Neighbor predation linked to natural competence fosters the transfer of large genomic regions in Vibrio cholerae

Natural competence for transformation is a primary mode of horizontal gene transfer. Competent bacteria are able to absorb free DNA from their surroundings and exchange this DNA against pieces of their own genome when sufficiently homologous. However, the prevalence of non-degraded DNA with sufficient coding capacity is not well understood. In this context, we previously showed that naturally competent Vibrio cholerae use their type VI secretion system (T6SS) to actively acquire DNA from non-kin neighbors. Here, we explored the conditions of the DNA released through T6SS-mediated killing versus passive cell lysis and the extent of the transfers that occur due to these conditions. We show that competent V. cholerae acquire DNA fragments with a length exceeding 150 kbp in a T6SS-dependent manner. Collectively, our data support the notion that the environmental lifestyle of V. cholerae fosters the exchange of genetic material with sufficient coding capacity to significantly accelerate bacterial evolution.

example by first absorbing foreign DNA and then recombining it into their genome. In this study, 48 significant differences in the pandemic A1552 strain compared to the environmental isolate 175 Sa5Y in terms of the absence/presence of genomic features and single nucleotide polymorphisms 176 (SNPs) in core genes that would allow us to measure HGT events occurring between the strains, 177 and several of these major differences are highlighted here. Indeed, as expected from its non-178 clinical origin, the environmental isolate lacked several genomic regions, including those that 179 encode major virulence features, namely Vibrio pathogenicity islands 1 and 2 (VPI-1, VPI-2), 180 Vibrio seventh pandemic islands I and II (VSP-I, VSP-II (45)), the cholera toxin prophage CTX 181 (46), and the WASA-1 element. In addition, the strain's O-antigen cluster differed significantly 182 from the O1-encoding genes of pandemic strain A1552 (Fig. 2). The region that differed the 183 most between both strains was the integron island, which is consistent with the role of this 184 assembly platform in fostering the incorporation of exogenous open reading frames (47). Given 185 these major differences between strain A1552 and Sa5Y and, in addition, an overall SNP 186 frequency of approximately 1 in 55 nucleotides for conserved genes, we concluded that HGT 187 events occurring between these two strains on chitinous surfaces could be precisely scored using 188 short-read sequencing. Apart from this important genomic information, we also noted that the 189 pandemic strains as well as Sa5Y contained previously unrecognized rRNA operons, with nine 190 or ten rRNA clusters in total compared to the initially reported eight (44). 191 192 Released DNA from T6SS-killed prey leads to the transfer of large genomic regions 193 As our previous study witnessed gene transfers between V. cholerae bacteria (5) though neither 194 scored the full extent of the transferred DNA region nor took T6SS-mediated neighbor predation 195 into consideration, we sought to next determine how much genetic material would be absorbed 196 and integrated by competent V. cholerae upon neighbor predation. To do this, we co-cultured the 197 predator (A1552) and prey (Sa5Y) strains on chitinous surfaces for 30 h without any deliberate 198 selection pressure. To be able to afterwards screen for the transfer of at least one gene, we first 199 integrated an aph cassette within the vipA of strain Sa5Y, which concomitantly deactivated the 200 prey's T6SS, to select kanamycin-resistant transformants of strain A1552. Using this system, 201 resistant transformants of A1552 were selected at an average frequency of 1.8 x 10 -4 after the 202 30 h co-culturing on chitin (Fig. S3), and 20 of those transformants were randomly picked for 203 further analysis. After three independent experiments, the whole genome of each of the 60 204 transformants was sequenced, and the reads were mapped to either the predator's or the prey's 205 genome sequence (see SI Appendix for detailed bioinformatic analysis). As shown in Figure 3, 206 apart from the common acquisition of the aph resistance cassette, the location and the size of the 207 prey-donated genomic region differed significantly between most transformants. Previous 208 estimates of the average length of total acquired DNA were made in experiments using purified 209 donor gDNA and were considered to be ~23 kbp (40). Importantly, we observed in these new 210 experiments that the average length of the total acquired DNA, meaning the DNA surrounding 211 the aph cassette plus any transferred regions elsewhere on either of the two chromosomes ( Fig.  212 S4), was almost 70 kbp and therefore significantly larger than the previous estimates. Around 15% 213 of all transformants acquired and integrated more than 100 kbp (Fig. 3B), which was previously 214 considered unlikely due to absence of such long DNA fragments in the environment. Consistent 215 with the principle of natural transformation, it should be noted that the new DNA was acquired 216 through double homologous recombination such that it replaced the initial DNA region and the 217 overall genome size did not significantly change. Further analysis indicated that about 50% of 218 the strains experienced a single HGT event around the aph cassette, while the others exchanged 219 regions in up to eight different locations on the two chromosomes (Fig. 3C). Finally, we 220 analyzed the length of continuous DNA stretches that were acquired from the prey and observed 221 that those ranged from a few kbp up to 168 kbp (Fig. 3D). Collectively, these data indicate that V. 222 cholerae can acquire large genomic regions from killed neighbors with an average exchange of 223 more than 50 kbp or ~50 genes. This finding contradicts the notion that natural transformation 224 -10 -cannot serve for DNA repair or acquisition of new genetic information due to the insufficient 225 length and coding capacity of the acquired genetic material. 226 227

Transformation by purified DNA only occurs if correctly timed 228
To better understand the DNA acquisition and integration potential of naturally competent V. 229 cholerae, we next compared the data described above, which we refer to from now on as 230 condition 1 using experiments varying the aspects of neighbor predation and DNA 231 supplementation (Fig. 4A). First, the acceptor strain was grown in a monoculture immediately 232 supplemented with purified genomic DNA (gDNA) derived from the same donor (prey) strain as 233 described above. Notably, when the gDNA was added at the start of the chitin-dependent culture, 234 no transformants were reproducibly detected from three independent biological experiments, 235 suggesting that free DNA is rapidly degraded under such conditions. This finding is consistent 236 with our previous work in which we demonstrated that V. cholerae produces an extracellular and 237 periplasmic nuclease Dns (16, 20) that degrades transforming material. At high cell density 238 (HCD), where competence is induced, dns is partially repressed through direct binding of HapR 239 (16,17), and this repression is reinforced by the transcription factor QstR (17, 18). We therefore 240 concluded that the simultaneous expression of both machineries, concomitantly with a strong 241 repression of dns, is a prerequisite for successful DNA transfer. Indeed, such coordinated 242 expression would ensure that T6SS-mediated attacks are exquisitely timed with low nuclease 243 activity so that the prey-released DNA can be efficiently absorbed. 244 As we previously showed that the addition of purified gDNA after ~20-24 h of growth on 245 chitin wasn't prone to degradation by Dns (48), we next choose this time point to probe the DNA 246 acquisition capability using purified DNA (condition 2; Fig. 4A). Doing so led to similar 247 transformation frequencies as those observed for the prey-released DNA caused by T6SS attacks 248 (condition 1; Fig. S3A). WGS of 20 transformants from two biologically independent 249 -11 -experiments likewise resulted in similar DNA acquisition patterns with average and maximum 250 DNA acquisitions of 70 kbp and 188 kbp, respectively, and the presence of multiple exchanged 251 regions of varying sizes ( Fig. 4 and Fig. S5). While we cannot entirely exclude that the 252 maximum length of individual DNA stretches was biased by the purification step, despite the 253 fact that we chose a method that was designed for chromosomal DNA isolation of 20-150 kbp 254 sized fragments (see methods), our results suggest that the maximum DNA acquisition length of 255 single fragments is probably reached between 100-110 kbp (Fig. S5). Moreover, the comparable 256 acquisition patterns between conditions 1 and 2 ( Fig. 4) imply that the prey-released DNA in 257 condition 1 is neither heavily fragmented nor is its accessibility or absorption by the competent 258 acceptor bacterium significantly hindered due to, for example, DNA-binding proteins. 259 260

Prey-exerted T6SS counter attacks do not change the DNA transfer pattern 261
Since the aph cassette was located within the T6SS sheath protein gene vipA in the above 262 experiments, we wondered if this T6SS inactivation biased the DNA transfer efficiency. We 263 therefore repeated the above-described experiments using prey strains that carried the aph 264 cassette on the opposite site of chr 2 (within gene VCA0747; condition 3). As shown in Figure  hypothesized that the now-restored T6SS-mediated killing capacity of the prey led to the 273 additional release of genomic DNA from the predator, which interfered with the uptake of prey-274 -12 -released DNA. To test this idea, we repeated condition 3 (e.g., aph within VCA0747) though 275 again inactivated the T6SS of the prey using a non-selected marker (cat; condition 4), expecting 276 the results to be similar to those of condition 1 if this hypothesis was correct. No statistically 277 significant differences were observed between both conditions (3 and 4) for all tested 278 characteristics including transformation frequency (Fig. S3B), number of exchanges, and 279 separate and collective length ( Fig. 4 and Figs. S6-S7), suggesting that predator-released DNA 280 does not interfere with the predator's overall transformability by the prey-released DNA. 281 However, we acknowledge that the technical limitations of the experimental setup did not allow 282 the identification of complete revertants that first acquired and then again lost the aph cassette. 283 284 T6SS-independent prey lysis rarely triggers DNA transfer and results in shorter DNA 285 exchanges 286 We next tested whether T6SS-mediated DNA release impacted the length of the exchanged 287 region, which would support the above speculation that the intimate co-regulation of type VI 288 secretion, nuclease repression, and DNA uptake ensures that freshly released DNA is rapidly 289 absorbed by the predator and is therefore less prone to fragmentation. Such co-regulation would 290 not hold true for T6SS-independent DNA release as a result of random cell lysis, so we tested the 291 transfer efficiency of the aph cassette under conditions in which both donor and acceptor strains 292 were T6SS-defective (condition 5; Fig. 4 and S8). Under such conditions, the transformation 293 frequency dropped by 99.7% (Fig. S3B), and WGS of 2 x 20 of these rare transformants showed 294 significant differences. Indeed, the average and maximal length of acquired DNA (Fig. 4A) and 295 the number of exchanged regions (Fig. 4B) were significantly different when T6SS+ versus 296 T6SS-acceptor strains were compared, with the latter exchanges never exceeding four events 297 compared to up to 13 events for T6SS-mediated DNA release (Figs. S7 and S8). Based on these 298 data, we conclude that T6SS-mediated DNA acquisition not only increases the transfer efficiency 299 by ~100-fold but also fosters the exchange of multiple DNA stretches of extended lengths. 300 301 T6SS-mediated DNA exchanges are not limited to the small chromosome 302 The experiments described above were designed to primarily score the transfer efficiency of 303 DNA fragments localized on chr 2. The rationale behind this approach was a recent population 304 genomic study on Vibrio cyclitrophicus that suggested the mobilization of the entire chr 2 and 305 caused the authors to speculate: "how often and by what mechanism are entire chromosomes 306 mobilized?" (49). In the current study, we were unable to experimentally show such large 307 transfer events. We considered four potential reasons for the absence of such large transfers: 1) 308 mild fragmentation of prey-released DNA that excluded fragments above ~200 kb; 2) limited 309 DNA uptake and periplasmic storage capacity of the acceptor strain (20, 21); 3) limited 310 protection of the incoming single-stranded DNA by dedicated proteins (such as Ssb and DprA; 311 (50, 51)); or 4) lethality of larger exchanges due to the presence of multiple toxin/antitoxin 312 modules within the integron island on chr 2 of V. cholerae (52). While technical limitations did 313 not allow us to address the first three points, we followed up on the last idea by repeating the 314 above-described experiments using prey strains in which the aph cassette was integrated on the 315 large chromosome 1 (inside lacZ). We used these to test three (co-)culture conditions in which 316 the prey strain was either T6SS-positive (condition 6), T6SS-negative (condition 7), or 317 replaced by purified gDNA (condition 8; Fig. 4A). As shown in Figure S3, the aph cassette was 318 again transferred with high efficiency from the killed prey strain to the acceptor strain. However, 319 comparing conditions 6 (co-culture conditions) and 8 (prey-derived purified gDNA as 320 transforming material) revealed a small but significant transformation increase (~ 4-fold; Fig.  321 S3A). Based on these data, we speculate that the larger size of chr 1 (~3 Mb) compared to chr 2 322 (~1 Mb) slightly lowers the probability of acquiring the aph cassette when released from killed 323 -14 -prey. This effect becomes negligible when purified gDNA is provided, most likely due to the 324 size constraints of the purification procedure (max. 150 kb). Consistent with this idea was the 325 finding that purified gDNA from all those prey strains described in this study resulted in the 326 same level of transformation no matter where the selective marker was located (Fig. S3D). 327 Next, we randomly picked 20 transformants from two biologically independent experiments 328 for each of these three experimental conditions (6 to 8) and sequenced their genomes (Fig. S9-329 S11). The analysis of these transformants showed that the average and maximum DNA 330 acquisition values were highly comparable to those described above for DNA exchanges on chr 2 331 ( Fig. 4A) and that multiple exchanged regions were likewise observed (Fig. 4B). We therefore 332 conclude that prey-derived transforming DNA can equally modify both chromosomes. Moreover, 333 our data suggest that consecutive stretches of exchanged DNA above ~200 kbp either do not 334 occur or occur at levels below the detection limit of this study, and that this size limitation is not 335 caused by the toxin/antitoxin-module-containing integron island on chr 2. 336 337

Conclusion 338
Based on the data presented above, we conclude that T6SS-mediated predation followed by 339 DNA uptake leads to the exchange of large DNA regions that can bring about bacterial evolution. shown for VchInd5 (57) and predicted for SXT (58)), which makes these pandemic strains less 348 -15 -likely to undergo HGT events. One could also argue that pandemic V. cholerae are rarely 349 exposed to competence-inducing chitinous surfaces due to the prevalence of inter-household 350 transmission throughout cholera outbreaks (59). Yet in vivo-induced antigen technology (IVIAT) 351 assays showed strong human immune responses against proteins of the DNA-uptake pilus that 352 fosters natural transformation, kin recognition, and chitin colonization (10,11,19,23), which 353 contradicts this idea. Indeed, the major pilin PilA was most frequently identified by IVIAT 354 together with the outer-membrane secretin PilQ (60), which suggests that the bacteria encounter 355 competence-inducing conditions either before entering the human host or after its colonization. 356 The latter option is not, however, supported by in vivo expression data from human volunteers 357 (61). Notably, our work shows the incredible DNA exchange potential that chitin-induced V.

Materials and Methods 363
Bacterial strains, plasmids, and growth conditions. The bacterial strains and plasmids used in 364 this study are described in SI Appendix, Table S1. Unless otherwise stated, bacteria were grown 365 aerobically in LB medium under shaking conditions or on solid LB agar plates (1.5% agar). 366 Growth on chitinous surfaces was performed as previously described (27,48). Additional details 367 are provided in the SI Appendix. 368 369 Preparation of genomic DNA. Genomic DNA (gDNA) was purified from a 2 ml culture of the 370 respective strain. DNA extraction was performed using 100/G Genomic-tips together with a 371 Genomic DNA buffer set as described in the manufacturer's instructions (Qiagen). After 372 precipitation, the DNA samples were transferred into Tris buffer (10 mM Tris-HCl, pH 8.0). This 373 was preferred over rapid gDNA isolation kits such as the DNeasy Blood & Tissue kit (Qiagen), 374 as the latter isolation kit is strongly biased towards shorter DNA fragments (predominantly 30kb 375 in length compared to up to 150kb for the 100/G columns, as stated by the manufacturer). predator/prey co-cultures. To induce natural competence, the WT or a T6SS-negative derivative 597 (A1552ΔvasK; T6SS -) was co-cultured on chitin with different prey strains (Sa5Y-derived) that 598 carried two antibiotic resistance cassettes: aph in vipA (chr 2) and cat at variable distances from 599 aph on the same chromosome or on chr 1, as indicated on the X-axis. Transformation 600 frequencies (Y-axis) indicate the number of transformants that acquired (A) a single resistance 601 cassette or (B) both resistance cassettes divided by the total number of predator colony forming 602 units (CFUs). (C) Natural transformation is not impaired in the T6SSacceptor strain. Purified 603 genomic DNA (gDNA) was added to competent WT or T6SSstrains. (A-C) Data represent the 604 average of three independent biological experiments (± SD, as depicted by the error bars). For 605 values in which one (#) or two (##) experiments resulted in the absence of transformants, the 606 detection limit was used to calculate the average. <d.l., below detection limit. Statistical 607 significance is indicated (*p < 0.05; **p < 0.01). 608 609