ParD Antitoxin Hotspot Alters a Disorder-to-Order Transition upon Binding to Its Cognate ParE Toxin, Lessening Its Interaction Affinity and Increasing Its Protease Degradation Kinetics

Abstract

Type-II toxin–antitoxin (TA) systems are comprised of two tightly interacting proteins, and operons encoding these systems have been identified throughout the genomes of bacteria. In contrast to secretion system effector–immunity pairs, TA systems must remain paired to protect the host cell from toxicity. Continual depletion of the antitoxin results in a shorter half-life than that of the toxin, though it is unclear if antitoxins can be effectively degraded when complexed with toxins. The current work probed the protein–protein interface of the PaParDE1 TA system, guided by an X-ray crystal structure, to determine contributions of antitoxin amino acids to interaction kinetics and affinity. These studies identified a “hotspot” position that alters the binding mode and resulting affinity (K_D) from 152 pM for a 1:1 model for wild type to 25.5 and 626 nM for a 2:1 model with mutated antitoxin. This correlates with an altered induced secondary structure upon complexation with PaParE1 and increased kinetics of Lon protease digestion of the antitoxin despite the toxin presence. However, the decreased affinity at this hotspot was essentially reversed when the antitoxin dimerization region was deleted, yielding insights into complex interactions involved in the tight association. Removal of the antitoxin C-terminal seven amino acids, corresponding to the site of a disorder-to-order transition, completely prevents association. These studies combine to provide a model for the initiation of the TA interaction and highlight how manipulation of the sequence can impact the antitoxin disorder-to-order transition, weakening the affinity and resulting in increased antitoxin susceptibility to degradation.

Graphical Abstract

INTRODUCTION

Protein–protein interactions mediate essential functions for life, and understanding the molecular basis of their specificity can provide valuable insights into their functions.¹ Interactions between effectors and cognate immunity proteins, such as those found in toxin–antitoxin (TA) systems, type-VI secretion systems, and contact-dependent inhibition systems, determine the protection of host cells and, in some cases, the targeting of other cells. The type-II TA systems are composed of a non-secreted toxin protein that is neutralized by a highly specific cognate antitoxin protein that serves an “immunity” function.^2,3 In other effector–immunity pairs, the neutralizing immunity proteins are known to be extremely stable and provide constant protection to the host as toxins are secreted.⁴ Remarkable similarity has been noted among these different effector–immunity pairs, in particular among toxin mechanisms of cell death involving nuclease activity.^4,5 In contrast to immunity proteins of effector–immunity pairs, antitoxins of TA systems are continuously depleted by ATP-dependent AAA⁺ proteases. This results in a shorter half-life of the antitoxin relative to the non-secreted toxin protein, such that continual translational replenishment of antitoxin is needed to maintain toxin neutralization.^6–13

TA systems were originally identified as mediators of plasmid inheritance through a post-segregational killing (PSK) mechanism that is predicated on the shorter half-life of antitoxin molecules and the subsequent liberation of toxin activity.³ However, TA systems are also integrated into the genomes of bacteria and archaea; the role of these systems within their host cells is an issue of active debate.^{2,3,7,8,10,14,15} For secreted effectors, the interaction with cognate immunity proteins in host cells is prevented by chaperones that deliver the effector to the secretion apparatus.¹⁶ Presumably in these cases, the interaction with a cognate immunity protein should be essentially irreversible to maintain host protection. For TA systems, the two protein components associate directly, and it is unclear if the tightly interacting antitoxin proteins can be degraded directly from a complex with toxin, as is implied for the PSK model of TA system function.¹⁷

Among a given TA system family, individually encoded antitoxins have low sequence similarity at the sites of these interactions, leading to high specificity for their cognate toxins with only limited “cross-talk” between homologous TA systems.^18–21 Type-II TA systems form extensive protein–protein interfaces, typically with nanomolar affinity and spanning surface areas of more than 1400 Ų.^19,21–25 A recent study detected the degradation of TA system antitoxins without an accompanying gain of toxin activity for ribonucleic acid (RNA) cleaving toxins within Escherichia coli, consistent with a lack of antitoxin degradation from pre-formed complexes.²⁶ This appears to contrast the canonical PSK model, which requires the generation of free toxin after loss of the plasmid encoding TA system genes,¹⁷ leading to questions of how tightly interacting TA proteins could be parted as antitoxin is degraded. This led us to further question if there might be “hot spots” within these TA interactions, such that interruption at one specific site could facilitate antitoxin removal from a TA complex. We hypothesized that the identification of such “hot spots” would provide insights into the vulnerable regions of interactions amenable to interruption by cellular components or by intentional external manipulation.^20,27–31

The type-II TA system ParDE was first identified on the broad-range plasmid RK2 as a plasmid stabilization element that causes growth inhibition upon plasmid loss.³² ParE toxins are structural homologs of the TA system RelE-superfamily of RNase enzymes, as well as colicin and barnase nucleases.³³ However, ParE toxins have been well-documented as causing growth inhibition through the inhibition of deoxyribonucleic acid (DNA) gyrase.^22,34,35 The ParE1 toxin from Pseudomonas aeruginosa (PaParE1) used in this study has been shown to inhibit DNA gyrase-mediated supercoiling, and its attenuated toxicity could provide target protection to cells treated with anti-gyrase antibiotics.²² Other ParE toxins, such as three ParE proteins from Vibrio cholerae, have been shown to select for chromosome II of V. cholera in a PSK-type function.³⁶ It has been demonstrated that ATP-driven proteases Lon and ClpP degrade intracellular antitoxins in vivo, including for the plasmid-derived RK2–ParDE system.^9,11,13,37 Given the overlap of ParE activity with the well-validated gyrase antibacterial target, there is interest in the manipulation of toxin activation via liberation from antitoxin.

The current study examined the ParD–ParE interactions of the chromosomally encoded PaParDE1 system with an aim of characterizing the important molecular determinants of complex formation and stability. Guided by a crystallographic structure of the PaParDE1 complex (PDB ID: 6XRW), site-directed mutagenesis experiments identified an amino acid change that was sufficient to decrease the interaction affinity (K_D) from picomolar to high nanomolar, concomitant with a change in the mode of binding. Further, C-terminal truncations identified that the last seven amino acids of the PaParD1 antitoxin are required for initiation of the interaction with the PaParE1 toxin. The interaction with the PaParE1 toxin proved critical for the protection of PaParD1 from in vitro Lon protease degradation. Overall, these results yield insights into how the complex formation is initiated, individual PaParD1 interfacial amino acid contributions to a disorder-to-order transition and thus to the strength of interaction, and how this in turn impacts protease degradation of antitoxin.

EXPERIMENTAL DETAILS

Expression and Purification of P. aeruginosa Pa-ParDE1.

The cloning and expression of the genes encoding PaParD1 and PaParE1 have been previously described.²² Mutants were generated using the QuikChange (Stratagene) or Q5 (NEB) mutagenesis kits (sequences and primers are given in Table S1). In brief, these genes were separately cloned into the pET-15b plasmid in frame with an N-terminal 6× his-tag and thrombin cleavage site. Protein expression was induced by the addition of 1 mM IPTG to the media; PaParD1 was expressed at 25 °C for 4–6 h to minimize in situ degradation, while PaParE1 was expressed at 16 °C overnight (possible due to its limited toxicity) prior to harvesting by centrifugation.

Protein purification was carried out by closely following previously reported procedures.²² In short, cell pellets were resuspended in a 50 mM Tris-HCl pH 8.5, 300 mM NaCl, 40 mM imidazole buffer and lysed using a high-pressure homogenizer (Emulsiflex). The clarified lysate was passed through a HisTrap FF Ni-NTA column (GE Healthcare) and eluted using a linear gradient with a 50 mM Tris-HCl pH 8.5, 300 mM NaCl, 500 mM imidazole buffer. The eluted protein was desalted in a 50 mM Tris-HCl pH 8.5, 300 mM NaCl buffer using a HiPrep 26/10 desalting column (GE Healthcare) and concentrated using a 3 kDa molecular weight cut-off (MWCO) Ultra-15 concentrator (Amicon) prior to separation by size-exclusion chromatography using a Superdex 75 10/300 GL (GE Healthcare) column pre-equilibrated with a 50 mM Tris-HCl pH 8.5, 300 mM NaCl buffer.

When necessary, cleavage of affinity (6×-his) tags was performed prior to size-exclusion chromatography using thrombin. Proteins were buffer-exchanged into 50 mM HEPES pH 7.5, 150 mM NaCl. Cleavage was performed using 1 U thrombin (bovine, Sigma) per 1 mg protein overnight at 4 °C. The uncleaved protein was separated from the cleaved protein using a HisTrap FF Ni-NTA column according to the method described above. Gel-filtration was then carried out using a Superdex 75 10/300 GL column (GE Healthcare) pre-equilibrated in 50 mM Tris-HCl pH 8.5, 300 mM NaCl.

The PaParDE1 complex was formed by first buffer exchanging purified PaParD1 and purified PaParE1 into 50 mM MES pH 6.0, 300 mM NaCl, 5% glycerol, prior to mixing 1:1 for 30 min to 1 h at 4 °C. After mixing, the complex was concentrated (if necessary) prior to purification using a HiLoad 16/600 Superdex 75 pg column (GE Healthcare) or Superdex 75 10/300 GL column (GE Healthcare) preequilibrated in a 25 mM MES pH 6.0, 150 mM NaCl, 5% glycerol buffer. Protein purity and integrity were assessed under denaturing and reducing conditions with 12% acrylamide gels in a Tris-HCl–tricine buffer system.³⁸

Structure Determination and Analysis of PaParDE1.

Initial broad screening of the PaParDE1 complex (10.5 mg/mL in 25 mM MES pH 6.0, 150 mM NaCl, 5%, glycerol) was carried out at 293 K in a 96-well plate using sitting drop vapor diffusion crystallography with the aid of a mosquito crystallization robot (TPP LabTech). Optimization screens were performed in 24-well plates using sitting drop vapor diffusion. Diffraction quality crystals were obtained after 72 h in a condition of 7.5% (v/v) 2-methyl-2,4-pentanediol (MPD), 20% PEG 6000, 25 mM ammonium sulfate, and 100 mM Tris-HCl pH 7.0. Prior to data collection, PaParDE1 crystals were plunged in liquid nitrogen directly from the crystallization mother liquor. X-ray diffraction was collected on a MicroMax 007HF microfocus X-ray generator (Rigaku) with a Pilatus 200K silicon pixel detector (Dectris). Data processing and scaling were performed using HKL3000R (HKL Research Inc.). Phasing was solved using the program phaser,³⁹ as implemented in the package Phenix, for molecular replacement. Many starting models were screened, with satisfactory solutions only arising using a hybrid template of a ParE–ParD dimer, lacking the ParD DNA binding domain, generated through manual inspection and curation of regions and loops; two of these dimer models were placed. Refinement of phases was initiated with AutoBuild (Phenix package) that placed missing protein regions, followed with iterative cycles with Phenix refine and manual adjustments in coot.^39,40 Statistics for data collection and refinement are presented in Table S2. Subsequent structure-based sequence alignments and visualization of PDB structures were performed using UCSF Chimera;⁴¹ sequence alignment figures were prepared using the ESPript 3 webserver.⁴² Surface area values and energetics were obtained using the PDBePISA webserver from EMBL.⁴³

Biolayer Interferometry.

Prior to biolayer interferometry (BLI), samples were desalted into a 50 mM Tris-HCl pH 8.5, 300 mM NaCl, 0.5% BSA, 0.05% Tween-20 buffer, and all solutions were matched to these buffer conditions. BLI was performed at 25 °C using an Octet RED96 interferometer (Pall ForteBio) by coupling 100 nM WT or mutant His-tagged ParD1 proteins to Ni-NTA BLI sensor pins. WT-untagged PaParE1 was serially diluted (10 to 0.3125 nM; I68A PaParD1 required 10× higher concentrations) and incubated with ParD1-coupled pins until the association approached saturation, and then the dissociation rates were measured. Rates were calculated by fitting the curves to a 1:1 Langmuir binding model, except for I68A which required a 2:1 binding model to be fitted, as performed with manufacturer-provided BioAnalysis software (v 8.2). Individual curves were analyzed to ensure high quality fits with R² values greater than 0.95 (see Figure S1); individual assays were repeated at least three times to obtain average K_D values, where possible, as calculated from the ratio of the dissociation rate to the association rate; in the absence of dissociation rates, the association rates were compared (the summary of replicates is given in Table S3).

Size Exclusion Chromatography-Multiangle Light Scattering.

Size exclusion chromatography–multiangle light scattering (SEC–MALS) experiments were carried out at room temperature using the flow cell of a Wyatt miniDAWN Treos system integrated with a refractive index (RI) detector. Samples were introduced from a liquid chromatography system and separated over Superdex 200 Increase 10/30 GL (GE Healthcare) equilibrated in 50 mM MES buffer, pH 6.0, 300 mM NaCl, 5% (v/v) glycerol. The analysis followed the manufacturer’s recommendations using the ASTRA software package (Wyatt Technologies, v. 6.1) to calculate the weight-averaged molecular weight; protein concentrations were determined from the RI signal (see Figure S2). Data were exported from ASTRA and re-plotted in GraphPad Prism (v. 6.0d) for graphic presentation.

Fourier-Transform Infrared Spectroscopy.

The protein secondary structure was measured via amide I and amide II stretching from wavenumbers between 1700 and 1500 cm⁻¹ measured at room temperature using a Tensor II Fourier transform infrared (FTIR) CONFOCHECK (Bruker) equipped with a BioATR unit. All samples were desalted into 50 mM MES pH 6.5, 300 mM NaCl, 5% glycerol and concentrated to 1 mg/mL prior to analysis. A blank run of buffer was performed and subtracted from each sample to eliminate the background buffer signal, and three technical replicates were performed for each measurement (see Table S4). The protein secondary structure values reported were calculated using the predefined α helix and β sheet methods for secondary structure determination provided with the CONFOCHECK software.

Bacterial Two-Hybrid Assay.

Plasmids and controls were purchased from Euromedex; cloning utilized a Q5-mediated gene amplification and HiFi DNA assembly mix (NEB). Competent E. coli BTH101 cultures were co-transformed with 30 ng each of pUT18C or pKT25 plasmids using a standard heat shock protocol. LB supplemented with kanamycin (50 μg/mL) and carbenicillin (100 μg/mL) was used to select for the two plasmids. Additionally, X-Gal (40 μg/mL) and IPTG (0.5 mM) were included in the plated media to assess the interactions via reconstitution of the adenylate cyclase activity, evident by the blue colony. Plates were incubated 24 h at 30 °C and then scored for the colony color, with blue indicating a positive interaction between bait and prey fusion proteins. Incubation was continued for 2–3 days longer at room temperature to look for weak or transient interactions. Miller units were determined using the Miller assay, as previously described⁴⁴ (see Figure S3). In brief, liquid cultures with both kanamycin and carbenicillin selection were grown overnight at 30 °C. Standardized volumes of these cultures were lysed using BugBuster (Sigma), and the absorbance at 420 nm arising from the β-galactosidase cleavage of oriho-nitrophenyl-β-galactosidase was monitored. The calculation of the resulting Miller units were performed following previously published methods.⁴⁴ Independent replicates were carried out by repeating the initial transformation with independently selected clones for each construct and following the subsequently described methods.

Expression and Purification of E. coli Lon.

E. coli Lon protease (pBAD33::EcLon) was obtained from Addgene (pBAD33.Lon was a gift from Robert Sauer, Addgene plasmid # 22145; http://n2t.net/addgene:22145; RRID/Addgene_22145) and expressed and purified essentially as previously described.⁴⁵ EcLon was expressed in E. coli strain DH5α, which was grown to an OD₆₀₀ of 0.8–1 prior to induction by the addition of arabinose to a final concentration of 0.2%. Protein expression was performed at 23 °C with moderate shaking for 4–8 h prior to harvesting by centrifugation. Cell pellets were resuspended in ice-cold 100 mM phosphate buffer pH 6.4, 1 mM EDTA, 1 mM DTT, 20% glycerol and lysed using a high-pressure homogenizer (Emulsiflex). Purification was performed using a batch affinity method with phosphocellulose resin (Whatman). Washing buffers used increasing phosphate concentrations until elution with 400 mM phosphate buffer pH 6.4, 1 mM EDTA, 1 mM DTT, 20% glycerol. The eluted protein was concentrated using a 30 kDa MWCO concentrator (Amicon) and further purified by SEC on a HiPrep 16/60 Sephacryl S400 HR column (GE Healthcare) pre-equilibrated in 50 mM HEPES pH 7.5, 1 mM DTT, 1 mM EDTA, 20% glycerol.

In Vitro Degradation of the ParDE1 Complex by EcLon.

Purified EcLon activity was confirmed by monitoring the degradation of casein (data not shown), essentially as previously described.⁴⁵ A 5× Lon assay buffer was prepared composed of 125 mM Tris-HCl pH 7.5, 500 mM KCl, 50 mM MgCl₂ 10 mM DTT, 50 mM ATP. Casein protein was prepared at 4.5 mg/mL in 50 mM HEPES pH 7.5, 50 mM NaCl. 100 μL assays were incubated at 37 °C with final concentrations of 5 μM EcLon, 0.5 mg/mL casein, 1× Lon assay buffer and monitored by gel electrophoresis.

Lon degradation of both WT and I68A PaParD1 and PaParDE1 complexes was performed in both the presence and absence of an ATP regeneration system utilizing pyruvate kinase and phosphoenolpyruvate (PEP).⁴⁵ A 5× ATP regeneration assay buffer was prepared composed of 125 mM Tris-HCl pH 7.5, 500 mM KCl, 50 mM MgCl₂, 10 mM DTT, 10 mM ATP, 50 U/mL pyruvate kinase, 50 mM PEP. 100 μL reactions for both WT and I68A ParDE1 complexes (final concentration 0.5 mg/mL) were incubated with 5 μM EcLon at 37 °C, and degradation was monitored by gel electrophoresis at 30 min, 1 h, and 24 h (see Figure S4).

RESULTS

Structural Similarities of ParDE Complexes and Their Interaction Surfaces.

The structure of the P. aeruginosa PaParDE1 complex was determined by X-ray crystallography to a resolution of 1.77 Å (PDB ID: 6XRW, see Table S2). The crystallographic asymmetric unit consists of two PaParD1–PaParE heterodimers, but these are not paired into the relevant tetrameric biological unit (Figure 1A). Instead, the biological heterotetramer is generated by a crystallographic two-fold axis of symmetry and consists of a PaParD1 dimer, with each PaParD1 bound to a single PaParE1 molecule, resulting in an overall Par(E1)₁-(D1)₂-(E1)₁ architecture (Figure 1B). Within the asymmetric unit, a ParE1–ParE1 interface is bridged by a well-ordered 28-atom long polyethylene glycol (PEG) molecule (Figure 1C). This PEG molecule is associated with a small cavity formed from the interaction of two PaParD1 α3 helices and two PaParE1 N-terminal helices. This small cavity has an additional small molecule with well-defined electron density modeled as a benzamidine molecule, likely present from the initial steps of protein purification (Figure 1D). At this time, no biological role for these specific crystallization molecules is expected, although they appear essential to crystal packing.

Figure 1.

Structure of the Pseudomonas aeruginosa PaParDE1 heterotetramer. (A) The asymmetric unit of the PaParDE1 complex contains two PaParD1 chains (grey ribbons) and two PaParE1 chains (blue and cyan ribbons). The red box highlights the presence of PEG and benzamidine molecules at a ParE–ParE crystal packing interface. (B) Biologically relevant heterotetramer of the PaParDE1 complex. Regions of functional significance are noted, including the red circles highlighting a cross-interaction between a PaParE1 toxin with the other PaParD1 chain found in the dimeric antitoxin. (C,D) A composite OMIT map (contoured at 1σ) of ordered small molecules observed at the PaParE1 crystal packing interface. (C) PEG and (D) benzamidine.

Comparison of the ParD antitoxins in the four available canonical structures reveals a conservation of dimerization through the N-terminal ribbon-helix-helix (RHH) DNA-binding domain, consistent with the classification as CopG-type transcriptional repressors, but with differences in the overall lengths of the α2 and α3 helices.^{19,25,46–48} Both PaParD1 (PDB ID: 6XRW) and Caulobacter crescentus (Cc) ParD1 (PDB ID: 3KXE) have considerably shorter α2 helices (48.1 and 46.8 Å, respectively) compared to the two Mesorhizobium opportunistum (Mo) ParD structures (PDB ID: 5CEG and a chimeric ParD in PDB ID: 6X0A), which share the same DNA binding domains and have α2 helices of 53.5 Å.⁴⁶ Differences in the length of the α3 helices are also noted, with PaParD1 among the shortest at 13.3 Å; E. coli (Ec) ParDE (PDB ID: 5CW7) has the longest, measuring just over 30 Å, with CcParDE1 at approx. 17 Å; and MoParD molecules at 21 and 23 Å. This places PaParD1 as the most compact ParD antitoxin structure known to date. Further, PaParD1 appears to have an unstructured C-terminus, with three amino acids in contact with the toxin beyond the α3 helix (Figures S5 and S6).

In all structures, the long helix (α2) of the antitoxin and the short C-terminal helix (α3) form the interaction interface with the toxin. Among all the complexes, the antitoxin slots into a surface channel to wrap around the toxin molecule, burying at least 1400 Ų between the two interacting partners. The complexes from Mo and Ec, however, have a larger buried surface of ≥1500 Ų mediated largely through a longer ParD α3 helix (Table S5, Figures S5 and S6). The interactions of ParD with ParE all have a generalized hydrophobic nature of the antitoxin-binding path, a property conserved among the RelE/ParE TA family, and of note, there are no water molecules within any of these protein–protein interfaces.^19,25 There are specificity imparting polar interactions, including salt bridges, interspersed along the edges of this hydrophobic path, as was originally identified in the first reported structure (PDB ID: 3KXE).²⁵ In general, approximately half of the buried surface area is hydrophobic (Table S5, Figure S5), with the two extremes being the complex from Ec (PDB ID: 5CZE) with only ~40% of the buried area on ParD being hydrophobic and one of the Mo complexes (PDB ID: 6X0A) at ~57%.

Comparison of the structure-based sequence alignments demonstrates that the ParE toxins contain higher sequence similarity than their cognate ParD antitoxins (Figure 2A). The ParE toxins each encode a highly conserved sequence in helix 1 (consensus: AXADLDDI), of which an Asp amino acid is observed to participate in a salt bridge with a C-terminal arginine (or lysine) conserved in the cognate ParD antitoxin in each structure (Figures 2B and S6). An additional absolutely conserved residue in ParE is an arginine at the start of β2, which is observed to be relatively solvent exposed in each structure.

Figure 2.

Sequence conservation and comparison to the contacting amino acids between PaParE1 and PaParD1. Structure-based sequence alignments of (A) PaParE1 and (B) PaParD1 as compared to other ParDE complexes with experimentally determined structures.^19,23,25 Completely conserved amino acids are denoted as white text on a red background, while conservation in 3 of the 4 sequences are in red text. The ParD1 sequence is annotated by black boxes to indicate the amino acids chosen for mutation, and a red asterisk is placed above the amino acid selected for substitution with a stop codon to generate C-terminally truncated PaParD1 proteins. (C) The molecular surface of a PaParE1 toxin is colored by Kyte–Doolittle hydrophobicity (orange, phobic; green, polar). The PaParD1 antitoxin ribbon is colored tan, and amino acid side chains chosen for alanine substitution are shown and labeled. Labels with red text indicate PaParD1 truncation mutants.

The interactions between ParE toxins and their cognate ParD antitoxins place the face of the ParE β sheet against the α2 helix of ParD; this interaction appears to be supported by Tyr or Phe amino acids found on ParE β3. ParE β3 and β4 are also enriched with smaller hydrophobic amino acids, many of which comprise the core of the ParE-like fold (Figure S6). The loop in ParD that connects α2 and α3 passes over one edge of the toxin β sheet (Figure S5). The α3 helix of ParD sits against the surface generated by the two helices of the toxin (Figure S5); and as noted above, this antitoxin α3 helix is significantly shorter in PaParD1 (Figures S5 and S6). We note prominent cavities are conserved on this surface of the toxins that face the antitoxin α3 helix of ParD molecules, and the volume and distributions appear tailored for interactions with side chains from cognate ParD antitoxins (Figure S5).

While the PaParE1 toxin is tightly wrapped by the C-terminus of a PaParD1 molecule, the PaParE1 toxin also makes limited interactions with the other PaParD1 molecule in the biological unit. Two points of contact are evident, one between PaParD1 α1 with a loop in PaParE1 that connects the two α helices with the β strands in the toxin (Figure 1B, red circled region) and another with PaParD1 α2 and the C-terminus of PaParE1. Together, these previously unrecognized interactions form an interface with ~360 Ų of the buried surface per copy in the PaParDE1 complex.

Interaction with PaParE1 Has a Hotspot at PaParD1 Ile68.

The PaParDE1 complex structure was used to identify potentially important sites on the α2 and α3 helices of the antitoxin for interactions with the toxin (Figure 2). To evaluate their individual contributions, site-directed mutagenesis was employed to exchange 10 hydrophobic residues of PaParD1 (in α2: W45, V47,150, L54, in the linker region: L62, L63, and in α3: I66, I68, W72, and L74, Figure 2) to alanine, although poor expression of the antitoxin containing L54A substitution precluded further analysis. Five additional alanine substitutions were made at polar amino acids (in the linker region: D56, E64, and in α3: H65, E69, and R71, Figure 2), yielding a total of fourteen positions probed along the PaParD1 protein. These mutated antitoxin proteins were used to evaluate the kinetics of binding and thus derive an equilibrium dissociation constant (K_D) for interaction with PaParE1. These were compared to the previously published values for the WT antitoxin, with a K_D value of 152 pM (Table S3).²²

The tight interaction of these proteins approaches the limit of detection for k_off; however, thirteen of the fourteen single point mutants showed no significant difference in binding kinetics (p-value < 0.05, one-way ANOVA) from that obtained with the WT PaParD1 (traces and analysis, Figures 3 and S7, Table S3). Further, size-exclusion elution profiles of these PaParD1 point mutants in complex with PaParE1 were identical to WT (data not shown), verifying that a stable complex was formed. The point mutation at Ile68 (I68A), however, has altered the binding behavior that both weakens the interaction and no longer conforms to a 1:1 Langmuir model of binding (Figure S8). The binding is instead parsed into two binding conformations per PaParD1 chain (a 2:1 heterogeneous ligand model), where one type of interaction has a considerably faster off-rate (p-value < 0.0001) compared to the WT interaction (Figures 3 and S7, Table S3) and thus a K_D value of 626 ± 468 nM, while the other has a slower on-rate, resulting in a K_D value of 25.5 ± 2.2 nM.

Figure 3.

Mutation of antitoxin PaParD1 amino acid Ile68 to Ala decreases the interaction affinity, while no other individual mutations of antitoxin amino acids alter the affinity of interaction with PaParE. Comparison of equilibrium values for K_D, as determined by BLI experiments, reveals that PaParD1 I68A decreases the affinity and is statistically significant. The binding of PaParE1 to I68A PaParD1, however, is parsed into two different types of interactions. Removal of the dimeric region of ParD, however, restores both I68A binding interactions to within the range for WT. Kinetic on and off rates are graphed in Figure S7, average values (n ≥ 3) for each parameter are given in Table S3, and traces with residuals for each BLI experiment are given in Figure S1.

PaParD1 I68A Hotspot Has a Minimal Effect on Affinity When Antitoxin Is Rendered Monomeric.

We reasoned that additional contributions to binding may arise from some uncharacterized cooperativity within the dimeric ParD, or potentially from cross-interactions of the PaParE1 toxin with the other PaParD1 monomer in the biologically relevant tetramer noted in the crystallographic packing (Figure 1B). To address this, N-terminal truncations of PaParD1 were produced, removing residues 1–41 encompassing the DNA-binding and dimerization domains of PaParD1; however, the resulting PaParD1 42–75 construct was insoluble. Therefore, both WT and I68A PaParD1 were fused to the C-terminus of the fluorescent protein mCherry to generate mCh-ParD1_42–75 and mCh-I68A ParD1_42–75 (Figure 3). These were both confirmed to be monomeric by SEC analysis (not shown). A lack of contributions from the mCherry fusion was verified by repeating the BLI assays with mCherry fused to full length ParD1, resulting in K_D values approximating the unfused WT PaParD1 interaction profile (225 ± 6.8 pM) (Figures 3 and S1, Table S3), as well as the mCherry protein alone, which demonstrated a lack of interaction. The equilibrium dissociation constant (K_D) for the truncated monomeric mCh-PaParD1_42–75 was determined to be essentially equivalent to the WT full length protein, at 828 ± 550 pM (Figures 3 and S1, Table S3). Like the full-length I68A PaParD1 mutant, the monomeric mCh-I68A PaParD1_42–75 mutant required fitting using a 2:1 heterogeneous ligand model. Two equilibrium constants of dissociation were determined for the mCh-I68A PaParD1_42–75 mutant: K_D1 was determined to be 29.06 ± 9.39 nM, while K_D2 was determined to be 2.23 ± l.95 nM (Figures 3 and S1, Table S3). While these values are still increased relative to the WT values, they are not statistically significant. However, the K_D1 value of mCh-I68A ParD1_42–75 is significantly different from the full-length I68A ParD1 mutant. This indicates that the impact of the I68A mutant on affinity is likely most attributable to altering the nature of the interactions within the heterotetramer rather than the loss of specific atomic contacts.

C-Terminal Seven Amino Acids of PaParD1 Are Required to Form a Complex with PaParE1.

Concomitant with our analysis of individual amino acid contributions to the interaction affinity, C-terminal truncations of the antitoxin were made to determine if it was required to establish a stable complex with PaParE1. Using site-directed mutagenesis, stop codons were created to replace residues N58, L63, or E69 (Figure 2). Based on BLI experiments and confirmed by SEC, none of these truncation mutants were able to form a complex with WT PaParE1 (Figure 4). For the truncation at amino acid 58, this was not surprising, as all of the last interacting helix were removed (α3, Figure 2). However, the abolishment of interaction with a truncation at amino acid 69 removes only a minimal part of the last helix. Of the truncated amino acids of E69X PaParD1, W72 and L74 form hydrophobic interactions with PaParE1. Additional contacts within this truncated region of PaParD1 are a polar interaction between E69 of ParD1 and K41 of PaParE1, as well as a polar interaction between R71 of PaParD1 with D14 of PaParE1. Single point mutations of each of these antitoxin residues had no impact on affinity by BLI (Figures 3 and S1, Table S3).

Figure 4.

SEC confirms a lack of interaction with PaParD1 antitoxin lacking the C-terminal seven amino acids. The black trace demonstrates the resulting profile when PaParE1 is mixed with either (A) WT or (B) truncated (E69X) PaParD1. Corresponding traces for either component alone are also shown.

Altered Interactions with PaParD1 Are Recapitulated in Bacterial Cells.

To assess the potential impacts of other cellular contents on the interactions of PaParD1 and PaParE1, the I68A and E69X constructs were generated for testing in a bacterial two-hybrid (BACTH) system. Each construct contained either the T25 or T18 fragments of the adenylate cyclase protein at the N-terminus of PaParD1 (WT, mutant, or truncation) and/or of PaParE1 (Figure 5). As we previously noted,²² the PaParE1 toxin is attenuated with respect to toxicity, which allows for expression in the absence of antitoxin in the BACTH system. Each construct was tested by spotting onto media with a colorimetric indicator (X-Gal) and by measurement of the Miller units of β-galactosidase activity resulting from an overnight induction. To ensure the detected interactions were protein-specific, each construct was also screened against an “empty” vector control, and this routinely yielded white colonies, as well as in the opposite vector orientation (fused to either T18 or T25 fragments) (Figure S3).

Figure 5.

A bacterial two-hybrid assay confirms the effects of PaParD1 mutations on complex formation within a cellular environment. Representative images of growth on indicator media are shown, and corresponding Miller units are indicated below each spot. These represent at least three independent transformations with subsequent measurements after an overnight induction. Additional measurements and control samples are given in Figure S3.

BACTH revealed that WT PaParD1 interacts strongly with PaParE1, evident by blue colonies and Miller units of 3600–4000, as expected for an in vitro apparent K_D of 152 pM. The interaction of I68A PaParD1 with WT PaParE1 results in a lighter blue colony than the WT interaction. Quantitation of Miller units reveals an approximately 2- to 4-fold decrease of the PaParE1 interaction with I68A PaParD1, as compared to the WT PaParD1 interaction with PaParE1 (Figures 5 and S3), which is consistent with the biophysical measurements. This demonstrates that the observed in vitro impact on binding of mutating amino acid Ile 68 of PaParD1 has an equivalent impact within a cellular environment, albeit in E. coli rather than the native host P. aeruginosa.

The lack of complexation between PaParE1 and the E69X PaParD1 antitoxin was verified in the BACTH assay format, which revealed white colonies and no Miller units above background (Figures 5 and S3). Therefore, the interactions of the E69X PaParD1 antitoxin truncation mutant with PaParE1 are also blocked within this cellular environment. Further, these results indicate that the last seven amino acids lacking in E69X PaParD1, which removes all amino acids after the Ile68 amino acid, are required to initiate the interaction with PaParE1.

PaParD1 “Hotspot” Mutation I68A Affects the Secondary Structure and Stability of the Resulting PaParDE1 Complex.

To further probe the impact of the PaParD1-I68A mutation’s effect on the antitoxin structure and complexation, we determined the secondary structures using FTIR.⁴⁹ Secondary structure measurements were performed for each component, as well as purified WT and I68A-containing complexes (Figure 6A, Table S4). In the absence of PaParE1, both WT and I68A PaParD1 have essentially identical secondary structure, containing approx. 35–36% α-helical and 28% β-sheet content, indicating that the I68A mutant does not impact the native structure. Upon binding to PaParE1, the WT PaParD1 undergoes a transition, resulting in an increase in α-helical content wherein each PaParD1 monomer contributes six amino acids, and this is consistent with a disorder-to-order transition at the C-terminal α3 helix of PaParD1 (Figure 6A, Table S4). Interestingly, upon complexation with PaParE1, the PaParD1-I68A mutant shows a larger increase in α-helical content relative to the WT complex (approx. 14 amino acids per PaParD1 monomer), indicating that it adopts an extended α-helical structure upon toxin binding. While we cannot preclude changes to the secondary structure of the ParE molecule, as was previously shown in the structurally related RelE toxin from E. coli, ParE toxins lack the region corresponding to the RelE gain of the helical content, and no available structures or models for association support such a change in the ParE structure.⁵⁰ The PaParD1 antitoxin is noted to contain a short β strand (L62, L63, E64); however, due to the limitations of FTIR only measuring β sheets, we cannot speculate whether this region is structured in the absence of or induced by the toxin. Given that few amino acids in the WT complexed PaParD1 antitoxin are not in an α helix, or part of the well-ordered RHH motif, it appears that the additional helical nature of the I68A antitoxin comes from the loop region between α2 and α3, as well as the short β strand (L62, L63, E64).

Figure 6.

The PaParD1 I68A mutation impacts the gain of the secondary structure as a function of binding to PaParE1 and results in altered occupancy of PaParE1 within the overall complex. (A) FTIR measurements were carried out using buffer-matched samples and repeated in triplicate (see Table S5 for individual measurements). (B) SEC–MALS analysis of complex formation with I68A antitoxin and compared to complexes containing WT PaParD1 antitoxin. The absolute molecular mass (right y-axis, dotted lines overlayed with peak traces) of the complexes were determined in the presence of excess PaParE1 toxin to ensure saturation. Peak traces are plotted from the refractive index (dRI) signal. Replicate measurements are shown in Figure S2.

We verified the association of the I68A PaParD1 mutant with PaParE1 using SEC, which indicates I68A is able to form a complex with PaParE1 in vitro. However, a small shift was observed in the elution volume for the I68A containing complex, as compared to the WT complex, suggesting a potential change in interaction (Figures 6B and S8). The data fitting results of the BLI experiments also suggest a change in binding for the PaParD1-I68A antitoxin (Figures 3 and S1), including faster off-rates resulting in weaker affinity. To further measure the impacts on the overall structures, we utilized SEC coupled to MALS to determine the absolute molecular mass of the components and the resulting complexes (Figure 6B).

We have previously characterized the PaParDE1 system, identifying a dimeric ParD antitoxin (14.74 ± 1.30 kDa, no affinity tag present) and monomeric toxin (13.5 ± 0.64 kDa) which form a heterotetrameric complex (44.5 ± 1.10 kDa).²² The SEC–MALS results recapitulated these previous measurements for monomeric PaParE1 (13.65 ± 0.18 kDa) and dimeric PaParD1, including for the PaParD1-I68A mutant (22.37 ± 0.38 kDa, with affinty tag, Figure 6B).²² In addition, a heterotetrameric complex of the WT proteins was measured as 47.15 ± 0.16 kDa, also consistent with previous measurements.²²

However, the PaParE1 toxin and the PaParD1-I68A mutant antitoxin complex consist of a 38.43 ± 0.38 kDa complex. The light scattering signal clearly indicates a shift in the molecular mass as compared to the WT complex and is interpreted as comprising a mixture of dimeric I68A antitoxin bound to either one or two PaParE1 toxin molecules. The loss of binding to one toxin results from the increased off-rate, which can reequilibrate as the chromatography progresses. This provides orthogonal BLI and SEC–MALS supporting an impact on the loss of affinity between the PaParD1 antitoxin and its cognate PaParE1 toxin, likely resulting from altered dynamics and C-terminal antitoxin ordering (measured by FTIR) due to the I68A mutation.

Conservation of PaParD1 Ile68 Interaction Site on the Surfaces of ParE Toxins.

Of note, this hydrophobic amino acid Ile68 sits in a well-defined hydrophobic cavity on the surface of PaParE1 (Figures 2C and S9). The I68 binding pocket is formed by the ParE1 residues Trp5, Ala10, Leu13, and Val37 (Figure S9). This binding pocket is conserved in two of the other four published structures (PDB ID: 5CZE, 5CEG) (Figure S9A,C) and contains conserved hydrophobic interactions. The E. coli ParDE complex (PDB ID: 5CZE) contains an additional conserved binding pocket that accommodates a conserved histidine residue (PaParD1_His65) (Figures 2C and S9A). In a recently posted description of the ParDE2 complex from V. cholerae, this same pocket is also highlighted as a conserved feature, although coordinates are not yet available to permit a more detailed comparison.⁴⁸ The I68 binding pocket on PaParE1 is flanked by two additional well-defined cavities that accommodate the ParD1 residues W72 and H65, neither of which have an apparent role in complex formation as evident by a lack of impact on binding when mutated to alanine (Figures 2B, 3 and S9, Table S3).

Altered TA Interaction Increases Accessibility for Lon-Mediated Antitoxin Degradation.

A pervasive feature of functional models for type II TA systems is the degradation of the antitoxin protein by the AAA⁺ ATPases Lon and ClpP.^2,3,13 However, it has been questioned if this degradation is of only the free antitoxin or if antitoxin could also be pulled from a tightly interacting complex with toxin.⁵¹ We further extended this question by testing if the Lon protease from E. coli (EcLon) was able to degrade PaParD1 in vitro when it was complexed to the PaParE1 toxin, and further, if the PaParD1-I68A mutant with weakened interaction affinity could be more readily degraded.

When the uncomplexed “free” PaParD1 antitoxins (WT or I68A) are incubated in the EcLon reaction, they are degraded almost completely by 24 h and appear to have the same kinetics of degradation (Figure 7A, blue lines). When a purified complex of PaParDE1 was incubated with EcLon, degradation of the PaParD1 component was evident (Figure 7A, green line), but only when an ATP regeneration system was also included in the reaction. This indicates that the relatively slow kinetics of in vitro degradation likely arise, in part, due to autoinhibition of EcLon by ADP,⁵² while under more ideal in vitro conditions, EcLon can access the antitoxin despite its tight interaction (Figure 7). When the PaParD1 I68A mutant is used in place of the WT antitoxin in the TA complex, the degradation kinetics increases such that by 24 h almost all of the antitoxin has been consumed (Figure 7, magenta line, Figure S4). No evidence of degradation of the PaParE1 toxin is observed when bound to either WT or I68A PaParD1 (Figures 7 and S4). This demonstrates that binding to toxin protects the antitoxin from degradation, such that a higher off rate (resulting from weaker affinity) such as noted for the antitoxin I68A mutation greatly decreases protection and increases the susceptibility of this antitoxin to degradation.

Figure 7.

The weakened affinity and altered interactions between PaPar1 and PaParD1 I68A permit a more efficient degradation by EcLon in vitro. (A) The PaParE1 toxin is not degraded, while the free antitoxins (WT or I68A) are readily degraded. Complexation of antitoxin with toxin, however, offers protection from EcLon cleavage. This protection is directly correlated with the interaction affinity, as the I68A mutant is depleted much more quickly than the WT antitoxin. (B) Stained gels of in vitro EcLon digestions used for (A). Replicate gel images are provided in Figure S4.

DISCUSSION

Toxin Surface Is Topologically Matched to the Cognate Antitoxin, Including a Conserved Central Cavity Comprised of Toxin Helices 1 and 2.

A comparative analysis of the available ParD–ParE interfaces reveals diversity throughout the cognate interactions (Figures 2, S5 and S6). Among these, however, the ParD antitoxins bind to a generally hydrophobic canyon with some specific polar interactions at the periphery of this canyon. These polar interactions include a highly conserved salt bridge between an acidic amino acid in the first helix of the toxin with a basic side chain at the C-terminus of the antitoxin. There are topological features on the ParE toxins that appear to complement the shape of hydrophobic side chains at the most C-terminal part of the cognate ParD antitoxins. A conserved surface cavity on the PaParE1 toxin surface is the binding site for the Ile68 antitoxin residue, which we demonstrate is critically important for both induced structural changes and initiation of binding of the PaParD1 to PaParE1. In addition, we identified a cross-interaction of the PaParE1 toxin with the antitoxin chain of the PaParD1 dimer that is not wrapped around the extensive hydrophobic interface. These interactions are facilitated by a shortened α2 helix which allows for contacts between the RHH motif of one antitoxin with the C-terminal loop of the opposite ParE toxin. While the in vivo impact of these specific interactions is not clear, removal of the dimeric nature of the ParD antitoxin, and thus the potential for these interactions to form, lessens the impact of the I68A mutation on affinity while maintaining the presence of two distinct types of binding.

Kinetics of Interaction Support a Disorder-to-Order Transition upon Binding and Suggest How a Form of Conditional Cooperativity May Rely on Dimeric Antitoxin Within a TA Complex.

The association rates for the PaParE1 toxin with the PaParD1 antitoxin and mutants thereof are relatively slow, consistent with a conformational change at the antitoxin C-terminus during association.⁵³ For related types of effector immunity pairs, such as used in type VI secretion systems, they typically have comparable overall affinity to the ParDE systems but with much faster on-rates; this may relate directly to the functional need to quickly and tightly interact with cognate toxins as protection from assault.^4,54 In contrast, TA systems must essentially be initiating interactions co-translationally.^3,6 Our data further support this model by localization of the “seeding” portion of interaction to the C-terminal of the antitoxin and the N-terminal of the toxin, which follows the order of the encoding operon. It is the folding of the antitoxin C-terminus upon binding that results in the high shape complementarity (CSS value) for the TA complexes, calculated as ≥0.9 for the ParDE structures, and where a value of 1.0 indicates a perfect fit.⁵⁵

The WT interaction has very slow, to essentially unmeasurable, off rates of binding. In earlier studies, we were able to capture this rate for the WT antitoxin protein likely due to a now recognized intrinsic degradation of purified PaParD1.⁵⁶ Some minor content of degraded PaParD1 in the reaction would facilitate changes in binding kinetics; however, the tight interaction in either situation is in the approx. 150 pM K_D range. While essentially non-exist off rates favor the model of irreversible interactions, such that an antitoxin is unlikely to be degraded from a complex with its cognate toxin,⁵¹ the I68A antitoxin point mutation is able to reduce affinity from pM to nM by altering the induced structural transition. Further, this effect is only present in a dimeric ParD antitoxin, as rendering the antitoxin monomeric essentially reduces the effect to a non-significant modest change. A reasonable interpretation is that the binding of a toxin molecule to one of the chains of a ParD dimer impacts the ability of the other antitoxin chain to undergo a structural transition and thus weakening the binding to the second ParE toxin. This highlights that external factors that perturb the interaction could feasibly alter or tune the interaction strength by similarly impacting this transition required for tight association.

While the PaParE1 Toxin Protects the PaParD1 Antitoxin from Degradation, the Kinetics of Their Interaction Ultimately Dictates Antitoxin Accessibility to Protease Digestion.

The altered conformations upon binding imparted by the I68A mutation is also strongly supported by our in vitro Lon protease degradation assays, which demonstrates an increased removal of the I68A PaParD1 antitoxin from complexation with toxin, relative to the WT antitoxin. Under these idealized in vitro conditions, PaParD1 was eventually digested even in the presence of PaParE1 toxin. It is tempting to speculate that the protease system was powerful enough to pull apart the picomolar interaction with the PaParE1 toxin, and more so that the weakened interaction with I68A was more easily parted. However, a simplified calculation using only the measured off rates (ignoring reassociation, which must be occurring, albeit slowly) and the number of molecules in the reaction results in theoretically essentially full removal of even WT antitoxin over the 24 h incubation time. The altered binding of the I68A antitoxin mutant, then, becomes critically important to the interpretation of this assay. While we note altered on rates and differences in the secondary structure associated with a disorder-to-order transition during association, the altered conformations adopted upon binding are also responsible for faster off rates, resulting in less toxin-mediated protection for I68A. Therefore, the increased degradation of the I68A antitoxin most likely results from the impact on the disorder-to-order transition that propagates to faster off-rates for this interaction.

It has been previously reported that toxin molecules can protect their cognate antitoxins from degradation, although the DinJ–YafQ system appears to be an exception.²⁷ Protection was identified in the kis–kid system and in the ParDE system from the RK2 plasmid, both of which are targeted for degradation by the ClpAP protease.^9,37 Interestingly, for the RK2 ParDE system, the chaperone ClpA could bind to the antitoxin even when it was complexed with toxin, and over longer time scales the antitoxin could be degraded,⁹ very similar to what we observe for PaParDE1. For RK2 ParD, signature sequences for ClpA targeting were identified; some weak similarity may be present in the PaParD1 sequence but are located in the DNA binding region (not shown).⁹ Another difference in these ParD molecules is that the RK2 ParD toxin binding region was demonstrated to be highly unstructured (PDB ID: 2AN7),⁵⁷ as is the ParD2 from V. cholerae,⁴⁸ in contrast to our measurements of the secondary structure for PaParD1 that indicate a modest gain of α-helical content (6 amino acids out of 75 total) upon complexation with PaParE1. The ParD from the CcParDE1 complex was also noted to be structured in the absence of ParE.²⁵ This intrinsic feature of ParD antitoxin may dictate the recognition and subsequent binding by specific proteases in the cell. The protease that natively degrades the PaParD1 antitoxin has not been identified, and we cannot exclude recognition by multiple proteases as found for the tripartite paaA–ParE systems in E. coli.²³ Our experiments using the Lon protease from E. coli, therefore, may not represent the kinetics of the native cellular environment but do support a direct correlation with toxin protection and thus TA affinity.

Molecular Model for ParD Antitoxin–ParE Toxin Interactions.

For PaParDE1, we have determined that the last seven amino acids are critical to initiate the interaction, although when we individually mutate each of the contacting residues to Ala in the context of the full length PaParD1, they did not significantly impact complexation or affinity. Within this region is a conserved electrostatic interaction between ParE α1 (Asp14) and ParD α3 (Arg71). We systematically mutated each of the contacting positions of PaParD1 in these C-terminal seven amino acids (E69A, R71A, W72A, and L74A), and no single point mutant affected the kinetic rates or overall affinity. Therefore, the loss of association upon C-terminal truncation of PaParD1 must be from the loss of multiple required interactions. It is striking, however, that the remaining 20 sites for interaction in this truncation mutant were not able to mediate any association, even at a weakened affinity.

Overall, it is clear for the PaParDE1 complex that the C-terminus of PaParD1, including part of α3 and the most C-terminal coil, initiates the association reaction. A similar requirement for the antitoxin C-terminus in assembly was previously identified for the DinJ–YarQ system from E. coli.²⁷ For the unique truncated antitoxin, called paaA, which is comprised only of the equivalent of a truncated α2 and the connecting loop with α3, interaction specificity was found to be encoded in the C-terminal α3.²³ Other studies have examined the specificity of interactions in ParDE systems through the lens of protein sequence co-evolution.^19,46 Critical positions for specificity in the ParD antitoxin of two M. opportunism systems were identified in the last four amino acids of α2 as well as in the unstructured linker.^19,46 Extrapolating from these previous publications and our current results, then, we reason that in a cellular setting, the PaParD1 antitoxin must be essentially completely translated prior to any interaction with the PaParE1 toxin.

This leads us to suggest that the PaParD1 interaction with PaParE1 undergoes a disordered-to-ordered sequential binding, initiated perhaps by the noted conserved electrostatic interaction within the C-terminal seven antitoxin residues. The interaction is stabilized or further directed by other contacts with the C-terminal seven amino acids that culminate at the hotspot position Ile 68. This site appears to serve as a “positioning factor”, concomitant with the induced helical formation for the antitoxin C-terminus. Given the tight interaction of WT proteins, such an order or cooperativity of antitoxin binding to toxin may be hidden within the speed of association. By diminishing the interaction strength, we have slowed the association and allowed capturing of these discrete events. Specifically for PaParD1, our data suggest that (1) the interaction is seeded by the C-terminal antitoxin amino acids, although the contribution of each individual contact is limited; (2) Ile68 is a critical position for interaction with a conserved toxin surface topology, triggering a subsequent gain in the secondary structure, whereas a less ideal contact with toxin perhaps allows more torsional freedom to gain additional secondary structure for PaParD1; (3) antitoxin α2 is secondary to the interaction affinity and is not involved in initiating the interaction or in the disorder to order transition; and (4) the dimeric nature of the antitoxin exacerbates the requirement for precise interactions associated with the disorder to order transition.

The identification of a hotspot for TA interactions demonstrates that vulnerable sites for interfering with preformed complexes exist, such that a small molecule could be designed to intercalate into the identified toxin crevices and artificially weaken the interaction of complexes. The current work contributes to an overall model of type-II TA system complex formation and subsequent stability, which directly correlates to protection from proteases, and that is contingent on the antitoxin C-terminus. Further, the high shape complementarity between cognate ParDE pairs combined with the essential role indicated for the equivalent of antitoxin I68 position permits the prediction of interactions of noncognate pairs, an area of active and on-going interest in TA system functionality.¹⁴

DATA AVAILABILITY

All data are included in this manuscript and supplement document; DNA clones for constructs made in this work are available upon request.