Type I toxin-antitoxin systems contribute to mobile genetic elements maintenance in Clostridioides difficile and can be used as a counter-selectable marker for chromosomal manipulation

Toxin-antitoxin (TA) systems are widespread on mobile genetic elements as well as in bacterial chromosomes. According to the nature of the antitoxin and its mode of action for toxin inhibition, TA systems are subdivided into different types. The first type I TA modules were recently identified in the human enteropathogen Clostridioides (formerly Clostridium) difficile. In type I TA, synthesis of the toxin protein is prevented by the transcription of an antitoxin RNA during normal growth. Here, we report the characterization of five additional type I TA systems present within phiCD630-1 and phiCD630-2 prophage regions of C. difficile 630. Toxin genes encode 34 to 47 amino acid peptides and their ectopic expression in C. difficile induces growth arrest. Growth is restored when the antitoxin RNAs, transcribed from the opposite strand, are co-expressed together with the toxin genes. In addition, we show that type I TA modules located within the phiCD630-1 prophage contribute to its stability and mediate phiCD630-1 heritability. Type I TA systems were found to be widespread in genomes of C. difficile phages, further suggesting their functional importance. We have made use of a toxin gene from one of type I TA modules of C. difficile as a counter-selectable marker to generate an efficient mutagenesis tool for this bacterium. This tool enabled us to delete all identified toxin genes within the phiCD630-1 prophage, thus allowing investigation of the role of TA in prophage maintenance. Furthermore, we were able to delete the large 49 kb phiCD630-2 prophage region using this improved procedure.


50
Clostridioides difficile is a medically important human enteropathogen that became a key 51 public health concern over the last two decades in industrialized countries 1, 2 . This strictly 52 anaerobic spore-forming Gram-positive bacterium is a major cause of antibiotic-associated 53 nosocomial diarrhoea in adults 3 . The main virulence factors of C. difficile are two toxins, 54 TcdA and TcdB, produced by all toxigenic strains 4 . A binary toxin CDT is also present in nature and the mode of action of the antitoxin led to the classification of TA modules into six 81 types 11 . In type I systems, the antitoxin is a small antisense RNA targeting toxin mRNA for 82 degradation and/or inhibition of translation, while in type III systems, the antitoxin RNA 83 binds directly to the toxin protein for neutralization 12,13 . For other TA types, both the toxin 84 and the antitoxin are proteins. In most studied type II TA systems, the proteinaceous antitoxin 85 forms a complex with its cognate toxin leading to toxin inactivation 14 . Major functions of TA 86 modules include plasmid maintenance, abortive phage infection and persister cell formation 87 15,16,17,18,19,20 . TA loci are commonly found on mobile genetic elements, in particular 88 plasmids in which they were initially discovered and extensively studied. However, the roles 89 of chromosomally-encoded TA modules, including those within prophage genomes, remain 90 largely unexplored. Likewise, the individual contribution from each of the chromosomal TA 91 modules in bacteria remains to be investigated. 92 We recently reported the identification of the first type I TA systems associated with CRISPR 93 arrays in C. difficile genomes 21 . The co-localization and co-regulation by the general stress 94 response Sigma B factor and biofilm-related factors suggested a possible genomic link 95 between these cell dormancy and adaptive immunity systems. Interestingly, two of these 96 functional type I TA pairs are located within the homologous phiCD630-1 and phiCD630-2 97 prophages in C. difficile strain 630. In the present work, we describe the identification of 98 additional type I TA modules highly conserved within C. difficile prophages and provide 99 experimental evidence of their contribution to prophage maintenance and stability. 100 The unique properties of type I TA systems offer interesting possibilities for biotechnological 101 and therapeutic applications. We demonstrate in the present work that inducible toxicity 102 caused by type I toxins largely improve the efficiency of allele exchange genome editing 103 procedures by promoting the elimination of plasmid-bearing cells. 104 Altogether, our data provide important insights into the function and possible applications of 105 type I TA systems in C. difficile. 106

107
Identification of novel type I TA pairs in C. difficile 108 Multiple TA modules have been discovered in bacterial chromosomes including 109 prophage regions 11 . In C. difficile, we have recently identified several type I TA pairs 110 adjacent to CRISPR arrays, two of them being located inside the phiCD630-1 and phiCD630- often overlapping coding regions. We therefore used the tBlastn program (protein query to 118 search against translated genomes) using a previously identified type I toxin CD0956.2 as a 119 query, as it can overcome annotation and ORF detection defects. We identified gene 120 CD0977.1 and two other novel putative genes, unannotated on the genome, that we named 121 CD0904.1 and CD0956.3. These genes code for small proteins of 47, 35 and 34 amino acids, 122 respectively ( Fig. 1A). Prophages phiCD630-1 and phiCD630-2 share a large region of 123 homology with almost identical sequences, which include a duplication of CD0977.1 and 124 CD0956.3 (Fig. 1B). In contrast, CD0904.1 is unique to phiCD630-1 and no other toxin gene 125 homolog could be identified within phiCD630-2. Transcript reads were detected in regions of 126 these putative genes by RNA-seq 22 (Fig. S1A and S1B) and a consensus RBS sequence 127 (AGGAGG) was present 7-8 nucleotides upstream of the respective ATG start codons, 128 suggesting that the corresponding proteins might be produced. In addition, all three putative 129 proteins carried a hydrophobic N-terminal region and a positively charged tail, which are 130 characteristic features of type I toxins (Fig. 1A). Analysis of our previous TSS mapping data 131 22 and sequence alignments (Fig. S1C) suggested the presence of potential antisense RNAs of 132 these toxin-encoding genes with the presence of TSS associated with Sigma A-and Sigma B-133 dependent promoter elements in both sense and antisense directions (Fig. S1). Antitoxins of 134 CD0977.1, CD0904.1 and CD0956.3, located on phiCD630-1, were hereafter named RCd11,

137
To determine whether these novel potential TA pairs are functional, pRPF185- Intriguingly, predicted riboswitches responding to the c-di-GMP signalling molecule, cdi1_4 149 and cdi1_5, precede RCd11 and RCd12 antisense RNAs 22 . We therefore sought to further 150 characterize the CD0977.1-RCd11 TA pair. In agreement with the data above, addition of 151 ATc to liquid cultures in exponential growth phase led to an immediate growth arrest of 152 strain 630Δerm/pT, unlike 630Δerm/p (Fig. S2A). In addition, the growth arrest was 153 accompanied by a drop of colony-forming units (CFUs) (Fig. S2B). Similarly to previous 154 observations with other C. difficile type I TA modules 21 , the analysis of liquid cultures by 155 light microscopy showed that toxin overexpression was accompanied by an increase in cell 156 length in about 10% of the cells (Fig. S2C). Their length was above the mean length value of 157 630Δerm/p control strain with two standard deviations (10.5 µm). The co-expression of the 158 entire TA module led to the partial reversion of this phenotype.

159
Using Northern blotting, we detected both toxin and antitoxin transcripts in the 630Δerm/p, Northern blotting using a riboswitch-specific probe. In contrast, abundance of toxin 170 transcripts was not affected by fluctuations of c-di-GMP levels. 171 We then mapped the transcriptional start (TSS) and termination sites for the genes of  Table S3). The results obtained agreed generally well with the transcript lengths deduced 174 from TSS mapping, RNA-seq and Northern blot. Taken together, these data suggest the 175 presence of two tandem TSS for RCd11, i.e. P1 associated with c-di-GMP-dependent 176 riboswitch and P2 located downstream from the riboswitch (Fig. S1).

177
It is in the nature of type I antitoxins to be short-lived in contrast to the stable toxin 178 mRNA 13 . To determine the half-lives of toxin and antitoxin RNAs of the CD0977.1-RCd11 179 module, C. difficile strains were grown in TY medium until late-exponential phase and 180 rifampicin was added to block transcription. Samples were taken at different time points after 181 rifampicin addition for total RNA extraction and Northern blot analysis with toxin and 182 antitoxin-specific probes. In a control strain 630Δerm/p carrying an empty vector, the half- antitoxin RCd11 RNA, we also observed a stabilization in strain depleted for RNase Y (Fig. 190 S4) suggesting that this ribonuclease could contribute to antitoxin RNA degradation.

191
To confirm the protein nature of CD0977.1 and assess its subcellular localization, we  Fig. S5A). In contrast, the strain carrying pDIA6817 did not grow in the 220 presence of 10 or 100 ng/ml ATc, similarly to the negative control strain. Similar results were 221 obtained when the strains were grown in an automatic plate reader for 20 h in liquid medium 222 in the presence of 5 ng/ml ATc ( Fig 3A). Interestingly, induction of toxin expression on BHI 223 plate with a lower dose of ATc (5 ng/ml) led to a partial reversion of the growth defect of the 224 strain carrying pDIA6816, unlike the negative control strain (Fig. S5A). These results suggest 225 that the short antitoxin transcript driven by promoter P2 is crucial for the efficient inactivation 226 of the toxin, while the longer antitoxin transcript directed by P1 is dispensable.

TA systems are involved in maintenance of phiCD630-1 in the host cells 275
Because the loss of an integrated phage from cells first requires its excision from the host 276 genome, we sought to determine whether spontaneous excision of phiCD630-1 from 277 chromosomal DNA occurred. To do so, we performed a PCR on genomic DNA from C. plates. Nearly 100% of cells from both strains were found to still be resistant to erythromycin 294 in these conditions, indicating that they had retained the prophage ( Fig S10D).

295
In an attempt to artificially increase the excision rate of phiCD630-1, we ectopically suggesting that TA systems do not affect phiCD630-1 excision (Fig. S11).

304
Strains ΔphiCD630-2 phiCD630-1::erm and ΔphiCD630-2 phiCD630-1-ΔT4::erm carrying 305 pDIA6867 were then inoculated at an initial OD600 of 0.005 in TY supplemented with 7.5 306 μg/ml Tm and 10 ng/ml ATc. After 24 hrs of incubation at 37°C, measurement of the OD600 307 revealed a dramatic growth defect of ΔphiCD630-2 phiCD630-1::erm compared to the 308 ΔphiCD630-2 phiCD630-1-ΔT4::erm (Fig. 5B). In addition, plating of the cells bearing 309 pDIA6867 on non-selective and erythromycin-containing agar plates revealed that 310 phiCD630-1::erm was still present in more than 90% of the total population while 311 phiCD630-1-ΔT4::erm remained in less than 10 % of the cells (Fig. 5C). These results thus 312 show that TA modules are important for phiCD630-1 maintenance after its excision and 313 highlight the impact of the toxin expression on the cell growth upon the loss of prophage.

315
Type I TA are prevalent in C. difficile phage genomes 316 Since we identified additional toxin variants of type I TA systems after careful inspection of 317 phiCD630-1 full genome, we decided to re-scan for possible ORFs in every available phage 318 genomes of C. difficile using the permissive algorithm of the NCBI ORFfinder software.

319
ORFs with minimal length of 60 nucleotides as well as nested ORFs were detected. A blastP 320 search against the corresponding proteins allowed the identification of toxin homologs in all 321 C. difficile prophage genomes (functional phages) (Fig. 6). Moreover, toxin sequence 322 alignments revealed the high conservation of the hydrophobic N-terminal region, as well as 323 the lysine-rich, positively charged region at the C-terminus. Hence, these data suggest the 324 functionality of the toxins and reinforce their proposed role for phage maintenance and 325 preservation.

326
Despite these conserved regions, alignment of toxins also revealed small variations among 327 sequences. We therefore sought to explore the possible relationship between phage 328 phylogeny and the observed toxin variants. A whole genome comparison of all phages 329 included in this study was performed to create phylogenetic groups (phiCD119-like viruses,

333
Interestingly, an extended search outside C. difficile phages revealed the presence of other 334 toxin homologs inside plasmids of C. difficile and Paeniclostridium sordellii, a closely 335 related species (Fig. S13). These findings imply that C. difficile phages could recombine with 336 plasmids to exchange genetic material, as already proposed for E. coli phages 27, 28 .

338
In this study, we identified and characterized novel functional type I TA modules in C.  It has been proposed that phages could recombine with plasmids during infection of the same 413 or different bacterial species to exchange genetic material 8,27,28 . It is therefore tempting to 414 speculate that this TA system could originate from P. sordellii, or at least that it has the 415 ability to disseminate, through horizontal gene transfer involving conjugation and 416 recombination, from one species to another. Intriguingly, it was previously noticed that a 1.9-417 kb region could have been transferred from the plasmid of a C. difficile 630 strain to the 418 phiCD38-2 prophage 8 . It was suggested that this recombination event had led to the 419 acquisition of parA, a gene assumed to help the newly created chimeric phage to 420 autonomously replicate and segregate as a circular plasmid. Our analysis of TA systems in 421 phiCD38-2 shows that this 1.9-kb region also carried a TA (gp33) that presumably 422 contributes to the phage maintenance and stability. It is interesting to observe that TA encoding regions can relocate from one mobile genetic element to another in this fashion, and 424 that genes in proximity to the TA being transferred (i.e. parA gene) have more chances to 425 become fixed in the newly integrated DNA. In the latter case, the region transferred seems to  Table   463 S1. C. difficile strains were grown anaerobically (5 % H2, 5 % CO2, and 90 % N2) in TY 40 or 464 Brain Heart Infusion (BHI, Difco) media in an anaerobic chamber (Jacomex). When All primers used in this study are listed in Table S2 (Table S2). For expression of the CD0912 excisionase, the 512 CD0912 gene (-37 to + 333 relative to the translational start site) was amplified by PCR and 513 cloned into the linearized pDIA6103 vector, yielding pDIA6867.

Mutagenesis approach and mutant construction 519
To improve the efficiency of the allele exchange mutagenesis in C. difficile, we made use of 520 the inducible toxicity of the CD2517.1 type I toxin that we previously reported 21 . To

532
The homology arms were amplified by PCR from C. difficile strain 630 genomic DNA (Table   533 S1) and purified PCR products were directly cloned into the PmeI site of pMSR vector using  (Table S1). The Pthl-ermB cassette was amplified from the Clostron mutant cwp19 24, 43 .

538
Purified PCR products were all assembled and cloned together into the PmeI site of pMSR 539 vector using NEBuilder HiFi DNA Assembly.

540
All pMSR-derived plasmids were initially transformed into E. coli strain NEB10β and all 541 inserts were verified by sequencing. Plasmids were then transformed into E. coli HB101 542 (RP4) and transferred by conjugation into the appropriate C. difficile strains. The adopted 543 protocol for allele exchange was similar to that used for the codA-mediated allele exchange 544 23 , except that counter-selection was based on the inducible expression of the CD2517.1 toxin 545 gene. Transconjugants were selected on BHI supplemented with Cs, Cfx and Tm, and then 546 restreaked onto fresh BHI plates containing Tm. After 24 h, faster-growing single-crossover integrants formed visibly larger colonies. One such large colony was restreaked once or twice 548 on BHI Tm plate to ensure purity of the single crossover integrant. Purified colonies were 549 then restreaked onto BHI plates containing 100 ng/ml ATc inducer to select for cells in which 550 the plasmid had been excised and lost. In the presence of ATc, cells in which the plasmid is 551 still present produce CD2517.1 at toxic levels and do not form colonies. Growing colonies 552 were then tested by PCR for the presence of the expected deletion.   (Table S2), which only results in PCR products when the prophage is excised.

591
PhiCD630-1 stability assays 592 Overnight cultures of C. difficile strain 630 ΔphiCD630-2 phiCD630-1::erm and 593 ΔphiCD630-2 phiCD630-1-ΔT4::erm were used to inoculate 10 ml of TY broth at an initial 594 OD600 of 0.05. Every 10 to 14 h, cultures were subcultured at an initial OD600 of 0.05. After 595 four passages, the cultures were serially diluted and plated on BHI plates to estimate the total 596 CFUs and on BHI plates supplemented with 2.5 µg/ml erythromycin to determine the number      to produce the heatmap of genome similarity. Similarity scores are based on a fragmented all-against-777 all pairwise alignment using BLASTn and the accurate alignment option (fragment size, 200; step 778 size, 100). The colours reflect the similarity, ranging from low (red) to high (green). Phages were 779 assigned to a genus if they clustered closely to another phage previously described as a member of 780 that genus. The phylogenetic tree is based on the sequence similarity scores from the same whole-781 genome comparison and was constructed using the neighbour joining method with the SplitsTree4 782 software (v 4.13.1). 783