ClpAP protease is a universal factor that activates the parDE toxin-antitoxin system from a broad host range RK2 plasmid

The activity of type II toxin-antitoxin systems (TA), which are responsible for many important features of bacterial cells, is based on the differences between toxin and antitoxin stabilities. The antitoxin lability results from bacterial protease activity. Here, we investigated how particular Escherichia coli cytosolic proteases, namely, Lon, ClpAP, ClpXP, and ClpYQ, affect the stability of both the toxin and antitoxin components of the parDE system from the broad host range plasmid RK2. The results of our in vivo and in vitro experiments show that the ParD antitoxin is degraded by the ClpAP protease, and dsDNA stimulates this process. The ParE toxin is not degraded by any of these proteases and can therefore cause growth inhibition of plasmid-free cells after an unequal plasmid distribution during cell division. We also demonstrate that the ParE toxin interaction with ParD prevents antitoxin proteolysis by ClpAP; however, this interaction does not prevent the ClpAP interaction with ParD. We show that ClpAP protease homologs affect plasmid stability in other bacterial species, indicating that ClpAP is a universal activator of the parDE system and that ParD is a universal substrate for ClpAP.


Results
ClpAP protease is responsible for ParD protein degradation. Since the stability of ParD and the potential protease responsible for its degradation have not been clarified, we conducted ParD stability tests. First, the stability of the ParD protein was verified in vivo in E. coli C600 cells. The plasmid pBAD24_ParD was used, which allows the arabinose-dependent expression of parD. One hour after induction, the protein translation was inhibited by the addition of tetracycline. Samples were collected at selected time points. After 120 min, a significant protein loss was observed by Western blot analysis with anti-ParD antibodies (Fig. 1A). E. coli strains with protease gene inactivation, lon(−), or AAA+ protease subunits clpA(−), clpX(−), and clpY(−) were also used for the analysis of ParD stability. We observed a significant (P value 1.3 × 10 −3 )(see Supplementary Table S1) increase in the stability of the ParD protein in the clpA(−) strain (Fig. 1B). This result indicates that the ClpAP protease is responsible for the efficient proteolysis of ParD.
We also analyzed whether the ParD protein contained known ClpA substrate recognition motifs. The ParD sequence analysis revealed that, indeed, ParD contains two putative motifs for ClpA recognition (see Supplementary Fig. S1). One motif is located in the protein C-terminal part and the other one, which is SsrA-like, is located closer to the N-terminus.
ParE protein is stable in host cells. The stability of the ParE protein was also tested. In vivo analysis showed that the ParE protein was stable after 120 min (Fig. 1A). ParE stability was also analyzed using in vitro proteolysis experiments with Lon, ClpAP, ClpXP, and ClpYQ proteases (see Supplementary Fig. S2). No degradation was observed after 120 min. DNA stimulates ClpAP protease to degrade ParD. The ClpA unit of the ClpAP protease is responsible for recognition of a substrate, ATP-dependent substrate unfolding, and its translocation into the central channel of the ClpP proteolytic subunit 42,43 . Previous results have shown that the ClpAP protease is able to bind DNA. Moreover, ClpAP-DNA interaction increases its ATPase activity and the efficiency of substrate proteolysis 44 . We tested whether DNA affects ParD proteolysis. In vitro experiments were performed using the Lon, ClpAP, ClpXP and ClpYQ proteases. A significant decrease in the amount of ParD protein (using SDS-PAGE analysis) was observed when ClpAP protease was in the reaction mixture ( Fig. 2A). We also noticed that the presence of supercoiled plasmid DNA increased the efficiency of the proteolysis ( Fig. 2A). To further analyze the influence of DNA on ParD degradation, various DNA variants were used containing or lacking the promotor sequence pparDE (ParD protein binds a site in the parDE operon promoter 30 ) and two DNA forms: supercoiled or linear (see Supplementary Fig. S3). In all cases a significant increase in proteolysis was observed. This suggests that ParD is efficiently processed by ClpAP in the presence of DNA regardless of the pparDE sequence or DNA form. To determine the dynamics of ParD degradation by ClpAP, a time-course in vitro proteolysis assay was performed (Fig. 2B). The analysis showed that approximately 50% of the antitoxin level decreased in less than 5 min. in the presence of DNA. The experiment was also performed in the absence of DNA, and the result showed that the time SCIentIFIC REPORTS | (2018) 8:15287 | DOI:10.1038/s41598-018-33726-y required to reach fifty percent was tripled. For statistical data analysis see also Supplementary Table S2. These results suggest that ParD degradation is rapid and is stimulated by DNA.
To test whether and how the preformation of complexes (protease, DNA, and/or ParD) affect the processing of ParD by ClpAP, we performed an experiment in which the addition order of proteolytic components was changed. In each case, DNA stimulated the degradation of the substrate; however, when the ClpAP protease was preincubated with DNA before the substrate ParD was added, the stimulation was the highest (Fig. 3). This sequential assembling of the proteolytic complex suggests that the ClpAP protease is stimulated by DNA to process the substrate.
We also tested whether DNA could stimulate ParE degradation. The presence of DNA in the reaction mixture did not result in ParE proteolysis in vitro by Lon, ClpAP, ClpXP or ClpYQ proteases (see Supplementary Fig. S2).

ParD stability increases when it is bound to ParE.
To investigate the effect of the ParE protein on the proteolysis of the ParD protein by the ClpAP protease, we performed a series of in vivo and in vitro experiments. The E. coli C600 strain was transformed with plasmid pAS12, which encodes the GST-parE gene (expression In the wild-type E. coli strain, the stability of ParD and ParE proteins was analyzed after inhibition of translation by tetracycline of ParD-(pBAD24-ParD) and ParE-(pRR46 and pAS12) overproducing cells. The assay was performed as described in Methods. (B) Cells of E. coli wt (•), clpA(−) (▼), clpX(−) (Δ), clpY(−) (○), and lon(−) (■) harboring plasmid pBAD24-ParD for overproduction of ParD were used. Samples were taken from the cultures at the indicated time points after the addition of tetracycline and were analyzed for ParD presence. The assay was performed as described in Methods. Full-length blots are included in the Supplementary Information file (Fig. S11). Each experiment was repeated three times, and the mean values with standard deviations (error bars) are presented as graphs. Numeric data and P value are shown in the Supplementary Table S1.
SCIentIFIC REPORTS | (2018) 8:15287 | DOI:10.1038/s41598-018-33726-y induced by IPTG), and the pRR46 plasmid for the constitutive expression of parD. Overnight cultures were diluted 1:100 in fresh LB broth, and two parallel cultures were grown at 37 °C to an OD 600 of 0.6. At this point, expression of ParE was induced in one culture, and the second culture was without IPTG induction of parE gene expression. One hour after the induction, the protein translation was inhibited by the addition of tetracycline. Samples were collected at selected time points and analyzed as described in Methods. The analysis was carried out using anti-ParD and anti-ParE antibodies. When the ParE toxin was overproduced in host cells, the stability of the ParD antitoxin increased (Fig. 4A, statistical analysis Supplementary Table S3). Taking into account the size of the ParD and ParE proteins and the sensitivity of the antibodies used, the estimated ParD:ParE stoichiometry in the reaction after tetracycline addition (time 0) was 1:1, indicating that ParD-ParE could assembled as a heterotetramer. In the E. coli clpA (−) strain, we did not observe any significant differences in the stability of the ParD antitoxin regardless of induction of parE expression (see Supplementary Fig. S4). In the in vitro proteolysis assay, an increasing concentration of ParE protein was used in the reaction mixtures containing ParD and ClpAP. SDS-PAGE followed by Coomassie brilliant blue staining showed that proteolysis of the ParD protein by the ClpAP protease was less efficient with increasing amounts of ParE (Fig. 4B). The estimated ParD:ParE stoichiometry of the reactions was 1:0.03 up to 1:1 and was comparable to that obtained in the in vivo tests (see Fig. 4A). The ParE toxin had no effect on the proteolysis of the other substrates by ClpAP protease (see Supplementary  Fig. S5). In a control experiment we also showed that the GST-tag had no effect on ParD proteolysis by ClpAP (see Supplementary Fig. S6).
By utilizing SPR, we performed ClpA, ParD and ParE interaction analysis. The obtained data showed that ClpA, the ATPase unit of ClpAP protease, is responsible for the recognition of the ParD substrate (Fig. 5A), while the ParE toxin is not recognized by ClpA (Fig. 5B). To analyze how the presence of ParE toxin affects the interaction of ParD with ClpA, we injected preincubated ParD and ParE proteins onto ClpA, which was immobilized on a sensor chip. A fixed concentration of 200 nM ParD was incubated with increasing concentrations of ParE (Fig. 5C). The SPR analysis results showed that the potentially created ParD-ParE heterotetamer was able to interact with the ClpA protein. To test whether the pre-formed ClpA-ParD complex might interact with the ParE toxin, we performed an injection of ParE onto the preformed complex of ClpA-ParD on the sensor chip. An increase in response was observed when ParE was injected onto the complex (Fig. 5D). These results indicate that ParE can bind to the ParD-ClpA complex by interacting with ParD. However, if ClpA interacts with another substrate such as TrfA, the ParE toxin does not interact with this complex (see Supplementary Fig. S7). Considering data obtained from SPR and in vivo and in vitro tests (see Fig. 4) it must be pointed out that formation of ParDE complex does not prevent ParD interaction with ClpA but it protects ParD against proteolysis (Fig. 5E).
The ClpAP protease is responsible for activating the TA parDE system of the RK2 plasmid in diverse bacterial species. Plasmid RK2 is present in a number of bacterial species. For this reason we tested whether the ClpAP protease homologs also function as activators of the parDE system in other species of bacteria. For this purpose, the operon parDE from the RK2 plasmid was cloned into pBBR1MCS-5, a broad host range plasmid lacking any TA system, as described in Methods. Plasmid pABD6-1 (pBBR1-parDE) was used to transform wild-type and clpA-deficient strains of E. coli, P. putida, and C. crescentus. Plasmid pBBR1MCS-5 was used as a negative control. The maintenance of plasmid pABD6-1 in wild-type E. coli (Fig. 6A), C. crescentus (Fig. 6B) and P. putida (Fig. 6C) cells was on average at the level of 75% after 150 generations compared to the complete loss of the control plasmid. However, in strains with inactive clpA genes, no statistically significant difference in the stability of the plasmids was observed (Fig. 6, Supplementary Table S4). These results clearly indicate that the ClpAP protease is responsible for maintaining the plasmid that contains the parDE system in different bacterial species. The lack of a functional ClpA unit most likely results in a significant increase in the stability of the ParD antitoxin (as shown in Fig. 1B), thus preventing the toxin activity. We also performed stability tests of plasmids pABD6-1 and pBBR1MCS-5 in the E. coli strain with inactivation of the clpX gene. No differences were observed compared to the wild-type strain (see Supplementary Fig. S8). To analyze the universality of recognition and degradation of the ParD antitoxin by ClpAP, both the ClpA ATPase and the ClpP protease from the bacterial species that host the RK2 plasmid were purified and analyzed for in vitro proteolysis (Fig. 7). The results of the in vitro proteolytic tests clearly indicated that the ParD antitoxin is degraded by ClpAP proteases from E. coli, C. crescentus and P. putida. Notably, in the case of C. crescentus and P. putida proteins, an increase in proteolytic efficiency significantly depends on the presence of DNA in the reaction mixture.

Discussion
Cellular proteases appear to be the key players activating multiple TA systems via degradation of the proteinaceous antitoxins 45 . Antitoxin stability has been examined in many E. coli TA systems using in vivo analysis, yet little is known about antitoxin stability in other bacteria species. In this work, we have evaluated the degradation of the ParD and ParE proteins of the plasmid RK2 type II TA system by different AAA+ cytosolic proteases of E. coli, C. crescentus and P. putida. The in vitro data showed that ClpAP proteases degrade the ParD antitoxin, and in vivo tests confirmed these results. In contrast to ParD, the ParE toxin was unaffected by all proteases tested. The results obtained are in agreement with the literature data regarding other type II TA systems 46-48 . This implies a differential stability of ParD and ParE proteins that is the key factor to explain the selective activation of the parDE system after plasmid loss. Reaction components ParD (marked as D), pKD19L (marked as DNA) and protease complex subunits ClpA (marked as A) and ClpP (marked as P) were mixed together in different orders. Mixtures containing two or three components were initially preincubated for 5 min, then the remaining components were added and the reaction continued for 30 min (lane 5-9). After a 5 min pre-incubation step, no components were added and the mixtures were incubated for 30 min (lane 3 and 4). The control reaction was stopped after 5 min of preincubation (lane 2). Molecular weight marker (lane 1). The most efficient degradation reaction is marked with an arrow (↓). The assay was performed as described in Methods. Full-length gels are included in the Supplementary Information file (Fig. S13). Each experiment was repeated three times, and the mean values with standard deviations (error bars) are presented as graphs. The experiment was carried out for 120 min. The assay was performed as described in Methods and analyzed with SDS-PAGE, followed by Coomassie brilliant blue staining. Given values are means from three independent repeats of each experiment. Full-length blots and gels are included in the Supplementary Information file (Fig. S14). Our in vitro proteolytic assays showed that ParD degradation is rapid; a major reduction of ParD was observed in less than 30 min. We estimated that ParD is degraded by ClpAP with a half-life of approximately 15 min. DNA stimulates this process giving a half-life of approximately 5 min, which is similar to the degradation of the Kis antitoxin by ClpAP 47 . The results of the experiments investigating the order of addition suggest that the observed increase in proteolytic efficiency is the result of direct stimulation of the ClpAP protease by DNA. Our data also show that ClpA ATPase is able to bind to ParD, but not ParE, and the stability of ParD is much higher when it interacts with ParE. Previous data has shown that ParD and ParE create a ParD 2 -ParE 2 complex 29,34 where toxicity   E.coli (lanes 3 and 4), P. putida (lanes 5 and 6) and C. crescentus (lanes 7 and 8) was performed in the presence (lanes 4, 6 and 8) or absence (3, 5 and 7) of DNA. In a negative control reaction, no protease was added (lane 2). Molecular marker (lane 1). The assay was performed as described in Methods. Full-length gels are included in the Supplementary Information file (Fig. S15). Each experiment was repeated three times, and the mean values with standard deviations (error bars) are presented as graphs.
SCIentIFIC REPORTS | (2018) 8:15287 | DOI:10.1038/s41598-018-33726-y of ParE is inhibited 36 . When both TA components are present in the host cell, the toxin protects the antitoxin from degradation. The potential cause of the ParD stabilization by the ParE toxin could be the structural change of the ParD antitoxin in complex formation. The structure stabilization of the intrinsically disordered ParD C-terminal due to interaction with ParE has been reported 34 . If the structure change is a reason for preventing ParD protein processing, it must be noted that it does not prevent its recognition by ClpAP. This is evidenced by the results showing that, regardless of whether ParD is associated with ParE, it is recognized by the ClpA subunit. It is not clear if the unstructured part of ParD or a specific motif is recognized by ClpA. The ParD antitoxin sequence alignment with sequences of known ClpA substrate recognition motifs revealed two putative recognition sites, including one motif similar to the E. coli SsrA recognition sequence. It would be interesting to test whether this is a universal signal for ParD recognition by ClpA homologs. It has been proposed that the N-terminal domain of ClpA facilitates an early binding step, contributing to the specific recognition of substrates for processing 49 . Although the conserved basic residues in the ClpA N-terminal domain were proposed to regulate the processing of several substrates, the molecular mechanism of substrate recognition remains unclear 49 . The mutation of two conserved arginine residues flanking a putative peptide-binding groove within the N-terminal domain of ClpA, specifically compromised the ability of ClpA to unfold and degrade selected substrates, but did not prevent substrate recognition 49 . It is possible that the recognition mechanism is identical among ClpA homologs, which would explain the universality of ParD recognition in various bacterial species. It is also very likely that not all ClpAP substrates would be universally recognized and processed by the homologous ClpAP proteolytic systems, and ParD as a protein of a broad host range replicon evolved as a universal ClpAP substrate. Additional studies on the mechanism of substrate recognition and substrate spectrum are needed to further elucidate this phenomenon.
Our experiments show that the hosts of the RK2 plasmid have a comparable mechanism responsible for the activation of the plasmid parDE system. It is interesting that although C. crescentus has up to four copies of the parDE system on the chromosome 50 , the activating factor of each individual systems has not yet been identified. Fiebig and colleagues showed that the expression of the parDE operons from C. crescentus may be dependent on environmental factors such as heavy metals, heat shock, or the culture growth phase 50 . Since RK2 parDE stabilizes the plasmid in C. crescentus, the chromosomal parDE systems most likely do not affect the plasmidic TA. However, it is possible that under stress conditions the plasmid could be affected by chromosomal parDE. Regarding reports of the ccdAB homologous system from E. coli and the F plasmid 13,14 , it would be very interesting to investigate whether the parDE system of the RK2 plasmid may also be involved in the response of the host cells to environmental stress. We observed a slight increase in the stability of ParD in the E. coli lon(−) strain. There are studies which have described the phenomenon of antitoxin degradation by another protease during environmental stress 51 , although this effect requires further investigation.
The knowledge gained of the TA system that is active in diverse bacterial species might be applied to the development of new strategies for managing pathogens. The plasmid RK2 TA system with its broad host range properties and a universal protease component can be considered as a potential model for the development of new antibacterial strategies.
DNA manipulations and transformation. Routine DNA recombinant techniques were performed as previously described 52 . Restriction enzymes and other enzymes were used according to the supplier's instructions. C. crescentus and P. putida chromosomal DNA was isolated as described in the A&A Biotechnology Genomic Maxi AX kit. E. coli transformation was performed using the standard calcium chloride method or by electrotransformation with a Gene-Pulser (Bio-Rad), according to the protocol described by Sambrook et al., 1989. Electrotransformation of C. crescentus, P. putida and protoplasts with plasmid DNA was performed as previously described 53,54 . Protein purification and determination of proteolytic activity. The experiments described in this study utilized highly purified proteins (90% or greater purity). Published protocols were applied for the purification of ParE 29 , ParD 30 , ClpA 55 , ClpP 56 , ClpY 57 , ClpQ 58 , Lon 59 and TrfA 60 . ClpX was purified using a combination of previously described ion-exchange chromatography methods 61 . The proteolytic activity of the Clp and Lon proteins was measured using a previously described method 44 , with α-casein as a substrate for ClpAP, ClpYQ and Lon, and λO protein for ClpXP. Proteolysis of excess substrate was performed as described for the in vitro proteolysis assay. The amount of degraded substrate was estimated after SDS-PAGE, Coomassie staining and densitometry analysis.
In vivo protein stability and Western blot analysis. In vivo stability tests were performed in E. coli C600 and its protease-deficient counterpart, as previously described 61 . These strains were transformed with pRR46 (for constitutive expression of parD) and pAS12 or pBAD24-ParD that allows the inducible overexpression of ParE and ParD, respectively. Overnight cultures were diluted 1:100 in fresh LB broth and were grown at 37 °C to an OD 600 of 0.6. At this point, expression of ParE or ParD was induced. After 1 h of induction, protein synthesis was inhibited by the addition of tetracycline (final concertation of 40 µg/ml) and the samples were collected at indicated time points and suspended in 4X Laemmli buffer. Samples of 15 µl with OD 0.1 were analyzed by 12.5% or 15% SDS-PAGE, followed by Western blotting using anti-ParD (see Supplementary Fig. S10A,B) and anti-ParE (see Supplementary Fig. S10C,D) polyclonal antibodies and a polyclonal goat anti-mouse IgG HRP conjugate. The purified protein was used as a marker. The proteins were visualized by the Chemi-Doc Image Lab 5.1 system (Bio-Rad). All figures were prepared in Microsoft PowerPoint without any modification except for resizing.
Plasmid stability. The plasmid stability was determined as follows: The tested strains were grown overnight in suitable conditions (27, 30, 32 or 37 °C) in liquid media supplemented with an appropriate antibiotic. Overnight cultures of each strain were diluted 1:100 in fresh LB with antibiotic and grown to an OD 600 of 0.6. Next, we calculated how many cells were needed to obtain the OD 600 0.6 after thirty generations by the equation = N C 0 2 OD n 600 , where N 0 is the initial number of cells, C OD600 is the number of cells in 1 ml of medium with OD 600 0.6 and n is the number of generations. The culture was then diluted by serial dilution and a proper volume of culture was transferred into 500 ml of fresh, antibiotic-free medium. When the culture reached an OD 600 of 0.6, the culture was diluted in the same way and left to grow again. Each 30-generation cycle (approximately 12 hours) was repeated until 150 generations of growth under nonselective conditions was reached. Every 30-generation cycle of each culture was plated (approximately 150-200 colonies per plate) on nonselective LB or PYE agar. After overnight incubation, the colonies were examined for resistance to the given antibiotic by replica plating on selective LB or PYE agar. The percentage of plasmid-free cells was estimated from the ratio of antibiotic-sensitive colonies to overall colony number. The percentage of plasmid loss per generation was calculated using the formula Protein interaction analysis. Protein interaction analysis was carried out by applying the surface plasmon resonance (SPR) technique using a BIAcore 2000 and following the manufacturer's manual, as previously described 55 . The proteins were immobilized on a CM5 Sensor Chip. The running buffer was HBS-EP (150 mM NaCl, 10 mM HEPES pH 7.4, 3 mM EDTA, 0.005% Surfactant P20). The flow rate was set to 15 µl/min, and the volume of injection was 30 µl. The results are presented as sensograms obtained after subtraction of the background response signal from control experiments with buffer injections. The assay was performed in triplicate, and representative sensorgrams are presented. The results were analyzed using BIAevaluation software version 3.2.

Statistical analysis.
To measure whether two sets of data are significantly different from each other we used the two-tailed, homoscedastic T-test. The P values were calculated with using data sets obtained from three repetitions of the analyzed experiments. The P value less than or equal 0.05 was considered statistically significant. The calculations were carried out in the Microsoft Excel software.