Toxin-Antitoxin Systems on the Large Defense Plasmid pSYSA of Synechocystis sp. PCC 6803*

Background: Cyanobacteria harbor a vast array of toxin-antitoxin modules, but their roles remain largely unknown. Results: The sll7003/ssl7004 system involves a VapC-type RNA endonuclease targeting single-stranded regions with little specificity. Conclusion: Seven different TA systems plus three free-standing components are located on plasmid pSYSA. Significance: Some bacterial plasmids are safeguarded against post-segregational loss by multiple TA systems and an aggressive type of RNA-targeting toxicity. Bacterial toxin-antitoxin (TA) systems are genetic elements, which are encoded by plasmid as well as chromosomal loci and mediate plasmid and genomic island maintenance through post-segregational killing mechanisms. TA systems exist in surprisingly high numbers in all prokaryotes, but cyanobacterial TA systems have been only very poorly experimentally characterized so far. Cyanobacteria are the only prokaryotes that perform oxygenic photosynthesis. As such, cyanobacteria are of high ecological importance and are considered promising for the production of biofuels. Here, we present the molecular characterization of the sll7003/ssl7004 TA system encoded on plasmid pSYSA of the model cyanobacterium Synechocystis sp. PCC 6803 as involving a Mg2+-dependent RNA endonuclease activity targeting single-stranded RNA regions and demonstrate the functionality of four more TA systems encoded on this 100,749-bp plasmid. Furthermore, one additional type I, one additional type II, and three freestanding TA system components are predicted on pSYSA, all of which appear active judged by their expression. By harboring at least seven simultaneously active TA systems, pSYSA appears as the plasmid most strongly selected for among all plasmids studied in this respect thus far. These results point to a high biological relevance of pSYSA, whose coding capacity is 75% devoted to three distinct clustered regularly interspaced short palindromic repeats (CRISPR) systems mediating antiviral defense.

free, these become inhibited and eventually die, caused by the depletion of CcdA antitoxin, which has a shorter half-life than CcdB (2). A very similar observation was made by Gerdes et al. (3) for maintenance of the E. coli plasmid R1, except that the antitoxin turned out to be an antisense RNA (asRNA). In the R1 TA system, the mRNA encoding the toxin hok (for host killing) is stable, whereas the asRNA sok (for suppressor of killing), which normal prevents translation of hok, is rapidly turned over and needs to be synthesized continuously. In the case of plasmid loss, the stable toxin mRNA hok becomes translated in the daughter cells, and the effect is similar as in the ccdB/ccdA system (4). Because both systems prevent the growth of plasmidfree cells, the underlying mechanism has also been called "postsegregational killing." Since the initial discovery, bioinformatics and experimental approaches have identified a wide variety of potential TA systems in various groups of bacteria. Mechanistically, at least five groups of TA systems can be differentiated. TA systems classified as type I employ small RNA or asRNA, rather than a small protein, as the antidote molecule (5). In many cases, the RNA component is not easily identified due to the variability among such RNA antidotes. However, there is at least one common denominator among these RNA antidotes; they repress the expression of proteins that are very frequently under 60 amino acids in length, are highly hydrophobic, and are toxic at high levels. Mechanistically, the antidote effect is achieved by base pairing across the ribosome binding site of the toxin mRNA, leading to a block in translation, and frequently, rapid mRNA degradation. In type III TA systems, the antitoxin RNA inhibits the protein toxin by binding (6,7), and in type IV systems, the protein antitoxin interferes with binding of the toxin to its target (8). TA systems of type V were recently characterized as systems in which the toxin mRNA is degraded by the antitoxin protein (9).
The TA systems that are probably most frequent are of type II. These are characterized by peptide antitoxins binding and inhibiting the toxin directly (10). Based on the known protein sequences, Pandey and Gerdes (11) analyzed 126 completely sequenced prokaryotic genomes (16 archaea and 110 bacteria) for the presence of type II TA systems and found 671 TA loci belonging to seven different classes. The work by Pandey und Gerdes (11) revealed that although most plasmids are likely to encode at least one TA system, the vast majority of TA systems are encoded at loci distributed throughout the chromosomal DNA. These findings and the observation that some TA systems are activated in response to environmental stress have led to the idea that TA systems not necessarily have functions only in maintaining certain genetic elements but may contribute to an enhanced stress resilience of the bacterial cell (12). In a more recent study, 750 complete genomes of archaea and bacteria were surveyed for the presence of type II TA systems, and an impressive number of 6797 TA pairs and a total of 51,319 putative toxin and antitoxin genes were identified (13). These include 175 individual toxin and antitoxin genes and 30 TA pairs in the model cyanobacterium Synechocystis sp. PCC 6803 (from here on "Synechocystis"). The most current update can be obtained in the Toxin-Antitoxin Database (TADB), where 10,753 type II TA gene pairs are listed within 1240 prokaryotic genomes (14), including 37 TA pairs in Synechocystis.
Because cyanobacteria are the only prokaryotes that directly convert solar energy into biomass using oxygenic photosynthesis, these are of tremendous relevance for primary carbon fixation in many ecosystems and of rising biotechnological relevance and promise for the production of biofuels (15). However, the only cyanobacterial TA system that has been experimentally characterized thus far is the chromosomally located relNE (ssr1114/slr0664) TA system of Synechocystis and its proteolytic regulation by ATP-dependent proteases (16).
For a better understanding of TA system functions, it is of uttermost importance which cellular processes and structures become targeted by the respective toxin. The main processes affected by the toxic component of type II TA systems so far identified are translation, DNA replication, and DNA metabolism, whereas interference with membrane organization and function is supposed to be frequent in the case of type I systems (5). It appears likely that additional processes become affected by some of the TA systems not experimentally studied so far.
With regard to the molecular mechanisms of type II toxins, most information has been gained for the seven TA families defined by Pandey and Gerdes (11). The well explored CcdB (TA pair CcdB/CcdA) and ParE toxins (ParD/ParE) are inhibitors of DNA gyrase (17,18) that abrogate cell reproduction by blocking DNA replication. The cellular target of CcdB is the DNA gyrase subunit GyrA. Binding has two possible outcomes; besides an inhibition of DNA supercoiling, the more severe result is that the enzyme within the gyrase-DNA cleavage complex becomes permanently locked in this state, leading to DNA double-strand breaks (18).
However, mRNA-specific endonuclease activity appears as the most common activity of type II toxins. These toxins are also termed interferases (19,20). HigB falls in this category (21), but interferases of the widespread RelE and MazF family are particularly well studied. RelE cleaves mRNAs in the ribosomal A site (22), whereas MazF cleaves single-stranded RNAs sequence specifically at ACA sites (23) or is even more selective, cleaving at UACAU, as in the case of the Clostridium difficile and Staphylococcus equorum MazF toxins (24,25).
Among studied TA systems, the VapBC family constitutes the most frequent, diverse, and complex class. Toxins belong-ing to this family possess a PIN domain (homologues of the "pilT N"-terminal domain). VapC-1 from Haemophilus influenzae shows an endoribonucleolytic activity (26), and vapBC LT2 and vapBC pMYSH from Salmonella LT2 and Shigella plasmid pMYSH6000 inhibit translation due to the RNase activity encoded by the respective toxin gene (27). However, the biological function and molecular targets of the diverse vapBC loci abundant in various other bacteria are not yet known.
In addition to its single chromosome, seven plasmids have been identified in Synechocystis. The chromosomal reference sequence of Synechocystis was determined as early as 1996 (28), and the sequences of the large plasmids were reported in 2003 (29), whereas the three small plasmids had already been sequenced before (30 -32). There are no genes on these seven plasmids known to have essential cellular functions. Nevertheless, the presence of all seven plasmids was verified during a recent resequencing study (33), suggesting the activity of efficient mechanisms for the maintenance of these plasmids. The plasmid pSYSA is a major defense plasmid because it encodes all components of three independent CRISPR systems, 3 thought to be involved in the recognition and degradation of invading alien DNA molecules (for reviews, see Refs. [35][36][37]. Due to two deletions in the repeat-spacer arrays of the CRISPR system in the Synechocystis substrain PCC-M used here, the total length of plasmid pSYSA is only 100,749 bp (GenBank TM accession number CP003267 and Trautmann et al. (33)) as compared with 103,307 bp for the reference substrain (29).
Among the 103 genes annotated on plasmid pSYSA are at least 12 genes coding for six different type II TA systems. Here we present a molecular characterization of the sll7003/ssl7004 TA system and demonstrate that at least four more TA systems encoded on plasmid pSYSA are functional.

Construction of Expression Plasmids pQE70::Toxin and pBAD::Antitoxin
Each toxin was cloned into the expression vector pQE70 (Qiagen). For sll7003, the ATG codon was shifted as no clones were obtained when using the pQE70 ATG codon and opti-mized Shine-Dalgarno region, likely due to leaky expression. The ampicillin resistance gene of expression vector pBAD/ Myc-His B (Invitrogen) was exchanged with the chloramphenicol or streptomycin resistance gene, respectively, resulting in plasmids pBAD_Cm and pBAD_Strep. Antitoxins were cloned into these vectors. Antitoxin gene ssl7004 was also cloned in expression vector pET28a (Novagen). All oligonucleotides with engineered restriction sites and plasmids are given in supplemental Table S1.

Growth Inhibition Assays
E. coli M15 pQE70::toxin pBAD::antitoxin strains were inoculated and grown for several hours at 37°C with shaking in LB medium containing appropriate antibiotics. These cultures were diluted to A 600 ϭ 0.1 in fresh LB medium also containing the appropriate antibiotics. Inducers IPTG and arabinose were added from the start or at an A 600 ϳ0.18. Cultures expressing only toxin, only antitoxin, both proteins, and no protein were set up in duplicates and monitored for growth by measuring A 600 in 15-min intervals (VICTOR3 TM multilabel plate reader, PerkinElmer Life Sciences).

Overexpression and Purification of Sll7003 and Ssl7004
Overnight cultures of E. coli M15 pQE70::sll7003 and E. coli BL21(DE3) pET28::ssl7004 were diluted in 1 liter of LB respectively and grown to an A 600 of 0.6 -0.8. Overexpression of Sll7003(His) 6 and Ssl7004(His) 6 was induced by adding IPTG for 3 h at 37°C with shaking. Protein purification was executed as described before (38) with slight modifications. For inhibition of proteases, phenylmethanesulfonyl fluoride and the Complete protease inhibitor mixture (Roche Diagnostics) were used. Wash buffer B composition was changed to 50 mM NaH 2 PO 4 , 300 mM NaCl, 40 mM imidazole (pH adjusted to 8.0 using NaOH).

In Vitro Synthesis of RNA
In vitro transcription for 16 S rRNA was realized with the MEGAscript kit (Invitrogen), whereas for the smaller 127-nt RNA fragment, antisense to gene sll0406 and its complement the MEGAshortscript TM kit (Invitrogen) was employed. PCR products containing a T7 RNA polymerase promoter sequence were used as templates, and transcription was carried out according to the manufacturer's instructions, including the optional DNase treatment and phenol/chloroform extraction.

RNase Activity
Fluorescence Assay-For the fluorescence assays, we employed the RNaseAlert TM (Integrated DNA Technologies) detection system, which uses a fluorescence-quenched oligonucleotide probe as substrate that emits a fluorescent signal after nuclease degradation. The substrate was incubated at 30°C in 25 mM Tris-HCl (pH 7.5), 60 mM KCl, 100 mM NH 4 Cl, 5 mM MgCl 2 , and 0.1 mM DTT for 10 -15 min before the addition of protein. Fluorescence was measured every minute with the Victor TM multilabel plate reader (PerkinElmer Life Sciences) with excitation at 480 nm and emission at 520 nm. To avoid evaporation during the measurement, samples were covered with 30 l of RNase-free immersion oil. Magnesium dependence was tested with 8.5 pmol of substrate, 250 ng of Sll7003 toxin, and 60 mM EDTA. Antitoxin protection was tested with 5.8 pmol of substrate, 250 ng of Sll7003, and ϳ200 ng of Ssl7004. Each experiment was assayed in technical replicates.

Determination of Cleavage Site Specificity
In vitro transcribed full-length 16 S rRNA and a 127-nt-long sll0406-asRNA fragment were incubated with purified Sll7003 protein. Cleavage products were analyzed by 5Ј and/or 3Ј rapid amplification of cDNA ends (RACE).
5Ј-RACE-5Ј-RACE was performed following Steglich et al. (39) with certain modifications. The 16 S rRNA in vitro transcript was treated with DNase (2 units of Turbo DNase, Invitrogen) and tobacco acid pyrophosphorylase (1 unit/1 mg of RNA; Epicenter). RNA was incubated with elution buffer and purified protein Sll7003 for 30 min at 37°C followed by a phenol/chloroform extraction and ethanol precipitation at each time. A synthetic RNA oligonucleotide (3 pmol) (supplemental Table S1) was ligated to Sll7003-treated RNA (320 ng) and, as a control, to mock-treated RNA using T4 RNA ligase (Fermentas; 3 units/100 ng of RNA) for 1 h at 37°C followed by phenol/ chloroform extraction and ethanol precipitation. Linker reactions with no oligonucleotide were performed as controls. For reverse transcription, linked RNA and controls were preincubated with reverse transcription primer (1 M final) at 65°C for 5 min followed by 42°C for 10 min before incubation with Omniscript reverse transcriptase (Qiagen) and a solution of all four dNTPs (0.5 mM each) in the provided reaction. Incubation was carried out at 42°C for 2 h with a final inactivation step at 95°C for 5 min. cDNA was amplified by PCR using an RNA oligonucleotide-specific primer (0.2 M) and several gene-specific primers (0.2 M) binding to different positions within the 16 S rDNA. Primers are given in supplemental Table S1. Amplified PCR fragments that appeared specific for the Sll7003treated sample were gel-excised and cloned into the pJET plasmid (Fermentas). After transformation into E. coli Top10FЈ, plasmid inserts were amplified by colony PCR and sequenced.
3Ј-RACE-In vitro transcripts were treated with DNase (2 units of Turbo DNase, Invitrogen), purified with phenol/chloroform, and incubated with elution buffer and purified protein Sll7003 for 30 min at 37°C. Samples were heated for 5 min at 95°C prior to 6 -10% 7 M urea PAGE separation. Bands of certain size ranges were cut out and gel-eluted overnight using RNA elution buffer (0.3 M NaAc; 1 mM EDTA, 0.1% SDS) fol-lowed by ethanol precipitation. 5Ј ends were blocked by dephosphorylation with FastAP (Fermentas) followed by phenol/chloroform extraction and ethanol precipitation. A 5Ј-monophosphorylated synthetic RNA oligonucleotide, with a 3Ј-dideoxynucleotide-blocked end, was ligated to elution buffer-and Sll7003-treated RNA as described under "5Ј-RACE." Reverse transcription and PCR were executed as described above. Primers are given in supplemental Table S1. Amplicons that appeared specific for the Sll7003-treated sample were gelexcised and cloned into pGEM-T (Promega). After transformation into E. coli Top10FЈ, plasmid inserts were amplified by colony PCR and sequenced.

RESULTS
TA Systems on Plasmid pSYSA of Synechocystis-We have recently characterized the main function of megaplasmid pSYSA of Synechocystis as harboring a sophisticated defense mechanism, indicated by the presence of three different types of CRISPR systems. 3 However, CRISPR systems are not essential, and pSYSA also does not possess any other gene with a known essential function; therefore the presence of TA systems preventing loss of the megaplasmid appeared likely. Indeed, Makarova et al. (13) identified six putative TA systems on plasmid pSYSA from a total of 30 type II TA pairs in Synechocystis. These TA systems are located at several distinct locations (Fig.  1), with five of the six putative toxins predicted as RNases, based on the presence of the respective domain. Three of these proteins possess PIN domains, and one protein each has a predicted RelE or MazF domain (Table 1). According to the criteria defined in a recent genome-wide mapping experiment of transcriptional start sites (TSS) (40), all these TA pairs appear to be transcribed, and distinct TSS could be identified (Table 1). It should be noted that the TSS of the ssl7039 toxin gene coincides with the start codon, pointing to a possible leader-less initiation of translation. For the ssl7007_sll7006 module, we noticed with 207 nt a quite long 5Ј-UTR based on the existing annotation. Moreover, the annotated start of ssl7007 is an alternative start codon (GTG), and there is not a single stop codon within the 5ЈUTR. Comparing the possible translation product of a polypeptide 5Ј-extended by 60 amino acids revealed a very high degree of conserved residues (Ͼ85% positive and Ͼ65% identical positions) with predicted protein sequences from several other cyanobacteria, such as SYNPCC7002_F0004, N9414_16309, or CWATWH0003_5642 from Synechococcus sp. PCC 7002, Nodularia spumigena CCY9414, and Crocosphaera watsonii WH 0003. We conclude that the annotation of ssl7007 should be 5Ј-prolonged and that the start of translation likely occurs from an ATT alternative start codon at positions 3766 -3768 on the reverse strand.
We also scanned for possible type I TA systems on pSYSA and identified with ssr7036 a gene coding for a short protein of high predicted pI (9.05) associated with very strong antisense transcription (Table 1), hence constituting candidacy for a type I TA system. Furthermore, pSYSA encodes at least three freestanding TA systems components, with ssl7046 and ssl7048 having close homologs on the plasmid pSYSG that are organized in the typical arrangement of a TA operon (genes sll8027 and ssl8028).
Expression of pSYSA TA Systems in E. coli Causes Growth Inhibition That Is Attenuated by the Associated Antitoxin-To test whether the pSYSA-borne potential TA pairs are bona fide TA systems, we heterologously expressed the toxin and antitoxin from different inducible promoters in E. coli. The respective toxins were cloned into the IPTG-inducible expression vector pQE70, resulting in the plasmid pQE70::toxin. The associated antitoxins were cloned into the arabinose-inducible expression vector pBAD, resulting in the plasmid pBAD::antitoxin. Five of the six predicted toxin-antitoxin systems (sll7003-ssl7004; ssl7007-sll7006; sll7033-sll7034; ssl7039-ssl7038; and slr7041-ssr7040) were successfully cloned into the expression vectors. For the sixth TA system (sll7030-sll7031), no positive clones were obtained, whereby it was excluded from further analysis. The different E. coli strains containing both pQE70::toxin and pBAD::antitoxin were grown in LB broth plus the respective antibiotics until they reached a sufficient optical density. To test the effects separately for toxin and antitoxin expression, as well as for the toxin-antitoxin coexpression, we diluted the cultures 1:100 in fresh LB (with the respective antibiotics) plus the corresponding inducer and measured the optical density over time in intervals of 15 min. Each of the strains showed an inhibition in bacterial growth when the toxin was expressed alone as compared with strains with no expression or lone expression of the antitoxin (Fig. 2). Co-expression of toxin and antitoxin attenuated the growth inhibitory effect for all five systems tested (Fig. 2). We conclude that all five tested systems exhibit typical properties of a TA system in vivo. In the case of Ssl7007, we observed toxicity despite the fact that the toxin was expressed according to the annotated reading frame and hence might have been missing about 60 residues at its N terminus.
Toxin Sll7003 Causes Growth Arrest and Decreases Colonyforming Units-We selected the toxic component Sll7003 to further test the toxicity of the protein by heterologously overexpressing it in E. coli strain pQE70::sll7003. The inducer IPTG was added at an appropriate optical density to ensure a sufficient amount of cell material for colony formation. Start of the Sll7003 expression arrested the bacterial growth for several hours (Fig. 3A). Toxicity was also seen by a significant decrease of colony-forming units over time (Fig. 3B).
The PIN Protein Sll7003 Is a Magnesium-dependent Ribonuclease-Ribonuclease activity of the purified PIN protein Sll7003 was tested with two different substrates. In the RNaseAlert TM nuclease detection system, a fluorescence-quenched oligonucleotide probe is used as substrate that emits a fluorescent signal after nuclease degradation. Upon the addition of the purified protein Sll7003, we observed a rapid increase of fluorescence over time, whereas there was no increase of fluorescence in the absence of Mg 2ϩ (Fig. 4A). To verify the ribonuclease activity of the purified toxin Sll7003 further, in vitro synthesized Synechocystis 16 S rRNA was used as natural substrate. The substrate RNA was incubated with purified toxin and analyzed by PAGE separation. As shown in Fig.  4B, the 16 S rRNA substrate was cleaved into several fragments by the toxin, resulting in a distinct cleavage pattern.
Antitoxin Ssl7004 Diminishes Toxin Sll7003 RNase Activity in Vitro-Co-expression of the respective toxin and antitoxin in vivo showed that the growth inhibition effect of the toxin could be attenuated by the antitoxin (Fig. 2). Therefore we tested whether the proposed toxic RNase activity of the purified toxin Sll7003 can be abolished by the purified antitoxin Ssl7004 in vitro. As a first substrate, we again used the RNaseAlert TM nuclease detection system. RNase activity of Sll7003 can be seen by a rapid increase of fluorescence over time after the addition of the toxin as compared with an even level of fluorescence in the buffer and Ssl7004 antitoxin control. Simultaneous addition of both toxin and antitoxin resulted in a significant lower increase of fluorescence (Fig. 5A). Furthermore, we investigated the diminishing effect of the antitoxin by incubating an in vitro synthesized RNA substrate with the purified toxin and antitoxin at the same time. Incubation with the toxin alone resulted in a distinct cleavage pattern, which could only be seen in a significantly extenuated form during synchronous incubation with toxin and antitoxin (Fig. 5B).
In Vitro Specificity of Sll7003 Ribonuclease-To determine the RNase specificity of the purified protein Sll7003, we applied 5Ј and 3Ј RACE on two different in vitro transcripts as substrates. We chose the 16 S rRNA and sll0406-as, a 127-nt RNA fragment antisense to the chromosomal gene sll0406, as substrates. The latter encodes another PIN domain-containing protein and is associated with a natural antisense RNA, but was selected here as an example for a short noncoding RNA. In total, we identified 34 cleavage sites, 31 in the 16 S rRNA and three in sll0406-as. Sequence analysis of these sites did not lead to the identification of a preferentially used recognition motif. However, the results indicated that cleavage is favored on the 3Ј side of purine residues (ϳ82% of the 34 sites).

TABLE 1 TA systems on pSYSA according to Shao et al. (14), Makarova et al. (13), and data obtained in this work
In addition to the respective gene loci IDs and the respective protein domains, the start site of transcription (TSS) is given according to the reference sequence (GenBank accession number NC_005230) or the actual sequence of the PCC-M substrain used in this work (GenBank accession CP003267), with an F indicating the location on the forward and an R indicating the location on the reverse strand. Moreover, the distance between the TSS and the annotated start of translation is given, as well as the type of TA system and the respective arrangement of toxin (T) and antitoxin (AT) component. Symbols and domain names: NK, not known; AbrB, AidB regulator domain; COG2442, Cluster of Orthologous Genes 2442; GNAT, GCN5-related N-acetyltransferases; PIN, PIN domain (homologues of the "pilT N"-terminal domain); RHH, ribbon-helix-helix; Xre, xenobiotic response element. sll0406-as, RNAfold prediction) showed that single-strand RNA regions were preferred (Fig. 6). Therefore we tested the capability of Sll7003 to cleave doubleversus single-stranded RNA molecules under defined conditions in vitro. The 127-nt sll0406-as fragment was annealed with the complementary sequence to create an RNA-RNA duplex. RNase activity was then tested against both ssRNA substrates and the dsRNA substrate. As shown in Fig. 7, Sll7003 was able to cleave the two single-stranded RNA substrates but not the double-stranded RNA target. We conclude that Sll7003 is a single-strand specific RNA endonuclease.

DISCUSSION
Here we have characterized the activity of the pSYSA-located Sll7003-Ssl7004 TA system. The toxic component Sll7003 is a Mg 2ϩ -dependent RNase, which is consistent with its annotation as a PIN domain-containing VapC-type protein (42)(43)(44).   The activity of Sll7003 is inhibited in the presence of the protein Ssl7004, characterizing it as the corresponding antitoxin. Sll7003 activity was not unspecific but exhibited a wide range of accepted substrate configurations, with a clear preference for single-stranded RNA regions and a slight preference for cleavage on the 3Ј side of purine residues. However, the fact that it cleaved the 16 S rRNA more than 30 times and the fact that it still cleaved the shorter 127-nt-long artificial substrate sll0406-as at three sites underscores its activity as a fairly wide range RNA endoribonuclease, at least in vitro. In contrast, the analysis of VapC of the Shigella flexneri 2a plasmid and of Vap-CLT2 of Salmonella enterica serovar Typhimurium LT2 in vivo indicated that enterobacterial VapC toxins effectively act as tRNases inhibiting translation globally by the site-specific cleavage of the initiator tRNA tRNA fMet between the anticodon stem and loop (34).
We demonstrate that pSYSA harbors at least five functional type II TA systems. However, in the genome-wide analysis of 750 prokaryotic genomes, Makarova et al. (13) identified six likely type II TA pairs on the Synechocystis pSYSA plasmid. Additionally, there is evidence for a possible type I TA system on this plasmid and at least three freestanding TA components (Table 1). In the TADB database of type II TA systems (14), 150 of 337 analyzed prokaryotic plasmids encode only a single TA system. In contrast, only 18 plasmids (5.34%) are listed to possess six or more type II TA systems. Thus, pSYSA is in this top category regarding the number of TA systems. However, extending these bioinformatic predictions by inferring that these are actually transcribed, and for five TA systems, proving their actual toxicity/antitoxicity, we present to our knowledge the highest number of simultaneously active TA systems on any plasmid described thus far. These numbers indicate a high positive selection for possible plasmid maintenance systems on pSYSA, suggesting that it is a very relevant cyanobacterial plasmid. Although pSYSA does not contain any known gene in the classical sense, it has recently been characterized as a major defense plasmid, devoting 75% of its coding capacity to encode all components of three independent CRISPR systems. 3  In vitro synthesized Synechocystis 16 S rRNA was incubated for 30 min with purified toxin Sll7003. Resulting RNA fragments were used to determine the cleavage sites by 5Ј and 3Ј RACE, indicated by the arrows. The RACE results indicate that toxin Sll7003 has an in vitro preference for ssRNA.