. 3, B and C). The catalytic RX4–6H domain is located in the base of the cleft, which is formed by the interlocking of the α4 helix and the α4–α5 loop of the two HEPN domain protomers. The cleft is ∼32 Å (long) × 18 Å (deep) × 13 Å (wide at its narrowest point) and the constituent residues are conserved. The width of the cleft may allow the favorable binding of a single-stranded RNA molecule (ssRNA, ∼11 Å width), whereas sterically excluding double-stranded DNA (dsDNA) (Fig. S8), which are consistent with the RNA/DNA cleavage experiment results above (Fig. 1 and Fig. S2). Moreover, inspection of the surface charge distribution of the SO_3166 dimer revealed there are notable positively charged protuberances with large continuous areas in and around the cleft (Fig. 3C). These patches with continuously positive surfaces may function as the wrapping path for the substrate binding. Therefore, the cleft in the SO_3166 dimer may provide a platform for the catalytic process as an active center. Our recent site-directed mutagenesis of SO_3166 revealed that three residues (Arg-97, Asn-98, and His-102) belonging to the RX4–6H motif, as well as six additional residues (Cys-15, His-56, Arg-70, Val-94, Leu-107, and His-118), are critical for the toxicity of SO_3166 (13). Mapping of these residues on SO_3166 showed they are located in the cleft (except Cys-15 and His-118) (Fig. 3D), indicating these residues are closely associated with SO_3166 RNase activity.

Antitoxin SO_3165 possesses an extra helix (α4) compared with its structural homologs

SO_3165 adopts a mixed α/β-fold, composed of a four-stranded twisted β-sheet (β1–β4) in the middle flanked by four helices (α1–α4) at the sides (Figs. 2A and 4A). The overall structure is well-ordered in the complex. Sucharge analysis of SO_3165 showed there is a positively charged protuberance distributed in helix α1 and around (Fig. 4B). Meanwhile, helices α2 and α4 are covered with dominantly negative charges (Fig. 4B), which mediated SO_3166 binding (discussed below). The solution behavior of purified apo SO_3165 using size exclusion chromatography indicated that the corresponding molecular mass (∼34 kDa) is larger than that of a monomer (∼15.5 kDa) (Fig. 2C and Fig. S5), suggesting a dimer may be formed in the absence of toxin SO_3166. Considering that SO_3165 exists as a monomer in the complex structure, it may be depolymerized to accommodate toxin binding in the inhibition process.

Figure 4.

Structural characteristics of MNT antitoxin SO_3165. A, structure-based sequence alignment of SO_3165 with its representative homologs as described in the legend to Fig. 3A. They include SO_3165 from S. oneidensis (S. oneidensis) and the homologs from P. carotovorum (P. carotovorum), O. marinus (O. marinus), P. syringae (P. syringae), P. caricapapayae (P. caricapapayae), and Halomonas ilicicola (H. ilicicola). The conserved residues are boxed in blue, identical conserved and low conserved residues are highlighted in red background and red letters, respectively. B, the molecular surface representation of SO_3165 (blue, +7.4KT; red, −7.4KT), colored by its local electrostatic potential. The helix α1 is covered with dominantly positive charges, whereas helices α2- and α4-mediated SO_3166 binding are covered with dominantly negative charges. C, structural superimposition of SO_3165 (light gray) and the homolog HI0073 (heavy gray, PDB ID 1NO5). The β1–β2 loop in HI0073 and the corresponding region in SO_3165 (also the helix α4) are highlighted in orange and green, respectively. They have significant differences in a β1–β2 loop conformation and the extra helix α4 in SO_3165, although they share overall similar folds. The binding zinc ion in HI0073 is shown as a yellow sphere.

A DALI search for globally similar proteins revealed that SO_3165 has remarkable similarities with several nucleotidyl transferases (NTs) from archaea and bacteria, including MNTs with a single MNT domain and aminoglycoside NTs with a fused MNT–HEPN domain. The closest structural homolog is HI0073 (PDB ID 1NO5) with a typical MNT-fold from H. influenza, with a DALI Z-score of 8.1 and a r.m.s. deviation of 2.4 Å for 100 Cα atoms (Fig. 4C). One important difference is the extra α4 helix in SO_3165, which play an important role in SO_3166 binding and inhibition (discussed below). The other difference is the long β1–β2 loop (interrupted by a small helix) in HI0073, which is absent in SO_3165. Meanwhile, no zinc ions were observed in the SO_3165 structure as found in HI0073 (16). The zinc-binding region and surroundings (Asp-46, Asp-48, and Asp-79) is suggested to be of functional importance and corresponds to the magnesium sites of the nucleotide-binding domains in the previously known structures of NTs (16). The structural variations in MNT domains probably reflect their function in the inhibition of distinct toxins.

The helices α2 and α4 of SO_3165 play a major role in SO_3166 binding

Structural analysis showed that SO_3165 has direct interactions with three SO_3166 molecules simultaneously. The buried surface areas of the three subunits (labeled using the subscripts A–C) at the interface are 565, 786, and 484 Ų, which is up to 7.1, 9.9, and 6.1% of the total surface area (7,971 Ų), respectively. These extensive contacts include 19 H-bonds and salt bridges (Fig. 5), as well as ∼100 van der Waals (VDW) contacts (not shown).

Figure 5.

Contacts analysis (H-bonds and salt bridges) between SO_3165 and SO_3166. A, direct interactions between SO_3166_A and SO3165_A. B, direct interactions between SO_3166_B and SO3165_A. C, direct interactions between SO_3166_C and SO3165_A. SO_3166 and SO_3165 are shown in dark gray and light gray, respectively. The interacting residues are shown as sticks and their colorings are the same as described in the legend to Fig. 2A. The RX4–6H motif is highlighted in red. The residues that are previously identified important for SO_3166 toxicity are highlighted using ellipses.

Notably, the two helices (α2 and α4) from SO_3165 are oriented toward the RX4–6H domains of the three SO_3166 molecules. These helices contribute to the majority of the hydrogen-bonding network and function as the main molecular recognition elements, thereby stabilizing the heterooctamer. The long α4 helix binds to the two subunits SO_3166_A and SO_3166_B simultaneously, and insert into the deep cleft of the SO_3166 dimer. The corresponding interacting elements in SO_3166 include the α2 helix (Glu-48, Asp-52, Asn-55, His-56, and Arg-59) in SO_3166_A, as well as the α2 helix (Asn-55 and Arg-59), the α2–α3 loop (Pro-66, Gln-67, and Ser-69), and the α4 helix (Arg-97) in SO3166_B, through extensive direct interactions (Fig. 5, A and B). The α2 helix of SO_3165 also falls close to the RX4–6H domain in one subunit of the symmetry-related SO_3166 dimer. The corresponding interacting elements in SO_3166 include the α4 helix (Lys-91, Lys-92, and Asn-98), the α2–α3 loop (Arg-70) and the α4–α5 loop (Leu-107) through extensive contacts (Fig. 5C). Moreover, the surfaces of the two helices are predominantly distributed with negative charges (Fig. 4B), which complements the positively charged regions in and around the cleft (Fig. 3C).

SO_3166 toxicity inhibition by SO_3165 in the active octameric form

Further structural analyses showed Rx4–6H motif residues Arg-97, Asn-98, and His-102, as well as Arg-70 and Val-94, are located in the bottom of the cleft and extend to a continuous large area (Fig. 6, A–D). More importantly, these residues, (such as Arg-70 and Arg-97) directly interacting with helices α2 and α4 of SO_3165, are key residues for SO_3166 toxicity. For example, the side chain (NH1 and NH2) groups of Arg-97 of SO_3166_B can form a salt bridge with the side chain carboxyl group (OE2) of Glu-117 (3.0 Å). The side chain (NH2) of Arg-70 in SO_3166_C can form two salt bridges with the side chain (OD1 and OD2) of Glu-117 (3.7 and 3.2 Å, respectively).

Figure 6.

Structural basis of SO_3166 toxicity suppression by SO_3165. A and B, heterooctamer structure of the SO_3166–SO_3165 complex from side view (A) and top view (B), respectively. SO_3166 and SO_3165 are shown as surface and schematics, respectively, and their colorings are the same as described in the legend to Fig. 2B. C and D, close-up views show the contacts between SO_3166 and SO_3165. The residues important for SO_3166 toxicity are highlighted in red. The helix α4 of SO_3165 can bind into the cleft formed by the SO_3166 dimer (C) and sterically block the two active sites of SO_3166 simultaneously. The helix α2 and α2–α3 loops of the symmetry-related SO_3165 dimer can bind to the edge of the cleft (D), and the narrowed cleft will probably cause the unfavorable binding of the substrate by SO_3166.

Consequently, the α4 helix can insert into the cleft formed by the SO_3166 dimer and may mask the RX4–6H catalytic domain and sterically block access of the substrate (Fig. 6, A and C). On the other side, the α2 helix (as well as the α2–α3 loop) from two SO_3165 molecules can also narrow the cleft formed by the other SO_3166 dimer formed by two symmetry-related molecules (Fig. 6, B and D). This helix may also affect the substrate binding access to the RX4–6H motif. Therefore, under the active octameric status, the complete neutralization of SO_3166 toxicities may require the cooperation of two SO_3165 molecules by their two helices simultaneously.

Functional studies confirm the important roles of the SO_3165 helices α2 and α4 in SO_3166 toxicity inhibition

Structure-based truncations of the helices α2 and α4 of SO_3165 were performed to confirm their roles in SO_3166 toxicity inhibition. The truncations including Δα4 (Gln-98–His-113), Δα4 (Leu-114–Val-125), and Δα2 (Asn-52–Ala-65) (Table S1), were constructed to test their neutralization capacity on SO_3166 toxicity by cell toxicity assays (Fig. 7A). All these variants can be produced in E. coli and have similar secondary structures to the WT (Fig. S9).

Figure 7.

The roles of helices α2 and α4 of SO_3165 in the inhibition of SO_3166 toxicity by functional studies. A, three truncations: Δ(98–113) and Δ(114–125) in the helix α4 and Δ(52–65) in the helix α2 of SO_3165 are constructed for SO_3166 toxicity inhibition studies. The regions important for SO_3166 toxicity are highlighted in red as described in the legend to Fig. 6. B, the viability (CFUs/ml) of BL21 hosts carrying the pET28a-based plasmids were induced with 0.3 mm IPTG added at A₆₀₀ ∼ 0.1. Three independent cultures were conducted, error bars indicate mean ± S.E. (n = 3). C, viabilities were tested after induced for 4 h. Three independent cultures of each strain were tested and only representative images are shown.

The results showed the inhibitory effect on SO_3166 toxicity is significantly affected by the Δα4 (Gln-98–His-113) truncation compared with the WT SO_3165, whereas Δα4 (Leu-114–Val-125) could still repress the toxicity to allow normal cell growth (Fig. 7, B and C). Deletion of the α2 helix also had a notable effect on SO_3166 toxicity inhibition and caused cell growth retardation and low viability. Therefore, these cell toxicity studies revealed the α2 and α4 helices play an important role in SO_3166 toxicity inhibition. The results are in a good agreement with the structure above, where the regions responsible for SO_3166 toxicity (the RX4–6H motif and surroundings) are dominantly masked by α4 (Gln-98–His-113) and α2, whereas Δα4 (Leu-114–Val-125) is not involved in this blockade.

Discussion

Although the HEPN–MNT module has been previously predicted as a novel type II TA system, the neutralization mechanism and cellular targets of this family remain unclear. Site-directed mutagenesis studies have shown that the conserved RX4–6H motif is responsible for SO_3166 toxicity, which is probably associated with its RNase activity (13). In this study, we first identified that HEPN toxin SO_3166 can specifically cleave mRNA, rather than tRNA, rRNA, or DNA. The structure of SO_3166 revealed that dimerization of the HEPN domains brings the RX4–6H motifs of the two HEPN domain protomers into close proximity, generating a composite RNase active site in the cleft. The situation is similar to that observed in type III-A CRISPR-associated protein Csm6, composed of the C-terminal HEPN domain and N-terminal CRISPR-associated Rossman-fold domain (17). HEPN-domain dimerization generates an ssRNA-binding cleft located at the interface, which harbors the RNase active center of Csm6. Therefore, we propose that dimerization of the HEPN-domains and the resulting juxtaposition of the RX4–6H motifs leads to the formation of a composite symmetric active center in SO_3166 that binds the substrate RNA asymmetrically.

The conserved RX4–6H motif has emerged as the most strongly conserved feature of the HEPN domain. Our recent site-directed mutagenesis on SO_3166 revealed three residues (Arg-97, Asn-98, and His-102) within the motif are critical for the toxicity of SO_3166 (13). Site-directed mutagenesis of the KEN domain of RNase L and the RNase domains of RloC and PrrC have shown that the arginine and histidine corresponding to the conserved R and H in the RX4–6H motif are essential for their respective nuclease activities (18–20). In our complex structure, the two RX4–6H catalytic motifs located in the cleft are sterically blocked by two helices (α2 and α4) of SO_3165. Structure-based mutagenesis further confirmed these helices play an important role in SO_3166-binding and toxicity inhibition. More importantly, the RX4–6H motif is usually lost in HEPN–MNT-fused aminoglycoside NTases (such as PDB IDs 1KNY, 1L8T, and 4CS6). Although the structures of the HEPN and MNT domains in SO_3166 and SO_3165 show remarkable similarities with these NTases, SO_3166–SO_3165 does not cause resistance to aminoglycoside antibiotics. Therefore, the presence of the RX4–6H motif may indicate that SO_3166–SO_3165 functions as a TA system with RNase activity in a HEPN toxin, rather than as an aminoglycoside NTase for antibiotic resistance via the nucleotidylation of antibiotic molecules.

Our structure-functional studies on SO_3166–SO_3165 showed the antitoxin SO_3165 may efficiently inhibit SO_3166 toxicity using its two helices, especially the C-terminal α4, which mediate the formation of the octameric complex at a ratio of 2:6. The α4 helix can directly insert into the cleft, and may block the composite RNase active site in the catalytic RX4–6H motifs of the two HEPN domain protomers. The α2 helix from two symmetry-related SO_3165 molecules can bind to a region close to the RX4–6H motifs on the edge of the cleft from opposite directions to reduce the cleft size. Therefore, the α2 helix may also affect the substrate RNA binding. Our functional studies on truncations of the two helices confirmed the strategies in SO_3166 toxicity inhibition under the active octameric status. Meanwhile, the N-terminal region distributed with dominantly positive charges (Fig. 4B) may be associated with the promoter-binding of the TA operon to repress its activity (13). The relative contributions of neutralization and transcriptional regulation to the inhibition of the toxic phenotype under different conditions warrant further investigations.

The closest structural homologue of SO_3165–SO_3166 is HI0073–HI0074 from H. influenza, which harbors a typical MNT-fold and HEPN-fold, respectively (Figs. 3E and 4C). Therefore, the HI0073–HI0074 pair may constitute a type II TA system like SO_3165–SO_3166. The previous study showed HI0073–HI0074 can form a tetramer with a 2:2 molecular ratio determined by size exclusion chromatography (21). Structural comparison of SO_3165 with HI0073 revealed the presence of the extra helix α4 in SO_3165 that can bind two SO_3166 simultaneously. Unlike the dual role of helices α4 and α2 in SO_3166 binding, only the helix α2 of HI0073 is probably responsible for HI0074 toxicity inhibition. Therefore, variable neutralization mechanisms of the HEPN–MNT TA family may exist depending on the presence or absence of the C-terminal α4 in MNT antitoxin.

Taken together, our structure–function studies on the SO_3166–SO_3165 HEPN–MNT pair present for the first time the structure of the complex with a unique organization, and reveal that one SO_3165 molecule can recognize and inhibit the toxicity of three SO_3166 molecules by sterically blocking the RX4–6H catalytic domains simultaneously. This work sheds light on the molecular architecture of the HEPN–MNT TA complex and the functional organization of its constituent domains.

Experimental procedures

Construction of plasmids and bacterial strains

Bacterial strains and plasmids used in this study are listed in Table S1. The pCA24N-SO_3166 and pHGE-SO_3166 plasmids were transformed into E. coli BW25113 and S. oneidensis MR-1 for SO_3166 production for RNA cleavage assay, respectively (13). The recombinant plasmids pET28b-SO_3165–SO_3166 and pET28b-SO_3165 were transformed into BL21(DE3) cells, respectively, for the expression and purification used in crystallization, small-angle X-ray scattering, and CD. Truncations of SO_3165 on the plasmid pET28a-SO_3165–SO_3166 (or pET28a-SO_3165) were performed by one-step PCR according to the QuikChange site-directed mutagenesis strategy (Stratagene). The primers are listed in Table S2.

SO_3166 RNA cleavage assay

To conduct mRNA cleavage assay, ompA mRNA was used as substrate for in vitro assay. For the synthesis of ompA mRNAs (1–306 nt from ompA), PCR products were obtained using the primers in Table S2. The PCR products were used as template for in vitro transcription with T7 RNA polymerase (New England Biolabs). The T7 RNA polymerase promoter sequence was included in the forward primers of ompA-T7-F. PCR products were gel-purified, and 0.5 to 1 μg of the PCR product was used as the template for the in vitro RNA reaction following the instructions. SO_3166 and the SO_3166–SO_3165 complex were purified as previously described (13). The reaction mixture for the SO_3166 RNase cleavage assay (8 μl) contained 0.5 μg of RNA, 20 mm Tris-HCl (pH 8.0), 300 mm NaCl, 25 mm EDTA, 60 mm KCl, 30 mm MgCl₂, and 2 μg of purified SO_3166 protein or SO_3166–SO_3165 complex. For the tRNA cleavage assay, E. coli total tRNAs (Roche Applied Science) were used as substrate. The reaction mixture was incubated at 37 °C for 20 or 40 min separately. To inactivate the SO_3166 and SO_3166–SO_3165 complexes, protein samples were heated at 95 °C for 10 min and were cooled before adding to the reactions. The reaction products from ompA mRNA and tRNAs were resolved by 15% TBE-urea gels (Invitrogen). To conduct rRNA cleavage assay, S. oneidensis carrying pHGE-SO_3166 and carrying empty plasmid pHGE were induced with 0.5 mm isopropyl 1-thio-β-d-galactopyranoside (IPTG) for 40 min, and total RNAs were isolated following the protocol of the Qiagen RNeasy Mini kit (Qiagen). Approximately 1.0 μg of total RNA was loaded to 1.0% agarose gel to check the integrity of 16S and 23S rRNAs. RNA was stained with SYBR safe.

SO_3166 DNA cleavage assay

To conduct DNA cleavage assay, plasmid pCA24N, λ dsDNA, and M13 ssDNA were used as substrate for the in vitro assay, respectively. Plasmid DNA (pCA24N) was isolated from strain E. coli BW25113/pCA24N, and λ dsDNA and M13 ssDNA were ordered from New England Biolabs. The reaction mixture for the SO_3166 DNA cleavage assay (15 μl) contained 1.0 μg of DNA, 20 mm Tris-HCl (pH 8.0), 300 mm NaCl, 25 mm EDTA, 20 mm CaCl₂, 20 mm MgCl₂, and 2 μg of purified SO_3166 protein or 1 μl of 2000 units/ml of DNase I (reaction devoid of EDTA). The reaction mixture was incubated at 37 °C for 0, 20, or 40 min separately. To inactivate SO_3166, protein samples were heated at 95 °C for 10 min and were cooled before adding to the reactions. The reaction was stopped by the addition of a stop solution (25% glycerol, 0.5% SDS, 0.05% bromphenol blue, and 50 mm EDTA) and was analyzed by electrophoresis on 1% agarose gels stained with SYBR safe (Invitrogen).

Protein expression and purification

Bacterial cells harboring pET28b-SO_3165–SO_3166 were grown to A_{600 nm} of 0.6 in LB media at 37 °C in the presence of 50 mg/ml of kanamycin. Induction of protein expression was initiated by adding IPTG to the culture to a final concentration of 1 mm, and cells were grown at 16 °C. Cells were pelleted after 18 h by centrifugation at 3500 rpm for 35 min at 4 °C. Cell pellets were suspended in the buffer containing 20 mm Tris (pH 8.0), 250 mm NaCl, 10 μg/ml of DNase, and 1 mm β-mercaptoethanol. The cell suspension was disrupted by a high-pressure homogenizer and then centrifuged at 16,000 × g for 50 min at 4 °C. The supernatant was then loaded onto a nickel-nitrilotriacetic acid column that was pre-equilibrated with 20 mm Tris (pH 8.0), 500 mm NaCl, 10 mm imidazole, 1 mm β-mercaptoethanol buffer. The His-tagged protein was eluted in 20 mm Tris (pH 8.0), 250 mm NaCl, and 250 mm imidazole. The complex was further purified by a Hitrap Q column (GE Healthcare) pre-equilibrated with 20 mm Tris-HCl, 100 mm NaCl, 1 mm dithiotheitol (DTT), pH 8.0, with a linear gradient of 100–1000 mm NaCl in 20 mm Tris-HCl (pH 8.0). Next the protein was purified by Superdex-200 chromatography on an ÄKTA Prime system (GE Healthcare) to obtain highly pure SO_3165–SO_3166 complex. The gel filtration buffer contained 20 mm Tris (pH 8.0), 100 mm NaCl and 1 mm DTT. The eluted fractions in all purification steps were analyzed by SDS-PAGE. The full-length SO_3165 gene was inserted into expression vectors pET28b, then pET28b-SO_3165 was transformed into BL21(DE3). SO_3165 was overexpressed and purified using the same producers described above.

The Se-Met SO_3166–SO_3165 complex were produced in minimal medium supplemented with 100 mg/liter of lysine, phenylalanine, and threonine, and 50 mg/liter of isoleucine, leucine, valine, and selenomethionine. The Se-Met protein production and purification were the same as described above.

Crystallization, data collection, structure determination, and refinement

The purified complex was concentrated to ∼10 mg/ml using a Millipore Amicon Ultra apparatus. The initial crystallization conditions were obtained through utilization of several sparse matrix screens (Hampton Research, USA) with the sitting drop vapor diffusion method at room temperature after 2–3 days. Crystal quality was optimized by adjust the concentration of the precipitant and buffer. The best crystal of SO_3165–SO_3166 was obtained in solution with 0.1 m Bicine (pH 8.6), 14% PEG6000, and 8% ethylene glycol at 20 °C.

The diffraction data from a single crystal of selenomethionine-substituted protein were collected on the beamline station BL19 U of SSRF (Shanghai Synchrotron Radiation Facility) using a Pilatus 6 M detector at a wavelength of 0.9788 Å. The total oscillation was 360° with 1° per image and the exposure time was 1-s per image. Before data collection, the crystals were soaked in the reservoir solution supplemented with 20% (v/v) glycerol for a few seconds and then flash-frozen in liquid nitrogen. All the data were processed by XDS (22). The Se-Met crystal structure of the SO_3166–SO_3165 complex was determined by the single wavelength anomalous dispersion method. The selenium atoms were located by the program Shelxd and then used to calculate the initial phases in Shelxe (23). The phases from Shelxe were improved in Resolve (24) and then used in Buccaneer (25) for model building. All structures above were refined with the program Phenix.refine (26) and manually corrected in Coot (27). The qualities of the final models were validated with the program MolProbity (28). Refinement statistics and model parameters are given in Table 1. The program PyMOL was used to prepare structural figures.

SAXS and low resolution model building

Synchrotron SAXS experiments were performed on the BL19U2 station of SSRF, equipped with a PILATUS 1 M detector (DECTRIS, Switzerland) (Table S3). The scattering was recorded in the range of the momentum transfer 0.018 Å⁻¹ < s < 0.321 Å⁻¹, where s = 4πsinθ/λ, 2θ is the scattering angle, and λ = 1.03 Å is the X-ray wavelength. The solutions were loaded in a flow-through quartz capillary cell with a diameter of 1 mm and a wall thickness of 10 μm, temperature controlled at 22 °C. The radiation damage was checked with 20 successive exposures of 1 s. To exclude concentration dependence, three different concentrations, 1, 3, and 5 mg/ml of purified SO_3166–SO_3165 complex (corresponding to 8.3, 25.0, and 41.7 μm, respectively) were prepared and measured. All SAXS data were processed with the program package ATSAS (29). The scattering of buffers were subtracted from that of the samples, and the forward scattering I(0) and the radius of gyration R_g were derived by the Guinier approximation I(q) = I(0) exp(−q² R_g^2/3) for q R_g < 1.3 using PRIMUS (30). The pair-distance distribution functions, p(r) and the maximal dimension of the macromolecule, D_max were calculated using indirect Fourier transformation and the program GNOM (31). To model the structures of the SO_3165 and SO_3166–SO_3165, 10 independent models were generated with the programs DAMMIN or GASBOR (32) in fast mode, compared and aligned with SUPCOMB (33), and averaged with DAMAVER (34) to determine common structural features and representative shapes. Theoretical scattering patterns I(s) from the available high resolution crystal structures were calculated by a program CRYSOL (35).

Circular dichroism (CD) spectroscopy

Purified SO_3165 (WT and its variants) was loaded onto a Superdex 200 column (GE Healthcare) equilibrated with PBS buffer (pH 8.0). The elution containing SO_3165 was subsequently concentrated to 0.5–1.0 mg/ml. The CD spectra were measured on the 4B8 station of Beijing Synchrotron Radiation Facility (BSRF) at 1-nm bandwidth with a 1-nm step resolution from 170 to 250 nm at 25 °C. The data were averaged over eight accumulations. The data are processed by CDtool (36) and further analyzed by Dichr web (http://dichroweb.cryst.bbk.ac.uk/html/home.shtml).⁵

Cell toxicity assay

E. coli BL21 harboring pET28a-based plasmids were grown in Luria-Bertani (LB) medium supplemented with 50 μg/ml of kanamycin at 37 °C and 0.3 mm IPTG were added at A_{600 nm} ∼ 0.1. The viability (CFUs/ml) was measured at different time points for individual cell cultures. Viability were calculated by serially diluting the cells in 10-fold steps and plated onto the LB agar. The plates were prepared for pictures after incubation at 37 °C for 12 h. Triplicate measurements were performed and similar results were obtained for each measurement unless stated.

Data availability

The atomic coordinates and structure factors of the SO_3166–SO_3165 complex have been deposited in the RCSB Protein Data Bank with PDB code 5YEP.

Author contributions

X. J., J. Y., and Z. G. data curation; X. J., J. Y., Z. G., Y. D., X. W., and H. Z. formal analysis; X. J. and Y. D. investigation; X. J., J. Y., and H. Z. writing-original draft; J. Y., Z. G., and G. L. validation; J. Y., Z. G., and G. L. methodology; Y. D. project administration; X. W. and H. Z. conceptualization; X. W. and H. Z. funding acquisition; X. W. and H. Z. writing-review and editing.