Novel Toxin-Antitoxin Module SlvT-SlvA Regulates Megaplasmid Stability and Incites Solvent Tolerance in Pseudomonas putida S12

Sustainable alternatives for high-value chemicals can be achieved by using renewable feedstocks in bacterial biocatalysis. However, during the bioproduction of such chemicals and biopolymers, aromatic compounds that function as products, substrates, or intermediates in the production process may exert toxicity to microbial host cells and limit the production yield. Therefore, solvent tolerance is a highly preferable trait for microbial hosts in the biobased production of aromatic chemicals and biopolymers. In this study, we revisit the essential role of megaplasmid pTTS12 from solvent-tolerant Pseudomonas putida S12 for molecular adaptation to an organic solvent. In addition to the solvent extrusion pump (SrpABC), we identified a novel toxin-antitoxin module (SlvAT) which contributes to short-term tolerance in moderate solvent concentrations, as well as to the stability of pTTS12. These two gene clusters were successfully expressed in non-solvent-tolerant strains of P. putida and Escherichia coli strains to confer and enhance solvent tolerance.

Following the completion of its full-genome sequence and subsequent transcriptome and proteome analyses, several genes have been identified that may play important roles in controlling and maintaining solvent tolerance of P. putida S12 (10)(11)(12). As previously reported, an important solvent tolerance trait of P. putida S12 is conferred through the resistance-nodulation-division (RND)-family efflux pump SrpABC, which actively removes organic solvent molecules from the cells (13,14). Initial attempts to heterologously express the SrpABC efflux pump in Escherichia coli enabled the instigation of solvent tolerance and production of 1-naphthol (15,16). Importantly, the SrpABC efflux pump is encoded on the megaplasmid pTTS12 of P. putida S12 (12).
The 583-kbp megaplasmid pTTS12 is a stable single-copy plasmid specific to P. putida S12 (12). It harbors several important operons and gene clusters enabling P. putida S12 to tolerate, resist, and survive the presence of various toxic compounds or otherwise harsh environmental conditions. Several examples include the presence of a complete styrene degradation pathway gene cluster, the RND efflux pump specialized for organic solvents (SrpABC), and several gene clusters conferring heavy metal resistance (12,17,18). In addition, through analysis using TADB2.0, a toxin-antitoxin database (19,20), pTTS12 is predicted to contain three toxin-antitoxin modules. Toxinantitoxin modules recently have been recognized as important determinants of resistance toward various stress conditions, like nutritional stress and exposure to sublethal concentration of chemical stressors (21,22). Toxin-antitoxin modules identified in pTTS12 consist of an uncharacterized RPPX_26255-RPPX_26260 system and two identical copies of a VapBC system (23). RPPX_26255 and RPPX_26260 belong to a newly characterized type II toxin-antitoxin pair, COG5654-COG5642. While toxin-antitoxin systems are known to preserve plasmid stability through postsegregational killing of plasmid-free daughter cells (24), RPPX_26255-RPPX_26260 was also previously shown to be upregulated during organic solvent exposure, suggesting its role in solvent tolerance (11).
In the manuscript, we further address the role of pTTS12 in conferring solvent tolerance of P. putida S12. Curing pTTS12 from its host strain might cause a reduction in solvent tolerance, while complementation of the srp operon into the cured strain may fully or partially restore solvent tolerance. Furthermore, we wished to identify additional genes or gene clusters on pTTS12 and putative mechanisms that might also play a role in conferring solvent tolerance to P. putida and non-solvent-tolerant E. coli.

RESULTS
Megaplasmid pTTS12 is essential for solvent tolerance in P. putida S12. To further analyze the role of the megaplasmid of P. putida S12 in solvent tolerance, pTTS12 was removed from P. putida S12 using mitomycin C. This method was selected due to its reported effectiveness in removing plasmids from Pseudomonas sp. (25), although previous attempts regarded plasmids that were significantly smaller in size than pTTS12 (26). After treatment with mitomycin C (10 to 50 mg liter Ϫ1 ), liquid cultures were plated on M9 minimal medium supplemented with indole to select for plasmidcured colonies. Megaplasmid pTTS12 encodes two key enzymes, namely, styrene monooxygenase (SMO) and styrene oxide isomerase (SOI) that are responsible for the formation of indigo coloration from indole. This conversion results in indigo coloration in spot assays for wild-type P. putida S12, whereas white colonies are formed in the absence of megaplasmid pTTS12. With the removal of pTTS12, a loss of indigo coloration and, hence, of indigo conversion was observed in all three plasmid-cured strains and the negative control P. putida KT2440 (Fig. 1A).
With a mitomycin C concentration of 30 mg liter Ϫ1 , 2.4% (3 out of 122) of the obtained colonies appeared to be completely cured of the megaplasmid, underscoring the high genetic stability of the plasmid. No colonies survived the addition of 40 and 50 mg liter Ϫ1 of mitomycin C, whereas all the colonies that survived the addition of 10 and 20 mg liter Ϫ1 of mitomycin C retained the megaplasmid. Three independent colonies cured from the megaplasmid were isolated as P. putida S12-6, P. putida S12-10, and P. putida S12-22. The complete loss of the megaplasmid was further confirmed by phenotypic analysis (Fig. 1) and by full-genome sequencing. Several operons involved in heavy metal resistance were previously reported in the pTTS12 (12). The terZABCD operon contributes to tellurite resistance in wild-type P. putida S12 with MICs as high as 200 mg liter Ϫ1 (Fig. 1B). In the megaplasmid-cured strains, a severe reduction of tellurite resistance was observed, decreasing the potassium tellurite MIC to 50 mg liter Ϫ1 (Fig. 1B).
Genomic DNA sequencing confirmed a complete loss of pTTS12 from P. putida genotypes S12-6, S12-10, and S12-22 without any plasmid-derived fragment being inserted within the chromosome, and genomic alterations by mitomycin C treatment were minimal. Complementation of pTTS12 into the plasmid-cured P. putida S12 genotypes restored the indole-indigo transformation and high tellurite resistance to a similar level as the wild-type strain (see Fig. S1 in the supplemental material). Repeated megaplasmid curing experiments indicated that P. putida S12 can survive the addition of 30 mg liter Ϫ1 mitomycin C with the frequency of (2.48 Ϯ 0.58) ϫ 10 Ϫ8 . Among these survivors, only 2% of the colony population lost the megaplasmid, confirming the genetic stability of pTTS12. In addition, attempts to cure the plasmid by introducing double-strand breaks as described by Wynands and colleagues (27) were not successful due to the pTTS12 stability.
Toxin-Antitoxin Module SlvTA in Pseudomonas putida S12 Applied and Environmental Microbiology Growth comparison in solid and liquid culture in the presence of toluene was performed to analyze the effect of megaplasmid curing in constituting the solvent tolerance trait of P. putida S12. In contrast with wild-type P. putida S12, the plasmidcured genotypes were unable to grow under toluene atmosphere conditions (data not shown). In liquid LB medium, plasmid-cured P. putida S12 genotypes were able to tolerate 0.15% (vol/vol) toluene, whereas the wild-type P. putida S12 could grow in the presence of 0.30% (vol/vol) toluene (Fig. 2). In the megaplasmid-complemented P. putida S12-C genotypes, solvent tolerance was restored to the wild-type level (Fig. S1D). Hence, the absence of megaplasmid pTTS12 caused a significant reduction of solvent tolerance in P. putida S12.
The SrpABC efflux pump and gene pair RPPX_26255-RPPX_26260 are the main constituents of solvent tolerance encoded on pTTS12. The significant reduction of solvent tolerance in plasmid-cured P. putida S12 underscored the important role of megaplasmid pTTS12 in solvent tolerance. Besides encoding the efflux pump SrpABC enabling efficient intermembrane solvent removal (12,13), pTTS12 carries more than 600 genes and, hence, may contain additional genes involved in solvent tolerance. Two adjacent hypothetical genes, RPPX_26255 and RPPX_26260, were previously reported to be upregulated in a transcriptomic study as a short-term response to toluene addition (11). We propose to name the RPPX_26255-RPPX_26260 gene pair as "slv" due to its elevated expression in the presence of solvent. In a first attempt to identify additional potential solvent tolerance regions of pTTS12, we deleted the srpABC genes (Δsrp), RPPX_26255-RPPX_26260 genes (Δslv), and the combination of both gene clusters (Δsrp Δslv) from pTTS12 in wild-type P. putida S12.
All strains were compared for growth under increasing toluene concentrations in liquid LB medium (Fig. 2). In the presence of low concentrations of toluene (0.10% [vol/vol]), all genotypes showed similar growth. With the addition of 0.15% (vol/vol) toluene, S12 Δslv, S12 Δsrp, and S12 Δsrp Δslv exhibited slower growth and reached a lower optical density at 600 nm (OD 600 ) than the wild-type S12 strain. S12 Δslv and S12 Δsrp achieved a higher OD 600 in batch growth than S12 ΔpTTS12 and S12 Δsrp Δslv due to the presence of the SrpABC efflux pump or RPPX_26255-RPPX_26260 gene pair.
Interestingly, S12 Δsrp Δslv (still containing pTTS12) exhibited diminished growth compared with S12 ΔpTTS12. This may be an indication of megaplasmid burden in the absence of essential genes for solvent tolerance. With 0.20% and 0.30% (vol/vol) toluene added to the medium, S12 Δsrp, S12 Δsrp Δslv, and S12 ΔpTTS12 were unable to grow, while wild-type S12 and S12 Δslv were able to grow, although S12 Δslv reached a lower OD 600 . Taken together, these results demonstrate an important role for both the SrpABC efflux pump and the slv gene pair in conferring solvent tolerance. We chose P. putida S12-6 for further experiments representing megaplasmid-cured P. putida S12.
Solvent tolerance can be exerted by ectopic expression of the SrpABC efflux pump and slv gene pair in Gram-negative bacteria. The functionality of the srp operon and slv gene pair was explored in the model Gram-negative non-solventtolerant strains P. putida KT2440, E. coli TG1, and E. coli BL21(DE3). We complemented srpRSABC (srp operon), slv gene pair, and a combination of both gene clusters into P. putida S12-6, P. putida KT2440, E. coli TG1, and E. coli BL21(DE3) using mini-Tn7 transposition. These strains were chosen due to their common application as model industrial strains while lacking solvent tolerance. P. putida KT2440 is another robust microbial host for metabolic engineering due to its adaptation toward physicochemical stresses; however, contrary to P. putida S12, this strain is not solvent tolerant (28). E. coli BL21(DE3), derived from strain B, is the common E. coli lab strain optimized for protein production due to its lacking Lon and OmpT proteases and encoding T7 RNA polymerase (29). E. coli TG1 was previously reported to successfully produce 1-naphthol with the expression of SrpABC (15,16), and therefore, this strain was included in this study as a comparison.
The chromosomal introduction of slv into S12-6 and KT2440 improved the growth of the resulting strains at 0.15% (vol/vol) toluene compared with S12-6 and KT2440 (Fig.  3). The introduction of srp or a combination of slv and srp enables S12-6 and KT2440 to grow in the presence of 0.30% (vol/vol) toluene. In KT2440, the introduction of both slv and srp resulted in a faster growth in the presence of 0.30% (vol/vol) toluene than the addition of only srp (Fig. 3B). Interestingly, the growth of S12-6 srp,slv and S12.6 srp is better than wild-type S12 (Fig. 3A). The observed faster growth of S12-6 srp,slv and S12.6 srp may be due to more efficient growth in the presence of toluene, supported by a chromosomally introduced srp operon, than its original megaplasmid localization. Indeed, replication of this large megaplasmid is likely to require additional maintenance energy. To corroborate this, we complemented the megaplasmid pTTS12 lacking the solvent pump (Tc r ::srpABC) into P. putida S12-6 srp, resulting in P. putida S12-9. Indeed, P. putida S12-9 showed further reduced growth in the presence of 0.20 and 0.30% toluene (see Fig. S2 in the supplemental material), indicating the metabolic burden of , and the combination of these gene clusters into P. putida S12 ΔpTTS12 (represented by strain S12-6) (A) and wild-type P. putida KT2440 (B) in liquid LB with 0%, 0.10%, 0.15%, 0.20%, and 0.30% (vol/vol) of toluene. Wild-type P. putida S12 was taken as a solvent-tolerant control strain. This figure displays the mean of three independent replicates, and error bars indicate standard deviation. The ranges of y axes are different in the first panel (0 to 6), second panel (0 to 3), and third to fifth panels (0 to 1.5).
carrying the megaplasmid. We conclude that the SrpABC efflux pump can be regarded as the major contributor to solvent tolerance from pTTS12. The slv gene pair appears to promote the tolerance of P. putida S12 at least under moderate solvent concentrations.
The intrinsic solvent tolerance of E. coli strains was observed to be clearly lower than that of P. putida (Fig. 4). The wild-type E. coli strains were able to withstand a maximum 0.10% (vol/vol) toluene, whereas plasmid-cured P. putida S12-6 and P. putida KT2440 were able to grow in the presence of 0.15% (vol/vol) toluene. With the introduction of slv and srp in both E. coli strains, solvent tolerance was increased up to 0.15% and 0.20% (vol/vol) toluene, respectively (Fig. 4). A combination of slv and srp also increased tolerance to 0.20% (vol/vol) toluene, while showing a better growth than the chromosomal introduction of only srp. However, none of these strains were able to grow in the presence of 0.30% (vol/vol) toluene. qPCR analysis of SrpABC expression (see Table S1 in the supplemental material) in P. putida S12, P. putida KT2440, E. coli TG1, and E. coli BL21(DE3) confirmed that srpA, srpB, and srpC were expressed at basal levels in all strains. In the presence of 0.10% toluene, the expression of srpA, srpB, and srpC was clearly upregulated in all strains. Thus, the lower solvent tolerance conferred by introducing the SrpABC efflux pump in E. coli strains was not due to the lower expression of the srp genes. An analysis of the codon adaptation index (CAI) (http://ppuigbo.me/programs/CAIcal/) (30) showed that for both the P. putida and E. coli strains, the CAI values of the srp operon are suboptimal, clearly below 0.8 to 1.0 (see Table S2 in the supplemental material). Interestingly, the CAI values were higher for E. coli (0.664) than for P. putida (0.465), predicting a better protein translation efficiency of the srp operon in E. coli. Hence, reduced translation efficiency is not likely to be the cause of lower performance of the srp operon in E. coli strains for generating solvent tolerance. Overall, our results indicate that, in addition to the solvent efflux pump, P. putida S12 and P. putida KT2440 are intrinsically more robust than E. coli TG1 and E. coli BL21(DE3) in the presence of toluene.
The slv gene pair constitutes a novel toxin-antitoxin system. BLASTp analysis was initiated to further characterize RPPX_26255 and RPPX_26260. This analysis indi- cated that RPPX_26260 and RPPX_26255 likely represent a novel toxin-antitoxin (TA) system. Through a database search on TADB2.0 (19,20), we found that RPPX_26260 is a toxin of the COG5654 family and typically carries an RES domain-containing protein, which has a conserved arginine (R)-glutamine (E)-serine (S) motif providing a putative active site; and RPPX_26255 is an antitoxin of the COG5642 family. Based on its involvement in solvent tolerance, we propose naming the toxin-encoding RPPX_26260 as slvT and the antitoxin-encoding RPPX_26255 as slvA.
Makarova and colleagues identified putative toxin-antitoxin pairs through genome mining of reference sequences in the NCBI database (31). They identified 169 pairs of the COG5654-COG5642 TA system from the reference sequences. Here, we constructed a phylogenetic tree of the COG5654-COG5642 TA system, including SlvA (GenBank accession no. AJA16859.1) and SlvT (AJA16860.1), as shown in Fig Fig. 5A and 6A. A 3D-model prediction of the SlvT and SlvA proteins using the I-TASSER suite for protein structure and function prediction (35) indicated that SlvT and SlvA showed the highest structural similarity to the MbcAT system from Mycobacterium tuberculosis ( Fig. 5C and 6C), which is reported to be expressed during stress conditions (34). Amino acid conservation between SlvAT and these few characterized toxin-antitoxin pairs is relatively low, as they do not belong to the same clade ( Fig. 5A and 6A). However, 100% conservation is clearly observed on the putative active side residues, namely, arginine (R) 35, tyrosine (Y) 45, and glutamine (E) 56, and only 75% consensus is shown on serine (S) 133 residue (see Fig. S3 in the supplemental material).
According to the model with the highest TM score, SlvT is predicted to consist of four beta sheets and four alpha helices. As such, SlvT shows structural similarity with diphtheria toxin which functions as an ADP-ribosyl transferase enzyme. Diphtheria toxin can degrade NAD ϩ into nicotinamide and ADP ribose (36). A similar function was recently identified for COG5654 family toxins from P. putida KT2440, M. tuberculosis, and Sphingobium sp. (32)(33)(34).
slvT toxin causes cell growth arrest by depleting cellular NAD ؉ . To prove that slvAT presents a pair of toxin and antitoxin, slvA and slvT were cloned separately in pUK21 (lac-inducible promoter) and pBAD18 (ara-inducible promoter), respectively. The two constructs were cloned into E. coli BL21(DE3). The growth of the resulting strains was monitored during conditional expression of the slvA and slvT genes (Fig. 7A). At the mid-log growth phase, a final concentration of 0.8% arabinose was added to the culture (*), inducing expression of slvT. After 2 h of induction, growth of this strain ceased, while the uninduced control culture continued to grow. Upon addition of 2 mM isopropyl-␤-D-thiogalactopyranoside (IPTG) (**), growth of the slvT-induced culture was immediately restored, reaching a similar OD 600 as the uninduced culture.
Bacterial cell division was further studied by flow cytometer analyses during the expression of slvT and slvA. After approximately 6 h of growth (indicated by gray arrow in Fig. 7A), samples were taken from control, arabinose, and arabinose ϩ IPTG-induced liquid culture. Cell morphology was analyzed by light microscopy, and the DNA content of the individual cells in the culture was measured using a flow cytometer with SYBR green II staining (Fig. 7B). Indeed, an absence of dividing cells and lower DNA content were observed during the induction of only slvT toxin with arabinose (Fig. 7B). Subsequent addition of IPTG to induce slvA expression was shown to restore cell division and 51 .
.   to an upshift of DNA content similar to that of control strain (Fig. 7B). While the expression of slvT was not observed to be lethal to the bacterial strain, this experiment showed that the expression of the slvT toxin stalled DNA replication and, subsequently, cell division. The induction of slvA subsequently restored bacterial DNA replication and cell division.

CWRILAPRWAYQPLSGEGAARNGGRWNPKGVPALYLSESIDTAFAEYQQT--LYVR--PGTFCAYRLLVAGVVDLC---DPTVLVELGVNEETLPCAWRE
To corroborate a putative target of SlvT, concentrations of NAD ϩ were measured during the induction experiment (Fig. 7C). Before the addition of arabinose to induce slvT (orange arrow on Fig. 7A), NAD ϩ was measured and compared to the strain harboring empty pUK21 and pBAD18 (Fig. 7B). On average, at this time point, the NAD ϩ level is similar between the slvAT-bearing strain and the control strain. NAD ϩ was measured again after arabinose induction when the growth of the induced strain has diminished (blue arrow on Fig. 7A). At this time point, the measured NAD ϩ was 32% (Ϯ 14.47%) of the control strain. After the induction of slvA, NAD ϩ was immediately restored to a level of 77% (Ϯ 9.97%) compared with the control strain. Thus, the induction of slvT caused a depletion of NAD ϩ , while induction of slvA immediately increased the NAD ϩ level, indicating that slvAT is a pair of toxin-antitoxin which controls its toxicity through NAD ϩ depletion.
slvAT regulates megaplasmid pTTS12 stability. In addition to its role in solvent tolerance, localization of the slvAT pair on megaplasmid pTTS12 may have an implications for plasmid stability. pTTS12 is a very stable megaplasmid that cannot be spontaneously cured from P. putida S12 and cannot be removed by introducing double-strand breaks (see above). We deleted slvT and slvAT from the megaplasmid to study their impact on pTTS12 stability. With the deletion of slvT and slvAT, the survival rate during treatment with mitomycin C improved significantly, reaching (1.01 Ϯ 0.17) ϫ 10 Ϫ4 and (1.25 Ϯ 0.81) ϫ 10 Ϫ4 , respectively, while the wild-type S12 had a survival rate of (2.48 Ϯ 0.58) ϫ 10 Ϫ8 .
We determined the curing rate of pTTS12 from the surviving colonies. In wild-type S12, the curing rate was 2% (see also above), while in ΔslvT and ΔslvAT mutants, the curing rate increased to 41.3% (Ϯ 4.1%) and 79.3% (Ϯ 10%), respectively, underscoring an important role for slvAT in megaplasmid stability. We attempted to cure the megaplasmid from the colonies by introducing a double-strand break (DSB), as previously described on Pseudomonas taiwanensis VLB120 (27,37). This indeed was not possible in wild-type S12 and ΔslvT strains; however, the ΔslvAT mutant now showed plasmid curing by a DSB, resulting in a curing rate of 34.3% (Ϯ 16.4%).
Since ΔslvT and ΔslvAT may compromise megaplasmid stability, we performed megaplasmid stability tests by growing S12 and KT2440 harboring pSW-2 (negative control) (37), pTTS12 (positive control), pTTS12 ΔslvT, and pTTS12 ΔslvAT in LB medium with 10 passages (Ϯ 10 generations/passage step) as shown in Fig. 8. Both KT2440 and S12 easily lost the negative-control plasmid pSW-2 (Fig. 8). Plasmid pTTS12 was not lost during this test, confirming that pTTS12 is indeed a stable plasmid. Furthermore, the ΔslvT genotypes also did not show a loss of the megaplasmid. Interestingly, the ΔslvAT genotypes spontaneously lost the megaplasmid, confirming that the slvAT module is not only important to promote solvent tolerance but also determines megaplasmid stability in P. putida S12 and KT2440.

DISCUSSION
Revisiting the role of pTTS12 and the SrpABC efflux pump in solvent tolerance. In this study, we conclusively confirm the role of the SrpABC efflux pump carried on pTTS12 and identify a novel toxin-antitoxin module playing an additional role in   .......................A.......P.......T.......G..............G..  conveying solvent tolerance to P. putida S12 (Fig. 9). Notably, megaplasmids may cause a metabolic burden to their host strains, and they can be a source of genetic instability (11). Our results show that, indeed, pTTS12 imposed a metabolic burden in the presence of an organic solvent (Fig. S2). This plasmid is very large and contains many genes that are not related to solvent tolerance. Hence, it may be interesting for biotechnological purposes to reduce the plasmid size and, consequently, the metabolic burden. In addition, a streamlined and minimal genome size is desirable for reducing host interference and genome complexity (12,13). We investigated the heterologous expression of the SrpABC efflux pump in strains of both P. putida and E. coli, which successfully enhanced their solvent tolerance in these strains (Fig. 3 and 4). Previous reports on the implementation of SrpABC in whole-cell biocatalysis successfully increased the production of 1-naphthol in E. coli TG1 (15,16). Production was still higher using P. putida S12, as this strain could better cope  with substrate (naphthalene) toxicity, while both P. putida S12 and E. coli TG1 showed similar tolerance to the product 1-naphthol (16). In our experiments, the E. coli strains clearly showed a smaller increase in toluene tolerance than the P. putida strains, although srpABC was expressed at a basal level and upregulated in the presence of 0.10% (vol/vol). These results indicate that besides having an efficient solvent efflux FIG 8 SlvAT is important for pTTS12 maintenance in P. putida. pTTS12 (variant with Km r ) maintenance in P. putida S12 and P. putida KT2440 growing in LB liquid medium without antibiotic selection for 10 passages (approximately 10 generations per passage). pSW-2 was taken as a negative control for plasmid stability in P. putida. This experiment was performed with three biological replicates, and error bars represent standard deviation.

FIG 9
Schematic representation of the gene clusters involved in solvent tolerance from megaplasmid pTTS12. The SrpABC efflux pump is the major contributor to the solvent tolerance trait from the megaplasmid pTTS12. This efflux pump is able to efficiently extrude solvents from the membrane lipid bilayer. A COG5654-COG5642 family toxin-antitoxin module (SlvT and SlvA, respectively) promoted the growth of P. putida S12 in the presence of a moderate solvent concentration and stabilized the pTTS12 plasmid. In the absence of SlvA, SlvT causes toxicity by conferring cellular NAD ϩ depletion and, subsequently, halts DNA replication and cell division. pump, P. putida S12 and P. putida KT2440 are inherently more robust in the presence of toluene and, presumably, other organic solvents than E. coli TG1 and E. coli BL21(DE3). The absence of cis-trans isomerase (cti), resulting in the inability to switch from cis-to trans-fatty acid in E. coli (38), may contribute to this difference in solvent tolerance. Additionally, P. putida typically has a high NAD(P)H regeneration capacity (39,40) which can contribute to the maintenance of proton motive force during solvent extrusion by the RND efflux pump. Further detailed investigation is required to reveal the exact basis for its intrinsic robustness.
Identification of the novel antitoxin-toxin module SlvAT. In P. putida S12, deletion of srpABC genes still resulted in higher solvent tolerance than the pTTS12cured genotypes (Fig. 2, panel 3). This finding indicated that within pTTS12 there were other gene(s) which may play a role in solvent tolerance. Two genes of unknown function were upregulated in a transcriptome analysis of toluene-shocked P. putida, namely, RPPX_26255 and RPPX_26260, suggesting a putative role in solvent tolerance (11). Here, we confirmed this finding and demonstrated that these genes together form a novel toxin-antitoxin module (Fig. 7). SlvT exerts toxicity by degradation of NAD ϩ , like other toxins of the COG5654 family, and expression of antitoxin SlvA immediately restored NAD ϩ levels. Depletion of NAD ϩ interfered with DNA replication and caused an arrest of cell division similar to another recently described COG5654-COG5642 family toxin-antitoxin pair (33). Indeed, the SlvAT toxin-antitoxin module was shown to be important for the stability of pTTS12 (Fig. 8).
Based on TADB2.0 analysis, pTTS12 encodes three TA pairs, namely, SlvAT and two identical copies of VapBC. VapBC was first identified from a virulence plasmid of Salmonella sp. and is known to prevent the loss of plasmid during nutrient-limiting conditions (41). A previous report showed that VapBC can stabilize/retain approximately 90% of the pUC plasmid in E. coli within 300 h of growth (42), which is similar to our result although demonstrated in a much smaller plasmid and under the control of the lac operon. Serendipitous plasmid loss due to double-strand break was reported in pSTY, which carries two identical copies of VapBC (27). Here, we observed a similar phenomenon in pTTS12 ΔslvAT. Hence, in the absence of SlvAT, two copies of VapBC were not sufficient to prevent the loss of pTTS12 on rich media without selection pressure and by double-strand break, indicating a major role for SlvAT.
A putative role of toxin-antitoxin module SlvAT in solvent tolerance. Toxinantitoxin modules are known to be important in antibiotic persistent strains as a trigger to enter and exit the dormant state, causing the cell to become unaffected by the antibiotic (42). Among Pseudomonas species, several toxin-antitoxin modules are reported to be involved in survival strategies, such as stress response, biofilm formation, and antimicrobial persistence (33,(43)(44)(45). Previous transcriptomic studies reported upregulation of the slvAT locus as a response toward toluene addition and its expression at 10 to 30 minutes after toluene addition (11). Here, we show that SlvAT improves solvent tolerance in P. putida and E. coli strains independent of pTTS12 or the SrpABC efflux pump. We hypothesize that SlvAT plays a role as a rapid response toward toluene addition. Activation of the SlvT toxin may halt bacterial growth, and this allows physiological adaptation and adjustments to take place (e.g., expression of extrusion pumps and membrane compaction) before resuming its growth and cell division in the presence of toxic organic solvent. It is interesting to note that P. putida S12 and KT2440 both carry another COG5654-COG5642 family toxin-antitoxin pair in their chromosome (locus tag RPPX_19375-RPPX_19380 and PP_2433-PP_2434, respectively). In P. putida S12, this TA module is not being induced during solvent stress, rendering it unlikely to play a role in solvent tolerance.
The putative regulation mechanism of toxin-antitoxin module SlvAT in P. putida S12. Expression of the slvAT locus with its native promoter region seemed to exert a similar physiological effect in solvent tolerance both in E. coli and P. putida ( Fig.  3 and 4). Typically, toxin-antitoxin can regulate its own expression by antitoxin binding to the promoter region (46). Unstable antitoxin is encoded upstream of the stable toxin, giving a transcriptional advantage for the production of antitoxin (47). While this study presents a role of the SlvAT module as a response to solvent stress, this toxin-antitoxin module may play a role in the response to various other stresses since pTTS12 itself encodes several modules involved in different stress response. It would be interesting to further study whether organic solvents directly induce the expression of the slvAT locus or intermediate signaling pathways are required. Several type II toxin-antitoxin modules are known to be regulated by proteases, such as Lon and Clp (48). These proteases degrade the antitoxin protein, promoting toxin activity, and thus upregulate the expression of the toxin-antitoxin locus. Indeed, our preliminary transcriptomic data show upregulation of specific protease-encoding loci after toluene addition. They may constitute putative regulatory proteases to the SlvAT module. Future research on the dynamics of slvAT locus regulation is required for revealing the details of the control mechanisms operating in vivo.
Conclusions. In summary, our experiments confirmed that the SrpABC efflux pump is the major contributor of solvent tolerance on the megaplasmid pTTS12 which can be transferred to other non-solvent-tolerant host microbes. In addition, the megaplasmid carries the novel toxin-antitoxin system SlvAT (RPPX_26255 and RPPX_26260) which promotes rapid solvent tolerance in P. putida S12 and is important for maintaining the plasmid stability of pTTS12. Chromosomal introduction of the srpRSABC operon genes in combination with slvAT confers a clear solvent tolerance phenotype in other industrial strains previously lacking this phenotype, such as P. putida KT2440, E. coli TG1, and E. coli BL21(DE3). Taken together, our findings show that both SrpABC and SlvAT constitute suitable candidate loci for exchange with various microbial hosts for increasing tolerance toward toxic compounds.

MATERIALS AND METHODS
Strains and culture conditions. Strains and plasmids used in this study are listed in Table 1  Wild-type P. putida S12 (ATCC 700801), harboring megaplasmid pTTS12 3 S12-1 P. putida S12, harboring megaplasmid pTTS12 with Km r marker This study S12-6/S12-10/S12-22 ΔpTTS12 This study S12-9 ΔpTTS12, Gm r srpRSABC::Tn7, complemented with megaplasmid pTTS12 (Tc r ::srpABC) This study S12-C P. putida ΔpTTS12 (S12-6/S12-10/S12- 22 (27). Deletion of srpABC, slvT, and slvAT was verified by PCR and Sanger sequencing (Macrogen B.V., Amsterdam, The Netherlands). Introduction of the complete srp operon (srpRSABC) and slvAT was accomplished using the mini-Tn7 delivery vector backbone of pBG35 developed by Zobel and colleagues (44). The DNA fragments were obtained by PCR using primer pairs listed in Table 2 and ligated into the pBG35 plasmid at PacI and XbaI restriction sites. This construct generated a Tn7 transposon segment in pBG35 containing a gentamicin resistance marker and srp operon with Tn7 recognition sites flanking on the 5= and 3= sides of the segment. Restriction analysis followed by DNA sequencing (Macrogen, The Netherlands) were performed to confirm the correct pBG-srp, pBG-slv, and pBG-srp-slv construct. The resulting construct was cloned in E. coli WM3064 and introduced into P. putida or E. coli strains with the help of E. coli WM3064 pTnS-1. Integration of the construct into the Tn7 transposon segment was confirmed by gentamicin resistance, PCR, and the ability of the resulting transformants to withstand and grow under toluene atmosphere conditions. Toxin-antitoxin assay. Bacterial growth during the toxin-antitoxin assay was observed in LB medium supplemented with 100 mg liter Ϫ1 ampicillin and 50 mg liter Ϫ1 kanamycin. Starting cultures were inoculated from a 1:100 dilution of overnight culture (OD 600 , 0.1) into a microtiter plate (96 well), and bacterial growth was measured using a Tecan Spark 10M instrument. To induce toxin and antitoxin, a total concentration of 0.8% (m/vol) arabinose and 2 mM IPTG were added to the culture, respectively. Cell morphology was observed using a light microscope (Zeiss Axiolab 5) at a magnification of ϫ100. A final concentration of 2.5ϫ SYBR green I (10,000ϫ stock; New England BioLabs) was applied to visualize DNA, followed by two times washing with 1ϫ phosphate-buffered saline (PBS), and analyzed using a Guava easyCyte single sample flow cytometer (Millipore). At indicated time points, NAD ϩ levels were measured using a NAD/NADH-Glo assay kit (Promega) according to the manufacturer's manual. The percentage of the NAD ϩ level was calculated by dividing the measured luminescence of tested strains with that of the control strains at the same time points. RPPX_26255 and RPPX_26260 were modeled using the I-TASSER server (35) and visualized using PyMol (version 2.3.1). Phylogenetic trees of toxinantitoxin module derived from the COG5654-COG5642 family were constructed using MEGA (version 10.0.5) as a maximum likelihood tree with 100 bootstraps and visualized using the iTOL Web server (https://itol.embl.de) (45).
Data availability. The sequence data for wild-type P. putida S12 and plasmid-cured genotypes P. putida S12 ΔpTTS12 have been submitted to the SRA database under accession number PRJNA602416.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. SUPPLEMENTAL FILE 1, PDF file, 7 MB.