All vectors were transformed into Rosetta (DE3) competent cells using the heat shock method. A single colony was inoculated in Luria-Bertani (LB) medium supplemented with the corresponding antibiotic, such as ampicillin (100 μg/ml) and kanamycin (50 μg/ml) at 310.15 K (37°C) and 220 rpm. Protein expression was induced with 1 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG) once the bacterial growth reached OD600 of 0.8 and further incubated at 289.15 K (16°C) for 20 h. Bacterial cells were harvested by centrifugation at 8,000 rpm for 6 min and 277.15 K (4°C), followed by resuspension in 0.3 M NaCl and 0.05 M Tris pH 8.0, and further lysed with ultrasonicator (Qsonica; USA) in an ice bath. After centrifugation at 12,000 rpm for 30 min and 277.15 K (4°C), the protein was purified from the supernatant utilizing Ni-NTA resin (GE Healthcare: USA), followed by Superdex 75 pg column (HiLoad™ 16/600; GE Healthcare; USA). The purity and concentration of the proteins were assessed with SDS-PAGE and OneDrop™ OD-1000 + spectrophotometer (WINS; China), respectively. In addition, antitoxin PrpA^FL (expressed by vector PrpA-N-his-pET28a) carries a 6 × his tag on the N-terminus after being translated, and PrpA²⁵⁴ (expressed by vector PrpA^2–54-C-his- pET28a) carries a 6 × his tag on the C-terminus. Toxin PrpT [expressed by vector PrpT-pET22b(+)] carries no tag. Their recombinant sequences could be accessed in Supplementary Table 1. The theoretical relative molecular weight of PrpT, PrpA^FL with N-terhexahistidine, and PrpA^2–54 with C-terhexahistidine is ∼11.4, ∼11.66, and ∼7.05 kDa, respectively. Moreover, all vectors and oligonucleotide fragments used in the current study were obtained from Sangon Biotech Co., Ltd (Shanghai, China).

The finely purified PrpA^2–54 and PrpTA were concentrated to ∼5.2 and ∼3.6 mg/ml, respectively, in 0.1 M NaCl with 0.05 M Tris (pH 8.0). Initial crystal screening was performed by mixing an equal volume of protein and reservoir solution utilizing the sitting drop vapor diffusion method at 289.15 K (16°C). Two days post-crystallization, the crystal of PrpA^2–54-form I was observed in a condition containing 0.2 M sodium acetate trihydrate, 0.1 M sodium citrate (pH 5.5), and 5% (w/v) PEG 4000. Similarly, crystals for PrpA^2–54-form II were obtained in 0.2 M magnesium chloride hexahydrate, and crystals for PrpTA complex were obtained in 0.1 M HEPES sodium at pH 7.5, 30% (v/v) PEG 400, 0.1 M sodium cacodylate (pH 6.0), and 15% (w/v) PEG 4000.

Crystals were picked with nylon loops and cryoprotected in a reservoir solution supplemented with 20% glycerol. All the diffraction datasets were collected at the beamlines in Shanghai Synchrotron Radiation Facility (SSRF) utilizing the single-wavelength small-angle oscillation method. All datasets were initially indexed and integrated with XDS, followed by scaling with an aimless module integrated into CCP4i (v7.1). The initial structural model of PrpA^2–54 was determined by Phaser in PHENIX (v1.19.2) with VcParD2 antitoxin (PDB: 7B22) as a search template utilizing molecular replacement method, followed by model building with Autobuild module in PHENIX. Similarly, the structure of the PrpTA complex was searched against the refined structure of the PrpA^2–54 monomer and Caulobacter vibrioides ParE toxin monomer. The structures were iteratively refined using Refmac5, WinCoot (v0.9.6), and phenix.refine offered by PHENIX. The detailed statistical information about these datasets is summarized in Supplementary Table 2.

SEC-multi-angle light scattering experiments were conducted using an AKTA™ pure HPLC system (GE Healthcare, USA) connected in-line with a DAWN HELEOS II (Wyatt Technology, Santa Barbara, CA, USA) eight-angle light-scattering detector, followed by a refractive-index detector (Wyatt Technology, Santa Barbara, CA, USA). SEC-MALS system was equilibrated with the corresponding running buffer at 0.5 mL/min for 12 h prior to the sample loading. A series of PrpA, PrpA^2–54, and PrpTA samples were prepared in several kinds of buffers (including 50 mmol/L tris–HCl and 100 mmol/L NaCl pH 8.0, 50 mmol/L tris–HCl and a300 mmol/L NaCl pH 8.0, 50 mmol/L tris–HCl and 500 mmol/L NaCl pH 8.0, 50 mmol/L MES and 500 mmol/L NaCl pH 5.5, and 100 Mm (NaH²PO⁴/Na²HPO⁴) pH 8.0), and 0.1 ml was injected into the loop for each dilution. Accordingly, the absolute molar mass of each protein sample could be determined based on the data processed by the ASTRA (v7.0.1) offered by Wyatt company.

Circular-dichroism spectra were recorded at room temperature using a Chirascan™ qCD Spectrometer (Applied Photophysics Limited, UK) at a concentration of 0.4 mg/ml in 100 Mm (NaH²PO⁴/Na²HPO⁴) pH 8.0 based on 0.5 mm optical path. For each CD experiment, the background was measured one time, the buffer was measured two times, and samples were measured three times to reduce error and noise. All data were acquired with the subsequent parameters, bandwidth: 1 nm, wavelength scanning range: 180–260 nm, and time per point: 0.5. Pro-Data (v4.5.1825.0), and BeStSel were used to view and process CD spectra data to consequently access information on secondary structure composition.

Sequence-specific interactions between antitoxin PrpA and duplex oligonucleotide are mostly mediated by N-terminal RHH domains of PrpA via recognizing the conserved 5’-(G/A)TTTG(T/A)AAT(A/G)-3’ motif which could be pseudo-palindromic or asymmetric (Ni et al., 2021). Based on the structure of transcriptional repressor CopG in complex with 22 bp dsDNA [PDB: 1EA4, which revealed a tetramer consisting of two dimers and associated by a crystallographic dyad, interacting in the same way as two dimers in the unliganded structure (Del Solar et al., 1989; Gomis-Ruth et al., 1998; Costa et al., 2001)], a possible dsDNA-RHH interaction mode was generated by NUCBIND. Combined with the structure of PrpA^2–54 antitoxin homotetramer, a knowledge-based model of PrpA^2–54 homotetramer in complex with prpAT promoter was proposed and later improved by energy minimization refinement and water refinement successively using refinement module offered by HADDOCK 2.4 with default parameters. The iteratively refined model was finally validated by the PISA server.

To investigate the assembly mechanism of the PrpTA system in solution, a series of SEC-MALS experiments were performed on purified PrpTA complex, PrpA^FL, and PrpA^2–54 samples in this study (Figure 1A and Supplementary Figures 1, 2). Moreover, we succeeded in crystallizing PrpTA but failed to crystallize PrpA^FL due to the flexibility in the CTD reported before (Ni et al., 2021). However, we were lucky to crystallize truncated PrpA^FL (PrpA^2–54) and finally solved two forms of PrpA^2–54 crystal structures and the high-resolution PrpTA crystal structure (Figure 1B and Supplementary Figure 3). Considering the theoretical relative molecular weight of PrpT (∼11.4 kDa) and PrpA (∼11.66 kDa), our SEC-MALS results reflected that PrpTA complex, PrpA^FL and PrpA^2–54 present an estimated absolute mass matching tetramer, tetramer and dimer, respectively. These results highlighted that deletion of PrpA^CTD (55–86 residues) could mediate the change in the oligomerization states of PrpA^FL from tetramer to dimer, suggesting the importance of CTD for the PrpA^FL homotetramer. Furthermore, it seems that the homodimer is actually the minimal assembly unit of PrpA^2–54 in the solution. Overall with the crystal structure of the PrpTA complex, it was certain that the PrpTA biological assembly unit indicated with PrpT:PrpA²:PrpT (Figure 1Ba) contains two PrpT monomers (Figure 1Bb), which were isolated from each other, and a PrpA homodimer (Figure 1Bd). In other words, PrpT:PrpA²:PrpT heterotetramer is maintained by one single PrpA^FL homodimer interface formed by N-terminal PrpA^RHH domains and two PrpT-PrpA^FL heterodimer interfaces, which is similar to its structurally relevant complex, such as CvParDE, and consistent with the conclusion that PrpA dimer interface and PrpT-PrpA heterodimer interface with an estimated CSS (Complex Formation Significance Score, calculated by PISA server) value of 1 contribute to PrpTA heterotetramer complexation. To sum up, the PrpTA complex exists as the PrpT:PrpA²:PrpT heterotetramer in solution and consists of two toxin monomers together with an antitoxin PrpA homodimer which appears to be depolymerized from homotetramer and the minimal functional unit of PrpA^FL.

PrpT toxin starts from a RHH fragment in the N-terminus, helices of which are connected by a single glycine residue, and the β1 tends to be packed with the EF-P OB-like fold that is mainly composed of C-terminal 3-stranded antiparallel β sheet (β3–β5), as shown in Figure 1Bb. To determine accurately the protein family of PrpT, the sequence alignment, phylogenetic analysis, and structural superposition analysis with its homologs were performed. The sequence alignment and phylogenetic tree reflected that PrpT and its sequence homologs could be divided into three subgroups (Supplementary Figures 4A,B). In addition to the residue Leu that is highly conserved in the α 1 helix of all known structures, a conservative (Gln/Gly)-Gly diad is specific to members in group 1, that is, ParE family. Accordingly, the superposition of all solved structures of toxin proteins from the sequence alignment above could also be divided into three subgroups based on their structure deviations with toxin PrpT (Figure 2Aa). The crystal structure of PrpT is similar to structures of ParE family members with an average estimated r.m.s.d. of 0.855 Å and the main difference lies in the conformation of α 2–β3 and β4–β5 loops, which might play a key role in adjusting and maintaining the conservative three-dimensional structure (Figure 2Ab). However, structures from group 2 or group 3 could be distinguished apparently from PrpT due to large structure deviations mainly resulting from the flexibility of loops and the length differences of α 1 and α 2 with corresponding helixes in PrpT (Figure 2Ac). Therefore, the toxin PrpT could be assigned to the ParE family.

As illustrated in Figure 1Bc, PrpA^FL begins with an extended conformation (β1) spanning from 5 to 9 aa, followed by the α 1 helix (11–23 aa) and the α 2 helix (30–45 aa). The C-terminal PrpT neutralizing domain (49–85 aa), which is also the tetramerization domain, is connected with the C-terminus of the α 2 helix by a hinge region composed of AGS triad [Figure 2B (upper)]. Due to the presence of glycine residue in the N-terminus of α 1 helix, the centroidal axis is rotated ∼70° in relation to helix α 2. For a similar reason, the direction of the helix α 1 is actually rotated ∼80° with respect to the strand β1 (Supplementary Figure 5c). Compared to the ∼47.9% helix content of PrpA^FL in solution [Figure 2B (bottom)], the total helix content of PrpA^FL in the PrpTA complex is ∼55.7%. Higher helix content seems to hint at a conformational change occurring in the PrpA^CTD upon PrpT binding. The DALI topological analysis revealed not only the C-terminal diversity of CopG/MetJ/Arc family with the conservative RHH motif but also the structure uniqueness of PrpA^FL. Despite the higher sequence similarity, PrpA^FL exhibits a difference in the secondary structural elements (α 2–α 3 region) with its homologs and obvious structural deviation in the backbone track of CTD (Figure 2C). Even though VcParD2 (PDB: 7B22) and EcParD (PDB: 5CZE) exhibit structural similarity to a different degree with the N-terminal PrpA^RHH domain (r.m.s.d. = 1.56 Å) and the PrpA^CTD (r.m.s.d. = 4.41 Å), respectively, the structure of PrpA^FL is still unique compared to structures already deposited in PDB together with the structure predicted by AlphaFold (r.m.s.d. = 5.41 Å). In summary, the folding of PrpA^FL is definitely novel and distinct from any other structures deposited in PDB.

To investigate the structural details of PrpT:PrpA²:PrpT, crystal structures of the PrpTA complex and PrpA^2–54 were all carefully analyzed. Like other members from the RHH family, PrpA monomer polymerizes in solutions and crystals in a highly symmetrical manner via a local twofold axis to form a two-stranded antiparallel β-sheet (Figure 3A). The symmetric dimerization interface could be found in PrpT:PrpA²:PrpT (excluding residue Ser3B) or either form of PrpA^2–54 crystal structures (Figures 3A,B). Moreover, the comparative structural analysis highlighted that PrpA^2–54 dimers are similar to each other with an r.m.s.d. value of ∼1.38 Å over 96 residues (Supplementary Figure 3). Thus, only the interface that existed in PrpA^2–54-form II will be described in detail here. PISA analysis reflected that the hydrogen bond/salt bridge interaction network of dimerization interface is primarily offered by Met6, Val8, Asp9, Ser31, and Arg35 from each monomer together with Arg15A/B, Thr10A/B, Arg54A, Arg4B, and Asp26B for the PrpA^2–54 (or symmetric Thr7A/B, Glu13A/B and asymmetric Ser3B for PrpA^FL in PrpTA complex). The dimerization interface in PrpTA or PrpA^2–54-form II buries roughly 1700 Ų with the nearly identical negative ΔG value (∼24 kcal/mol) corresponding to hydrophobic interfaces. It is presumably because bonds are broken by even small shifts in distance or orientation caused by the tag that interface symmetry of terminus that carries the tag of PrpA^FL or PrpA^2–54 is always worse than that of the other one. In addition, a number of hydrophobic interactions between α helices could also partly contribute to the formation of a stable PrpA^FL homodimer. Crystal structures of PrpT:PrpA²:PrpT heterotetramer complex and PrpA^2–54 suggested the significance of RHH dimer for the PrpT-PrpA recognition and the formation of PrpA^FL multi-dimer. Overall with LigPlus analysis (Supplementary Figure 6), it is rational to assert that the 2-stranded antiparallel β-sheet formed by two adjacent ribbons spanning from 5 to 9 aa is the basis of oligomerization of the PrpTA system.

To investigate whether PrpA^CTD undergoes a conformational transition upon toxin PrpT binding, structural superposition of PrpA^2–54 monomers with PrpA^FL was conducted. As demonstrated in Figure 4A, the results uncovered the obvious orientation change of α 3 upon PrpT binding, which is consistent with the flexibility feature of the PrpA^CTD. Except for the asymmetric Arg54A-ASP26B contact in PrpA^2–54, the protein backbones of PrpA^2–54 or PrpA^FL project away from the dimer interface without further intra- or inter-contacts, suggesting the flexibility of PrpA^CTD (Figure 3B). That is, the PrpT binding could result in the dramatic flexibility-to-stability transition of PrpA^CTD, which provokes the de-tetramerization of PrpA^FL antitoxin and the rotation of PrpA^CTD by ∼80° around the AGS residue triad, further leading to the extra interaction between toxin PrpT and RHH domains of PrpA homodimer (Figure 4B). PrpT:PrpA²:PrpT heterotetramer is formed via a local twofold axis and highlights two extremely similar PrpT-PrpA^FL contact interfaces. Therefore, only one heterodimer interface will be discussed here. PrpA and PrpT interact via two hydrophobic subdomains surrounded by electrostatic interactions (Supplementary Figures 5, 7). The α 3 and α 4 make up the core PrpT binding region, the surface polarities of which are exactly opposite to the two subdomains of PrpT, which are mainly comprised of N-terminal ribbon–helix-helix and the C-terminal 4-stranded anti-parallel β-sheet (β2–β5), respectively (Figure 4C). The heterodimer interface dominated by a series of salt bridges/hydrogen bonds covers a surface of ∼2000 Ų with an estimated ΔG value of −17 kcal/mol. Even though the number of extra contacts between PrpT and PrpA^RHH accounts for just ∼20% of all PrpT-PrpA^FL interactions according to PISA, they are deemed to limit the swinging ability of PrpA^CTD in the local space, thus further enhancing the stability of PrpT:PrpA²:PrpT heterotetramer. Together, we infer that the conformational changes in the PrpA^CTD are significant for neutralizing the toxin.

Furthermore, compared with relevant toxin–antitoxin complexes reported before, as demonstrated in Figure 5, PrpT-PrpA²-PrpT heterotetramer is the only toxin–antitoxin complex whose toxin monomer could simultaneously interact with the CTD and the NTDs of the antitoxin homodimer, thus neutralizing the toxin. A similar strategy is also found in MtRelBE2 (Figure 5F), of which adjacent α 2–α 3 loop could interact with each other limiting the flexibility of CTD, meanwhile, blocking the contacts between antitoxin NTD and toxin monomer. In brief, the PrpA dimerization interface, PrpA^CTD-PrpT contacts, and PrpA^RHH-PrpT contacts all contribute to PrpTA oligomerization. In actual fact, except for EcParDE (Figure 5E) and PhRelBE (Figure 5G) whose oligomerization is maintained by toxin–toxin contacts that are actually absent in PrpT-PrpA²-PrpT heterotetramer, the oligomerization of CvParDE (Figure 5B), PaParDE (Figure 5C), MoParDE (Figure 5D), MtRelBE (Figure 5F), and MjRelBE (Figure 5H) all only depend on their NTDs, especially the N-terminal antiparallel β sheet. To sum up, the PrpT-PrpA²-PrpT heterotetramer displays a novel interaction profile between toxin and antitoxin, which contributes to the stability of the PrpTA complex.

Similar to the arrangement of CopG (PDB: 2CPG) (Gomis-Ruth et al., 1998; Costa et al., 2001) monomers in the asymmetric unit (AU), PrpA^2–54-form I monomers are arranged in a cogwheel-like form in the AU [Supplementary Figure 3 (right)]. Apart from a type A-D dimer, either B or C could further dimerize with its symmetric mate generating type B-B# and type C-C# dimers (# represents symmetry mate). Since the dimer–dimer interfaces in AD-BB# and AD-CC# are similar to each other with an overall r.m.s.d. of 1.05 over 144 residues, only the AD-CC# interface will be, thus, discussed at length here. As demonstrated in Figure 6A (left), a functional PrpA^2–54 tetramer is defined by inter-dimer crystallographic contacts between RHH domains with positively charged DNA-binding surfaces exposed to the solvent environment. The AD and CC# interact via a surface that is electrostatically complementary and somewhat slightly hydrophobic, especially the interface formed by α 2 helixes of PrpA. Salt bridges and hydrogen bonds below 4 Å are distributed mainly through the α 1–α 2 turn and α 2 helices of molecule AD together with almost equivalent parts of a molecule CC# [Figure 6A (right)]. In addition, the interface generated by AD and CC# dimers covers a total surface area of ∼1100 Ų, which is a little smaller than that of the dimerization interface. The inter-dimer interface of PrpA^2–54 is thought to be strong enough since it can still exist as a multi-dimer in certain cases in the absence of PrpA^CTD (Figure 7A). A similar dimer–dimer interface dominated by hydrogen bonds and salt bridges was also reported in VcParD2 tetramer [Figure 6B (upper)], especially the symmetric residues E32, R35, and R39 (matching E31, R34, and R38 in SaCopG, respectively) are extremely conserved in primary and tertiary structures. However, they are all totally different with inter-dimer interface dominated by hydrophobic van de Waals interaction that existed in SaCopG tetramer bound to DNA [Figure 6B (bottom)]. However, the inter-dimer interface of PrpA^2–54 displays a novel feature that all monomers could participate in the formation of tetramer suggesting a much more extensive and strong contact. In contrast, few residues from one single monomer from each dimer, that is, only two chains mediate the corresponding inter-dimer interface for SaCopG and VcParD2. The difference in the inter-dimer interface presumably lies in the different conformations caused by residue substitutions, such as Asp9, Glu43, and Lys28 (corresponding to Thr8, Gly27, and Asn42 in VcParD, respectively), which allows much more complex contacts between PrpA^2–54 monomers (Figure 6C). Moreover, an unfavorable positive–positive interaction is abolished by Gly47 substitution matching the Lys46 in VcParD. In general, the dimer–dimer interface in PrpA^2–54 multi-dimer possesses much more extensive inter-monomer contacts.

A circular hexadecamer (∼112.58 kDa) consisting of 8 PrpA^2–54 dimers could be found in both forms of PrpA^2–54 crystals (Figure 7B). Higher-order form of PrpA^2–54 is maintained by a dimer interface and inter-dimer interface. The maximum diameter of the annular doughnut-like hexadecamer is ∼80 Å and that of the hole in the center is approximately 22 Å. The C-terminal α 2 helix of each PrpA^2–54 monomer stretches outward from the inner side of the hexadecamer; overall with the flexibility of CTD, it is reasonable to infer that the replacement of PrpA^FL would provide an entropic penalty for oligomerization, which is similar to the effect made by the IDR of VcParD2. To further investigate which factors could mediate the oligomerization of PrpA^2–54, a series of SEC-MALS experiments were performed (Figure 7A). PrpA^2–54 tends to present an absolute molecular weight matching PrpA^2–54 dimer in most cases, and no clear dependency on the ion strength or pH of solutions could be observed. However, the increase in protein concentration within a limited range seems to result in a higher absolute molecular weight (41.38–42.41 kDa, probably hexamer). In addition, the hexadecamer could be occasionally observed in solution with an extremely smaller proportion (∼4%). The PrpA^2–54 has a tendency to assemble into a relatively stable homotetramer in phosphate buffer (PB), which is distinct from its performance in Tris–HCl buffers. Taken together, the increase in the protein concentration and PB are favorable for the oligomerization of PrpA^2–54 in the solution.