Specialised functions of two common plasmid mediated toxin-antitoxin systems, ccdAB and pemIK, in Enterobacteriaceae

Toxin-antitoxin systems (TAS) are commonly found on bacterial plasmids and are generally involved in plasmid maintenance. In addition to plasmid maintenance, several plasmid-mediated TAS are also involved in bacterial stress response and virulence. Even though the same TAS are present in a variety of plasmid types and bacterial species, differences in their sequences, expression and functions are not well defined. Here, we aimed to identify commonly occurring plasmid TAS in Escherichia coli and Klebsiella pneumoniae and compare the sequence, expression and plasmid stability function of their variants. 27 putative type II TAS were identified from 1063 plasmids of Klebsiella pneumoniae in GenBank. Among these, ccdAB and pemIK were found to be most common, also occurring in plasmids of E. coli. Comparisons of ccdAB variants, taken from E. coli and K. pneumoniae, revealed sequence differences, while pemIK variants from IncF and IncL/M plasmids were almost identical. Similarly, the expression and plasmid stability functions of ccdAB variants varied according to the host strain and species, whereas the expression and functions of pemIK variants were consistent among host strains. The specialised functions of some TAS may determine the host specificity and epidemiology of major antibiotic resistance plasmids.


Introduction
Toxin-antitoxin systems (TAS) were originally discovered on bacterial plasmids in the 1980s, [1,2] but have since also been recognised on bacterial chromosomes [3]. TAS cassettes typically consist of two gene loci, governed by a common regulation mechanism [4], encoding a stable toxin that induces cell death or arrests growth, and a labile antitoxin that neutralises the toxin through binding to the toxin or other means. While the toxin is always a protein, the antitoxin can be protein or RNA based, and thus TAS can be categorised into six different types (Types I-VI), based on the nature and mechanism of action of the antitoxin [5,6]. The Type II system, in which both toxin and antitoxin are proteins, is the typical model of a TAS and the best studied, and is probably the most common in bacteria [7]. It has long been known that plasmid-mediated TAS play a role in plasmid maintenance through postsegregational killing and/or inhibition of growth of plasmid free cells [2,8,9], while chromosomal TAS have been found to be involved in a range of other roles, such as in the bacterial stress response, antibiotic tolerance, persister cell formation, biofilm formation, bacterial virulence and intestinal colonisation [10][11][12][13][14]. However, several recent studies identified the involvement of plasmid-mediated TAS in a range of bacterial physiology beyond plasmid maintenance. We have recently showed that the plasmid-mediated parDE TA system is involved in antibiotic and heat tolerance in Escherichia coli [15]. Some other plasmid-mediated TA systems such as ccdAB contributes antibiotic persistence in E. coli [16], while mpvAT is involved in virulence by maintaining virulence plasmids in Shigella and Salmonella species [17,18]. Within Enterobacteriaceae, TAS are common among conjugative plasmids including antibiotic resistance (AbR) plasmids, and often associated with certain plasmid incompatibility (Inc) types. For example, the type II TAS pemIK and vagCD are usually found on IncF and IncL/M plasmids, and IncF and IncHI2 plasmids respectively [19,20].
The distribution of TAS in the plasmids found in Escherichia coli is well described [21][22][23][24][25], but there has been as yet no comprehensive study of the TAS in all known plasmids residing in Klebsiella pneumoniae species. Existing studies of a collection of clinically relevant K. pneumoniae genomes (encompassing both plasmids and chromosomes) has provided insights into the distribution of TAS in the chromosomes of this species and the association of some TAS with certain plasmid replicon and antibiotic resistance regions, suggesting possible strain specific specialisation of TAS, with some being localised to one species while others are found across species [26,27]. Previous studies have also identified two distinct groups of vagCD TAS on plasmids located in K. pneumoniae [28], with each of the toxins and antitoxins shown to be functional. vagCD is a member of the large TA family vapBC, with the toxin VagD being a PilT N-terminal (PIN) domain containing endoribonuclease that inhibits translation [29,30].
TAS with identical names infer similar roles, regardless of the initial source and sequence variations, but functional differences in TAS from different genetic contexts have previously been noted [31,32]. For example, IncF plasmid-mediated ccdAB TAS (ccd F ) from pathogenic E. coli O157:H7 is involved in post-segregational killing but its chromosomal counterpart (ccd O157 ) is unable to mediate PSK [32]. However, common TAS functions (e.g. plasmid maintenance or antibiotic tolerance) for plasmid-mediated TA variants were not investigated.
In this study, we analysed the distribution of type II TAS in 1063 fully sequenced plasmids found in K. pneumoniae retrieved from GenBank. Variation in DNA sequence, expression and function of two common TAS found in both E. coli and K. pneumoniae plasmids, ccdAB and pemIK, were examined.
ccdAB is a well-studied, primarily plasmid-associated type II TAS, although copies have also been found on bacterial chromosomes where it appears to be involved in the bacterial stress response [33,34]. The toxin, CcdB, binds to and disrupts the action of DNA gyrase, causing double strand DNA breaks and the induction of the bacterial SOS response [35,36]. pemIK, another well studied type II TAS that is related to the mazEF system, is an mRNA endoribonuclease that inhibits protein synthesis [37]. Here, we provide vital information about the similarity, specificity and functions of these two TAS, with broad implications for their role in the spread of antibiotic resistance.

Identification of TAS on K. pneumoniae plasmids
The names and complete sequences of all plasmids from K. pneumoniae species available in GenBank at the time of search (August 2019) were retrieved (https://www.ncbi.nlm.nih.gov/ genome/?term=klebsiella+pneumoniae), and the sequences of all plasmids >30 kb examined with TA Finder (http://202.120.12.133/TAfinder/TAfinder.php) [38] using the default parameters to identify potential type II TAS. Plasmids <30 kb in size were excluded as smaller plasmids generally do not carry TA systems. The plasmid incompatibility (Inc) type of each plasmid was defined using PlasmidFinder (https://cge.cbs.dtu.dk/services/PlasmidFinder/) [39].

Alignment of TAS sequences
Representative examples of the two most common TAS, ccdAB and pemIK, were chosen, and the nucleotide and amino acid sequences of the toxin and antitoxin coding regions retrieved from GenBank (https://www.ncbi.nlm.nih.gov/genbank/) using coordinates obtained from TA Finder. To verify that these sequences were in fact representative of their specific variant, each of the sequences were used as queries for nucleotide Basic Local Alignment Search Tool (BLAST) searches (https://blast.ncbi.nlm.nih.gov/Blast.cgi) [40]. The resulting matches were compiled for each TAS, then aligned in MEGA7 (http://www.megasoftware.net/mega7/) [41] using the ClustalW algorithm, followed by construction of phylogenetic trees using the Maximum Likelihood method.
Variants were compared with one another by nucleotide alignment as described above. The amino acid sequences were then used to predict the secondary structures of the proteins using PsiPred (http://bioinf.cs.ucl.ac.uk/psipred/) [42].
TAS promoters were predicted using BPROM (http://www.softberry.com/berry.phtml? topic=bprom&group=programs&subgroup=gfindb) (Softberry) with default parameters, with the input being the 500 bp region upstream of the ATG start codon of the antitoxin gene. The putative promoter sequences were then aligned as described above.

Plasmids, bacteria, primers and culture conditions
Tables 1, 2 and 3 list plasmids, bacterial strains and primers respectively. Bacteria were grown in Luria-Bertani (LB) broth (BD Biosciences, NJ, USA), with kanamycin (50 μg/mL) or chloramphenicol (20 μg/mL) (Sigma-Aldrich, MO, USA) added as indicated. Insertion of TAS into vectors was carried out using standard restriction digestion and ligation cloning protocols, and chemical transformation and electroporation into host strains was also performed using standard protocols. The plasmid construct details can be found in Table 1. Each solution used was rendered sterile either through autoclaving or filter sterilising at 0.22 μm.

Measurement of relative promoter strength
Relative strengths of putative TAS promoters were determined from TA promoter-gfp constructs using methods described previously [47]. Briefly, the predicted TAS promoters were cloned upstream of a promoterless gfp in the expression vector pANT3, and four strains each of E. coli and K. pneumoniae were transformed with these constructs. Overnight cultures were inoculated from single colonies into LB broth with kanamycin, and grown with shaking at 37˚C. The cultures were then diluted 200 x in LB broth and grown under the same conditions for a further 3-4 h. Cells were harvested by centrifugation and the pellets resuspended in 0.85% sodium chloride (saline). The concentrations of each sample were then standardised to OD 580 of 1.0 (~8.0 x 10 8 cfu/mL), as determined using the DensiCHEK TM Plus nephelometer (bioMérieux, France). Fluorescence was analysed using a Victor3 plate reader (Perkin Elmer, MA, USA), with an excitation wavelength of 485 nm and an emission wavelength of 535 nm. Experiments were performed in triplicate, and readings were averaged and corrected for background fluorescence by subtracting the pANT3 (no promoter) reading from each sample. The  positive control pANT5, which constitutively expresses GFP from the ptac promoter, is a derivative of pANT3 [43]. Both plasmids carry the pRSF1010 origin of replication, which is a low copy broad-host range IncQ plasmid [48].

Plasmid stability assays
To assess plasmid stability, the whole TAS (including the putative promoter and ribosome binding site) was cloned into a low and a high copy plasmid (pACYC184 and pBSCK+ respectively) and two strains each of E. coli and K. pneumoniae were transformed with these constructs. Plasmid stability was assessed as described previously [15]. Briefly, a single colony of E. coli or K. pneumoniae bacteria carrying the relevant plasmid was grown in LB broth at 37˚C with shaking at 225 rpm for 72 or 96 hours, without antibiotic selection. Bacterial cultures were transferred into fresh LB medium at 1:1000 dilution at 16, 24, 40, 48, 64 and 72 hrs for high copy plasmids. Additional dilutions were performed at 88 and 96 hrs for low copy plasmids, as these are not lost as quickly as the high copy plasmids. Samples were taken before every transfer, diluted in saline and plated on to LB agar without antibiotic and incubated at

Results and discussion
Distribution of type II TAS in the plasmids found in K. pneumoniae strains Twenty-seven different putative type II TAS were identified (Table 4), with ccdAB, pemIK and vagCD most common among them (S1 Table). We compared the distribution of the TAS identified here in K. pneumoniae plasmids with previously reported TAS in E. coli plasmids [21]. E. coli and K. pneumoniae are closely related members of the Enterobacteriaceae family, sharing a large number of mobile antibiotic resistance genes, mainly via plasmids. However, only three TAS (ccdAB, pemIK, and vagCD) were common in both, with ccdAB and pemIK most commonly shared. Many, such as MNT-HEPN-like, GNAT-RHH-like, Bro-Abr-like and Bro-ArsRlike TAS appear to be most important in Klebsiella plasmids and have not previously been reported in E. coli plasmids [21,23]. Plasmids found in K. pneumoniae species belonged to 11 different Inc types (Table 4), although around 10% (114/1063) were not assigned an Inc type in PlasmidFinder. More than half of the plasmids were from the IncF replicon group (Fig 1), which is the most common plasmid replicon type in Enterobacteriaceae [49]. Replication gene variation can be used to further subdivide IncF plasmids (e.g. IncFIA, IncFIB, IncFIC, IncFII etc) and IncFII plasmids are usually further subdivided with a subscript to indicate the typical host species, e.g. IncFII K and IncFII S for the IncFII plasmids found in Klebsiella and Salmonella species, respectively. For simplicity, all IncF plasmid subtypes found in K. pneumoniae strains were grouped as IncF in this study. A number of K. pneumoniae plasmid types (L, M, X, A, C, N and H) ( Table 4, Fig 1) are also commonly found in E. coli and other Enterobacteriaceae, [50] but some that are quite Other associations are noteworthy. For example, ccdAB is somewhat specific to IncF plasmids in K. pneumoniae, while pemIK is common to both IncL/M and IncF plasmids (Table 4, Fig 1, S1 Table), in line with previous observations [19]. vagCD is another TAS predominantly found on the bla CTX-M -carrying IncF plasmids in E. coli (also as previously noted) [21], and in K. pneumoniae (Table 4, Fig 1). However, vagCD was also common in IncH (37/58) and IncR (35 of 76) plasmids in K. pneumoniae but not in E. coli. Most IncX plasmids lacked an easily identifiable TAS, save for a few carrying hicBA, also as previously noted [51]. ccdAB sequences differ between species, while pemIK is highly conserved ccdAB and pemIK were the two most predominant TAS modules across both E. coli and K. pneumoniae plasmids (Table 4, [21]), and were thus chosen for further study.
Representative sequences of ccdAB were chosen from IncF plasmids derived from two species (E. coli and K. pneumoniae; henceforth termed "ccdAB EC " and "ccdAB KP " respectively), while variants of pemIK were chosen from two plasmid incompatibility types, IncL/M and IncF (henceforth termed "pemIK LM " and "pemIK F " respectively). These sequences were confirmed as representative of their specific variant by phylogenetic analysis (Fig 2). ccdAB shows two clear groups, with one dominated by E. coli and one by K. pneumoniae, representing ccdA-B EC and ccdAB KP respectively (Fig 2A). Each variant was also found, in rare instances, in other species, however given that these are self-transmissible conjugative plasmids, this is to be expected. Similarly, pemIK also branched into two groups, with one being dominated by IncF plasmids, and the other by IncL/M plasmids (Fig 2B), irrespective to species. These results confirm that each variant chosen is representative of their specific group.
Alignment of ccdAB EC and ccdAB KP revealed 85% identity in the nucleotide coding sequence (S1A Fig), translating to 75% and 93% amino acid sequence identities in the antitoxin and toxin respectively (Fig 3A). Predicted secondary structure analysis indicated that although the antitoxin CcdA had the same basic structure in both cases, the toxin CcdB derived from K. pneumoniae had an additional small α-helix near the C-terminal end (S2 Fig). The putative promoters were less similar, with only a 33% identity throughout the promoter region (-10, -35 and spacers) (Fig 3C). Thus, it is expected that the expression of these TA modules in different species would vary. It has previously been reported that differences in the promoter regions of chromosomal and plasmid mediated ccdAB variants was associated with differences in expression and function [32].
By contrast, pemIK sequences differed with the plasmid Inc types, rather than with the host bacterial species. Alignment of pemIK LM pemIK F revealed 92% identity in the nucleotide

PLOS ONE
coding sequence (S1B Fig), translating to 95% amino acid sequence identities in both proteins (Fig 3B). The PemI antitoxins and PemK toxins also had identical secondary structures (S3 Fig). Similarly, the putative promoters had 94% nucleotide identity (Fig 3C) with only one SNP in the -10 sequences, indicating that pemIK TAS is quite conserved within different plasmid types.

Variation in expression of ccdAB in different host strains is greater than expression of pemIK
GFP expression from both pemIK promoter variants were relatively strong in all strains/species tested, with expression higher than from either of the ccdAB promoters tested, as well as the workhorse tac promoter [52] (Fig 4), in line with the apparently minor variation in pemIK promoters. pemIK is the sole TAS evident in published IncL/M plasmids [19], and these data suggest that pemIK alone is sufficient to provide stability to the plasmids carrying them, regardless of the host bacterial species.
In contrast, expression of GFP from ccdAB promoters is generally lower than that of pemIK and showed host species specificity. GFP expression from the E. coli specific ccdAB promoter was higher in E. coli than in K. pneumoniae strains. The expression of GFP from the K. pneumoniae specific ccdAB promoter also varied between strains in each species but was relatively lower than from all other promoters tested here. The association between differences in expression of ccdAB from different genetic contexts with differences in function in E. coli has previously been noted [32].

ccdAB function is specific to host strain, whereas pemIK is not
To assess whether the apparent specialisations within ccdAB (by host species) and pemIK (by plasmid type) are also reflected functionally, we assessed plasmid stability in two host strains of each species, in both low and high copy number plasmids. The low copy plasmid data are presented (Fig 5) as the natural plasmids carrying these TAS are generally low copy number. Consistent data for high copy plasmids can be found in the supplementary material (S4 Fig). The plasmid stability conferred by ccdAB variants varies between host species (Fig 5) and is generally consistent with the GFP expression data, with ccdAB EC stabilising plasmids in both E. coli strains but not in K. pneumoniae strains. ccdAB KP , on the other hand, appears to be less specialised, conferring some plasmid stability in all strains, regardless of species. Given the relatively low level of expression from ccdAB promoters and discrepancies in the conferred plasmid stability across species and strain, it is perhaps unsurprising that it is typically one of several TAS in IncF plasmids [21]. By contrast, the more reliably and vigorously expressed pemIK variants conferred significant plasmid stability in all strains tested (Fig 5). Overall, pemIK appears less specialised than ccdAB. The sequences and expression of pemIK variants differed relatively less, and plasmid stability functions of both variants were more consistent. In contrast, ccdAB appears to be more variable in sequence and function, different variants conferring plasmid stability in different strains.
ccdAB is one of the most well studied TAS and has been reported on both plasmids and chromosomes of different bacterial species [33,34]. Their sequences and functions also varied based on their genetic context or source of origin (plasmids and bacterial species), and functional differences between plasmid-mediated and chromosomal ccdAB have been demonstrated previously [32]. The finding that the CcdB toxin found in the Salmonella virulence plasmid pSLT, which has a single amino acid substitution (R99W), is non-toxic and unable to provide plasmid maintenance [53], suggests that even minor variations in TAS protein sequences may lead to changes in their functions. It has also been demonstrated that the antitoxin CcdA originating from one source does not necessarily protect bacteria from toxicity mediated by the toxin CcdB from a different origin, even though both toxins followed the same mechanism of action. For example, CcdA from Vibrio fischeri superintegron could not protect from the toxicity of plasmid mediated CcdB F [54].
The inability of ccdAB to confer significant plasmid stability in certain host strains raises the possibility that it is performing another role in these strains. It would be of interest to investigate whether these plasmid-mediated ccdAB variants are involved in other known functions of TAS within these strains, such as persister formation, antibiotic and heat tolerance, and other bacterial stress responses [10,12,15,33,55,56]. Within E. coli, both an IncF plasmid mediated ccdAB and chromosomally located variant have previously been shown to be able to confer some level of protection during antibiotic treatment [33], and the toxin CcdB from IncF plasmid has also been linked to increases in persister cell formation [16].
Variation in ccdAB has previously been noted to influence the spread of certain Salmonella serovars [53], and we believe that the variation in expression and plasmid stability function of ccdAB and pemIK may contribute to and influence the epidemiology of the conjugative antibiotic resistance plasmids which carry them. It is hoped that an increased understanding of these effects may be able to aid in the surveillance of these plasmids and the spread of the resistance genes carried by them.