A semi-lethal CRISPR-Cas system permits DNA acquisition in Enterococcus faecalis

Antibiotic resistant bacteria are critical public health concerns. Among the prime causative factors for the spread of antibiotic resistance is horizontal gene transfer (HGT). A useful model organism for investigating the relationship between HGT and antibiotic resistance is the opportunistic pathogen Enterococcus faecalis, since the species possesses highly conjugative plasmids that readily disseminate antibiotic resistance genes and virulence factors in nature. Unlike many commensal E. faecalis strains, the genomes of multidrug-resistant (MDR) E. faecalis clinical isolates are enriched for mobile genetic elements (MGEs) and lack CRISPR-Cas genome defense systems. CRISPR-Cas systems cleave foreign DNA in a programmable, sequence-specific manner and are disadvantageous for MGE-derived genome expansion. A unique facet of CRISPR biology in E. faecalis is that MGEs that are targeted by native CRISPR-Cas systems can be transiently maintained. Here, we investigate the basis for this CRISPR tolerance. We observe that E. faecalis can maintain self-targeting constructs that direct Cas9 to cleave the chromosome, but at a fitness cost. Interestingly, no canonical DNA damage response was observed during self-targeting, but integrated prophages were strongly induced. We determined that low cas9 expression is the genetic basis for this transient non-lethality and use this knowledge to develop a robust genome editing scheme. Our discovery of a semi-lethal CRISPR-Cas system suggests that E. faecalis has maximized the potential for DNA acquisition by attenuating its CRISPR machinery, thereby facilitating acquisition of potentially beneficial MGEs that may otherwise be restricted by genome defense.

transcriptional response distinct from the response to levofloxacin (LVX), a clinically relevant 105 fluoroquinolone antibiotic. Robust induction of the SOS response genes with LVX treatment, and 106 the concomitant lack of induction of these genes by CRISPR targeting, led us to conclude that 107 CRISPR self-targeting does not induce an SOS response in E. faecalis. However, CRISPR self-108 targeting induced all seven integrated prophages in V583. Finally, we demonstrate that increased 109 expression of cas9 leads to CRISPR lethality and contributes to bacteriophage resistance. We 110 utilize this knowledge to develop a robust CRISPR genome editing platform for E. faecalis. These 111 findings, coupled with our previous results, reveal a mechanism used by a bacterial pathogen to 112 overcome the limitations of possessing a genome defense system while preserving population-113 level protection against foreign DNA. We previously reported the ability of E. faecalis to transiently maintain CRISPR targets (19). It 119 has also been postulated that CRISPR targets can be temporarily maintained through plasmid 120 replication that proceeds faster than CRISPR targeting (27). To account for this possibility, the 121 experiments in this study utilize vectors that direct Cas9 to target the chromosome; this ensures 122 that CRISPR-Cas complexes would not need to compete with plasmid replication. To generate a 123 vector for facile generation of chromosome-targeting constructs, we modified a previously 124 developed plasmid bearing a synthetic CRISPR that targeted ermB (19). We removed the first 125 repeat upstream of the ermB spacer and introduced the promoter for pPD1 bacA (PbacA), which is 126 strongly constitutive (28). Subsequently, we introduced pheS* to allow for counterselection on 127 para-chloro-phenylalanine (p-Cl-Phe) (29). The resulting plasmid was designated pGR-ermB 128 (GenBank Accession: MF948287), which has advantages over its parent plasmid. In addition to 129 counterselection, removal of the first repeat reduces the probability of spacer deletion while also 130 allowing the spacer to be easily altered through PCR-directed mutagenesis (19). We 131 subsequently modified the spacer to target different regions of the chromosome of E. faecalis 132 V649 (V583 + cas9) (19). We assumed that the number of instances a protospacer target was 133 present in the genome was proportional to the number of DSBs that would be caused via CRISPR 134 self-targeting. We constructed four derivatives of pGR-ermB that were predicted to generate one 135 DSB (targeting vanB, a gene for vancomycin resistance) or up to ten DSBs (targeting the IS256 136 transposase). A control predicted to generate no DSBs (pGR-tetM, targets tetracycline resistance 137 gene tetM, which is not present in V583) was also constructed. Consistent with our previous 138 observations of CRISPR escape (19,26), a large number of transconjugants arose despite 139 chromosomal CRISPR targeting, and no change in conjugation frequency (CF) was observed 140 between pGR-vanB (1 DSB) and pGR-IS256 (10 DSBs) ( Figure 1a). This suggested that total 141 CRISPR lethality could not be achieved even with constructs that theoretically cleaved the 142 genome in 10 distinct locations, in contrast to previous investigations of CRISPR self-targeting in 143 other species (25,30). This result was also observed in M236, an engineered derivative of Merz96 144 that encodes cas9 (Figure S1a), and OG1RF, which natively encodes the entire CRISPR1-Cas 145 system (described later), demonstrating that this phenotype is not strain-specific. 146 147 Transconjugants of V649 pGR-IS256 were subsequently examined for phenotypic characteristics 148 of this apparent non-lethal CRISPR self-targeting. Transconjugants that maintained CRISPR self-149 targeting constructs displayed slower colony growth relative to control constructs on media with 150 vancomycin (for selection of V649) and chloramphenicol (for selection of pGR-IS256) ( Figure S2). 151 Furthermore, V649 pGR-IS256 transconjugants possessed an extended lag phase in 152 chloramphenicol broth relative to controls and were two-fold more sensitive to LVX and 153 ciprofloxacin (Figure 1b-c, Table S1). A growth defect was also observed in M236 pGR-vanB 154 transconjugants ( Figure S1b-c). These findings demonstrate that CRISPR self-targeting 155 constructs confer deleterious but not lethal fitness effects on E. faecalis. We previously 156 demonstrated that these phenotypes are associated with the transient maintenance of CRISPR 157 conflicts without mutation of the CRISPR machinery in E. faecalis (19,26). 158 159

Transcriptional responses to CRISPR-and fluoroquinolone-induced damage 160
It is possible that CRISPR-Cas self-targeting in E. faecalis induces a robust SOS response, which 161 has been previously observed in E. coli (31). To assess this hypothesis, we performed RNA 162 sequencing to examine changes in gene expression due to CRISPR and LVX-induced damage. 163 To assess CRISPR damage, V649 pGR-tetM (control) and V649 pGR-IS256 (test) 164 transconjugants from vancomycin/chloramphenicol selection were pooled and RNA harvested. 165 To assess LVX-induced damage, RNA was harvested from cultures prior to and two hours after 166 LVX administration at the minimum inhibitory concentration. 167 168 After statistical filtering, 999 genes in V649 were significantly differentially expressed by either 169 LVX or CRISPR self-targeting (Dataset S1). 227 genes were significantly up-regulated during 170 CRISPR self-targeting and 626 were significantly up-regulated by LVX, with 162 genes up-171 regulated in both conditions. Therefore, 71.4% of genes up-regulated during CRISPR self-172 targeting were also up-regulated by LVX, but only 25.9% of genes up-regulated by LVX were also 173 up-regulated by CRISPR ( Figure S3). Prophage genes were up-regulated by both CRISPR and 174 LVX ( Figure 2). 70% of the significantly up-regulated genes by CRISPR self-targeting alone were 175 located in prophage elements. Increases in circular Phage01 DNA and infectious phage particles 176 were detected in LVX and CRISPR treatments ( Figure S4). This correlates well with observations 177 of prophage induction upon ciprofloxacin exposure (32). Importantly, induction of the SOS 178 response, including recA, dinP, and EF1080 (predicted umuC), was observed with LVX, but not 179 by CRISPR self-targeting (Dataset S1). Furthermore, various regions of the genome were 180 regulated discordantly between our two experimental conditions. LVX treatment up-regulated 181 genes on two integrated plasmids, but CRISPR did not. Interestingly, a cluster of genes in the 182 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; vancomycin resistance transposon were up-regulated by CRISPR but not differentially regulated 183 by LVX ( Figure 2). Collectively, these data demonstrate that E. faecalis responds to CRISPR self-184 targeting in a manner distinct from a fluoroquinolone-induced SOS response. Taken together with 185 our previous findings, we directly demonstrate a unique, semi-lethal phenotype associated with 186 CRISPR targeting in E. faecalis, characterized by prophage induction but no canonical DNA 187 damage response. We hereafter refer to this transient maintenance of CRISPR targets and the 188 corresponding phenotypes as "CRISPR tolerance". 189 190

Genetic basis for CRISPR tolerance 191
We hypothesized that increasing the abundance of certain components of the CRISPR machinery 192 would potentiate CRISPR chromosome targeting and lead to lethality. We introduced PbacA 193 upstream of cas9 and examined CFs of CRISPR-targeted plasmids. 27-fold up-regulation of cas9 194 was verified with RT-qPCR ( Figure S5). We previously showed that pKHS67, targeted by spacer 195 67 on the V649 CRISPR2 locus, possesses markedly reduced CFs relative to pKH12, which lacks 196 a protospacer target (19). When cas9 expression is increased (strain V117; V583 PbacA-cas9), a 197 significantly greater reduction in CF is observed, and pKHS67 transconjugants fall near or below 198 levels of detection ( Figure 3A). Similarly, we observe very few V117 transconjugants that arise 199 from chromosomal targeting with pGR-vanB ( Figure 3B). We then hypothesized that the few V117 200 transconjugants that accepted CRISPR targets were mutants with inactivated CRISPR-Cas. To 201 investigate this, we assessed plasmid maintenance in the absence of selection. Our previous data 202 showed that CRISPR-dependent plasmid loss in the absence of selection is one of the 203 phenotypes of CRISPR tolerance (19,26). Expectedly, V649 pGR-IS256 transconjugants 204 demonstrate marked plasmid loss after two days of passaging without selection, characteristic of 205 the CRISPR tolerance phenotype and consistent with pGR-IS256 conferring a fitness defect to 206 host cells ( Figure 3C). However, V117 pGR-IS256 transconjugants on average show no 207 significant plasmid loss, indicating that these are true CRISPR mutants ( Figure 3C). We verified 208 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; that these observations extend to E. faecalis strains natively encoding cas9 by investigating 209 OG1RF, which natively possesses the functionally linked CRISPR1-Cas and CRISPR2 loci. 210 Consistent with results obtained in V649 and M236, we observed a 2-log reduction in CF with 211 pKHS5, which is targeted by S5 on the OG1RF CRISPR2 locus, relative to the control. We then 212 inserted PbacA-cas9 into OG1RF, creating strain OG117. We observed significant 5-log reductions 213 in CFs for pKHS5 relative to pKH12 in OG117. CF of a chromosome-targeting construct, pCE-214 pstSCAB (described later) was similarly reduced in OG117 ( Figure 3D). These results collectively 215 demonstrate that increased cas9 expression overcomes CRISPR tolerance and results in 216 CRISPR lethality, and implicate low cas9 expression as the genetic basis for CRISPR tolerance. 217

CRISPR genome editing in E. faecalis 227
Knowing that cas9 overexpression leads to lethality of CRISPR self-targeting, we sought to 228 develop an efficient CRISPR editing scheme for E. faecalis, since none had been reported. We 229 modified pGR-vanB to encode a homologous recombination template which conferred a 100 bp 230 deletion of vanB ( Figure S6a). Successful edits would abolish vancomycin resistance, and 231 therefore allowed us to utilize a rapid screen. The new plasmid, designated pCE-vanB, was 232 conjugated into V649 (V583 + cas9) and V117 (V583 + PbacA-cas9); transconjugants were 233 selected on erythromycin (for V649 or V117 selection) and chloramphenicol (for pCE-vanB 234 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; selection). After two days, V117 transconjugant colonies appeared at low frequencies. 235 Interestingly, two colony morphologies were observed for V649 transconjugants; some were large 236 and appeared after two days, but most were slower-growing and apparent after three days. We 237 distinguished these phenotypes as "early" (the larger colonies) and "late" (the smaller colonies). 238 Transconjugants from all three groups (V117, V649 early, and V649 late) were restruck on 239 chloramphenicol agar and then screened for vancomycin sensitivity. Remarkably, 83% of V117 240 transconjugants were vancomycin-sensitive. 50% of the early V649 transconjugants and 22% of 241 the V649 late transconjugants were vancomycin-sensitive ( Table 1). The restreak on 242 chloramphenicol was essential for CRISPR editing of vanB, as V117 pCE-vanB transconjugant 243 colonies on the initial erythromycin/chloramphenicol selection still possessed some cells that were 244 vancomycin-resistant ( Figure S6b). Vancomycin-sensitive clones were passaged and plated on 245 counterselective media to identify clones that lost pCE-vanB, and these were screened for the 246 desired edit by PCR ( Figure S6C). All vancomycin-sensitive clones that were PCR-screened 247 contained a 100 bp deletion of vanB. Editing in V649 reveals that homologous recombination can 248 rescue these cells from the effects of CRISPR tolerance, albeit at markedly lower efficiencies than 249 when cas9 is overexpressed (Table 1). 250

251
To further evaluate CRISPR editing efficiency, we designed a CRISPR editing construct to delete 252 genes encoding the putative phosphate transporter pstB2 or the entire operon consisting of pstS2, 253 pstA, pstC, pstB2, and pstB (hereafter referred to as pstSCAB) (Figure 5a-b). 67% and 56% of 254 V117 clones screened by PCR had deletions in pstB and pstSCAB, respectively. Furthermore, 255 pstSCAB deletion by CRISPR editing in was highly efficient in OG117 (OG1RF + PbacA-cas9) (95% 256 editing success), demonstrating that CRISPR editing can be achieved in different E. faecalis 257 strains ( Figure 5C). 258 259 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; During these experiments, the conjugation frequency of chromosomal CRISPR targeting 260 constructs into V117 (V583 + PbacA-cas9) was low (only ~100 CFU/mL transconjugants were 261 obtained in some experiments). We sought a method to increase CF and avoid plating extremely 262 high cell densities to detect modified clones. The New England Biolabs REBASE (34) predicted 263 a type IV restriction endonuclease in V583 (EF3217), for which a homolog was biochemically 264 assessed in S. aureus (35). The predicted recognition site (SCNGS) from S. aureus corresponded 265 to known 5-methylcytosine methylation sites in the E. faecalis OG1 derivatives OG1RF and 266 OG1SSp (G m5 CWGC) (36). Since the donor used for conjugation in our experiments is also 267 derived from OG1, we hypothesized that deletion of EF3217 in the recipient would increase CF 268 of CRISPR editing constructs. We therefore generated strain V200, a V117 derivative which lacks 269 EF3217, using CRISPR editing. CFs of all plasmids, even those targeting the chromosome, were 270 significantly greater for V200 recipients compared to V117 ( Figure S7). We also successfully 271 performed CRISPR editing in V200 (V583 + PbacA-cas9 ΔEF3217), demonstrating that successive 272 CRISPR edits are possible in our system ( Figure 5C, Table 1). 273 274 "Side effects" of lethal chromosome targeting 275 Since the genomes of E. faecalis clinical isolates typically possess multiple repetitive elements, 276 we sought to assess if CRISPR editing could drive large genome deletions or rearrangements. 277 We used pGR-ermB, which targets ermB on pTEF1; pTEF1 is a 66 kb pheromone-responsive 278 plasmid conferring erythromycin and gentamicin resistance that naturally occurs in V583 and its 279 derivatives. Since ermB is flanked by two IS1216 elements, we hypothesized that CRISPR 280 targeting of ermB in the absence of an exogenous recombination template could result in 281 erythromycin-sensitive mutants that had undergone recombination between the repetitive IS1216 282 sequences. Indeed, multiple erythromycin-sensitive clones were recovered when ermB was 283 targeted in strain V200. Whole genome sequencing was performed on two of these mutants. One 284 clone (V202) deleted the entire region between the IS1216 transposases, including ermB. 285 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; Remarkably, the other clone (V204) had lost ~75% of pTEF1 (~45 kb deletion). V204 was also 286 sensitive to gentamicin via deletion of aac6'-aph2'. The mechanism for this large deletion was 287 recombination between IS1216 and IS256 sequences on pTEF1 and pTEF3, which resulted in 288 deletions in both plasmids ( Figure S8). Our findings demonstrate that CRISPR chromosome 289 targeting can promote larger and unintended recombination events where repetitive DNA is 290 abundant, in agreement with previous data (37). 291 292 Finally, we investigated potential off-target mutations that arose as a result of CRISPR genome 293 editing, including whether unintended mutations occurred as a consequence of cas9 294 overexpression. In addition to sequencing the genomes of V202 and V204 as described above, 295 we sequenced V117 pCE-vanB and V200 (see Figure S8 for a diagram of strain derivations). 296 These strains collectively represent three independent CRISPR editing events. V200 (V583 + 297 PbacA-cas9 ΔEF3217) and V204 (V583 + PbacA-cas9 ΔEF3217, erm S , gent S ) were identical (except 298 for the aforementioned recombination events), while V117 (V583 + PbacA-cas9) and V202 (V583 299 + PbacA-cas9 ΔEF3217, erm S ) differed from V200 by two and one single nucleotide 300 polymorphisms, respectively ( Figure S8). The low frequency of genetic variations between the 301 four clones confirms the highly specific nature of CRISPR genome editing in our system. Taken 302 together, we validate CRISPR editing as a highly efficacious platform for genetic manipulation in 303

Discussion 306
In this study, we investigated the intrinsic non-lethality of chromosomal targeting by the native E. 307 faecalis CRISPR1-cas9. We show that maintenance of chromosomal targeting constructs results 308 in the induction of prophages, but no canonical SOS response. Furthermore, when cas9 is 309 overexpressed, a highly significant reduction in the number of transconjugants that accept 310 CRISPR targeting constructs is observed. These transconjugants appear to be phenotypic 311 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; CRISPR mutants. Using this knowledge, we subsequently developed a rapid and robust CRISPR 312 genome editing platform in E. faecalis. 313

314
Although we were able to map the transcriptomic response to chromosomal CRISPR targeting, 315 the exact events that occur inside a cell upon CRISPR self-targeting are unclear. We are uncertain 316 if low expression of cas9 alone accounts for the ability to survive chromosomal CRISPR targeting. 317 It is tempting to speculate that trans acting elements (anti-CRISPR proteins, regulatory RNAs, 318 etc.) may regulate the expression or activity of cas9. Furthermore, overexpression of cas9 does 319 not lead to complete phenotypic lethality in all cases, exemplified by the fact that some individual 320 transconjugants harboring chromosomal CRISPR targets show modest plasmid loss in the 321 absence of selection ( Figure 3c). Additionally, during CRISPR editing to remove vancomycin 322 resistance, the initial transconjugant colony still possesses vancomycin-resistant cells ( Figure  323 S6b). Nevertheless, we show that low expression of cas9 is largely responsible for the ability of 324 E. faecalis to transiently tolerate CRISPR targets, which may be advantageous to allow some E. 325 faecalis cells to accept foreign DNA. This phenotype may also protect strains that accidently 326 acquire a self-targeting spacer. During preparation of this manuscript, a study by Jones et al. 327 demonstrated that kinetics of a catalytically inactive Cas9 are particularly slow at low 328 concentrations (38). The investigators suggest that in order for Cas9 to quickly find its target, both 329 Cas9 and the crRNA would need to be present at high concentrations. It is therefore possible that 330 the CRISPR tolerance we observe here and in our previous work is actually the direct phenotype 331 of slow Cas9 kinetics in nature. Nevertheless, the advantage of CRISPR tolerance in the context 332 of beneficial MGEs is clear. When CRISPR targets that may be beneficial are encountered by a 333 population, it is advantageous for a large fraction of that population to be CRISPR tolerant and 334 "sample" the effect of possessing the MGE. If the MGE is helpful for survival, the cell can still 335 proliferate; if it is not, the MGE can be removed or MGE-containing cells outcompeted. 336 337 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; The ability to maximize DNA acquisition appears to come at the cost of compromised 338 bacteriophage defense. This phenomenon has yet to be observed in other bacteria, and 339 underscores the puzzlingly low expression of cas9 in E. faecalis. While it is possible that E. 340 faecalis has lost CRISPR function in the context of phage defense altogether, we hypothesize 341 that it is more likely that CRISPR-Cas, specifically cas9, is somehow induced under certain 342 conditions. The extent to which CRISPR tolerance occurs for E. faecalis in the gastrointestinal 343 tract will be the subject of future investigations. 344 345

Materials and Methods 346
Detailed materials and methods can be found in SI Materials and Methods.       Table 1.
Successful edits and appropriate negative controls are shown as indicated. All clones were verified to be chloramphenicol sensitive, indicative of plasmid loss. Bacterial strains, growth conditions, and routine molecular biology procedures 520 Enterococcus faecalis was routinely cultured at 37°C in Brain Heart Infusion (BHI) without 521 agitation; Escherichia coli was routinely cultured at 37°C in Lysogeny Broth with agitation at 220 522 rpm. Routine PCR was performed with Taq DNA polymerase, and PCR for cloning purposes was 523 performed with Q5 DNA polymerase (New England Biolabs). T4 Polynucleotide Kinase (New 524 England Biolabs) was used for routine phosphorylation. PCR products were purified with the 525 PureLink PCR Purification Kit (Invitrogen). Plasmids were purified using the GeneJet Plasmid 526 Purification Kit (Fisher). Primers were synthesized by Sigma-Aldrich. Routine DNA sequencing 527 was performed at the Massachusetts General Hospital DNA Core facility. E. coli EC1000 was 528 used for routine plasmid propagation (39). E. faecalis and E. coli competent cells were prepared 529 as described previously (19). Genomic DNA was extracted using the MO BIO Microbial DNA 530 Isolation Kit (Qiagen). Antibiotics were used in the following concentrations: chloramphenicol, 15 531 μg/ml; streptomycin, 500 μg/ml; spectinomycin, 500 μg/ml; vancomycin (van), 10 μg/ml; 532 erythromycin (erm), 50 μg/ml; rifampicin, 50 μg/ml; fusidic acid, 25 μg/ml; tetracycline, 10 μg/ml; 533 gentamicin (gent), 300 μg/ml. A full list of primers can be found in Table S2. 534 535

Strain and plasmid construction 536
A schematic of the plasmid construction used in this study is shown in Figure S9. All strains and 537 plasmids used in this study are shown in Table S3. CRISPR edited strains are shown in Table 1. 538 All CRISPR editing plasmids can be derived in a single step from pGR-ermB (accession number: 539 MF948287). The derivation of pGR-ermB is described below. 540 541 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; To generate chromosomal targeting constructs, pCR2-ermB was linearized to remove 160 bp 542 upstream of the ermB spacer and simultaneously introduce the promoter of bacA from pPD1, 543 which is constitutive (PbacA) (19,28). This procedure also removed the upstream repeat. The linear 544 product was phosphorylated and self-ligated to generate an intermediate plasmid referred to as 545 pSR-ermB. This plasmid was once again linearized around cat and a fragment containing cat and 546 pheS* from pLT06 was blunt-end ligated (40). The original cat was deleted to simplify the cloning 547 procedure. The final plasmid was designated pGR-ermB, and was fully sequenced (accession 548 number: MF948287). 549

550
To modify the spacer, pGR-ermB was linearized at PbacA and the downstream repeat; primers 551 contained the entirety of the spacer sequence to be inserted. The exception was pGR-IS256, 552 which was generated without ligation by taking advantage of the ability of E. coli EC1000 to 553 recombine linear DNA (i.e., linear DNA was recombined in vivo). All pGR derivatives were 554 sequence-verified to ensure spacer integrity prior to introduction into C173 for conjugation. 555 Homologous recombination templates were introduced using the NEB HiFi DNA Assembly Master 556 Mix (New England Biolabs). For simplicity, the spacer was included as overhangs during Gibson 557 assembly, and therefore a plasmid containing two fragments for homologous recombination and 558 the appropriate spacer could be generated in a single step. The same linearization-559 phosphorylation-ligation procedure was used to modify the plasmid to insert PbacA upstream of 560 cas9. Knock-in protocols were performed essentially as previously described (41). A streamlined 561 protocol for CRISPR-Cas9 genome editing in E. faecalis using our system is outlined in Figure  562 S10 and the primer schematic for generating CRISPR editing plasmids is shown in Figure S9. 563 564 For CRISPR editing, the appropriate plasmid was first transformed into E. faecalis C173 or 565 CK111SSp(pCF10-101). Conjugation was then performed into the desired recipient strain, and 566 transconjugants were selected on agar media containing chloramphenicol and appropriate 567 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; antibiotics for recipient strain selection. Transconjugant colonies were re-struck for isolation on 568 agar media containing chloramphenicol, and single colonies were inoculated into 1-5 mL of BHI 569 broth lacking antibiotics and incubated at 37°C until turbid. Cultures were then struck on MM9YEG 570 + para-chloro-phenylalanine (p-Cl-Phe) to counterselect for the plasmid backbone. By this point, 571 the recipient strain will have received the CRISPR editing plasmid, recombined with the editing 572 template, and then lost the backbone plasmid. In total, this procedure can take as little as two 573 days once transconjugants are obtained. We observed that an additional passage in MM9YEG + 574 p-Cl-Phe was helpful for eliminating residual chloramphenicol resistance, since the 575 counterselection is imperfect. This extra passage was utilized whenever frequencies needed to 576 be determined and there was no marker to phenotypically screen for, since preliminary 577 experiments occasionally yielded some chloramphenicol-resistant clones which interfered with an 578 accurate assessment of successful editing rates. Once presumptive CRISPR-edited mutants 579 were obtained, colony PCR to confirm the desired edit was performed in all cases except for 580 deletion of pstB; the larger amplicon required that genomic DNA be extracted. 581 582

Conjugation assays 583
Conjugation assays were performed essentially as described (19)

Transcriptomics Analysis 590
To assess the transcriptional response to CRISPR self-targeting, transconjugants of V649 pGR-591 tetM (control) and V649 pGR-IS256 (test) selected on vancomycin and chloramphenicol were 592 incubated on agar media for 2 days. Cells were scraped from plates, resuspended in RNA-Bee 593 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; (Tel-Test), and lysed by bead-beating in lysis matrix B (MP Biomedicals). After RNA-Bee 594 extraction, the aqueous layer was subject to ethanol precipitation. The RNA was treated with 595 DNase (Roche) and concentrated using the GeneJet RNA Cleanup and Concentration Kit 596 (Fisher). For assessment of the transcriptional response to levofloxacin (LVX)-induced stress, 597 cells were treated essentially as previously described (19). Briefly, overnight cultures of V649 598 were diluted in fresh medium and grown to OD600nm = 0.3, at which point cultures were split. Some 599 cells were harvested for control transcriptomic analysis, and LVX was added to remaining cells at 600 a concentration of 1 μg/ml. After two hours of incubation with LVX, the remaining cells were 601 harvested. RNA was isolated and treated with DNase as described above. Three biological 602 replicates were performed with both experimental conditions. were first removed and the unmapped reads were mapped to the V649 reference genome. 616 Transcripts per million (TPM) values were used to quantitate expression. False discovery rate 617 (FDR)-adjusted P value was used to assess significance. Genes were filtered first by removing 618 those for which both CRISPR self-targeting and LVX treatment yielded FDR-adjusted P-values 619 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; >0.05. Subsequently, genes for which both LVX and CRISPR self-targeting had fold changes <2 620 were removed. The remaining list consisted of genes that were significantly up or downregulated 621 by either LVX or CRISPR self-targeting. 622 623 RT-qPCR to verify increased cas9 expression was performed as previously described (19) Approximately 10 5 -10 6 PFU/mL of ΦNPV-1 was added to 5 mL of M17 + chloramphenicol soft 628 agar and overlaid on BHI + chloramphenicol agar (33). Overnight cultures of OG1RF and OG117 629 containing pGR-tetM or pGR-NPV1 were spotted on the soft agar containing ΦNPV1. pGR-NPV1 630 targets a predicted phage lysin gene. A simultaneous control lacking soft agar and phage was 631 included to enumerate total bacterial CFU. Using identical amounts of ΦNPV-1 in each 632 experiment was essential for consistent results. Cultures were induced with LVX as described in a previous section. Induced cultures were 644 pelleted, and the supernatant was filtered using 0.2 μm polyethersulfone filters. Similarly, 645 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; transconjugant colonies of V649 pGR-tetM and V649 pGR-IS256 were scraped from agar plates 646 using 2 mL PBS (identical to protocol used for transcriptomics analysis), pelleted, and the 647 supernatant filtered. Filtrates were spotted on soft agar containing lawns of E. faecalis ATCC 648 29212, which is susceptible to infection by V583 prophages (42). To prepare the lawns, overnight 649 cultures of ATCC 29212 were diluted in fresh medium and cultured to OD600nm=0.4. 10 μL culture 650 was added to 2 mL melted soft agar (BHI broth, 0.2% agarose, 10 mM MgSO4) and the mixture 651 was poured on a 100 mm diameter standard BHI agar plate (1.5% agar). We observed that 652 varying the amount of bacteria added and the thickness of the soft agar affected visibility of phage 653 plaques; the protocol we present here yielded the clearest zones of lysis. 654 655

Genome sequencing 656
Whole genome sequencing was performed at MR DNA (Molecular Research LP). Briefly, libraries 657 were prepared using the Nextera DNA Sample preparation kit (Illumina) using 50 ng of total 658 genomic DNA. Libraries were pooled and sequenced paired-end for 300 cycles using the Illumina 659 HiSeq system. Reads were mapped to the V117 genome in CLC Genomics Workbench. Mapping 660 graphs were generated to identify deleted (zero coverage) regions, and basic variant detection 661 was performed on read mappings to identify smaller SNPs, deletions, and insertions using the 662 default parameters. 663 664 Dataset S1. Changes in gene expression resulting from CRISPR self-targeting and LVX. 665 The fold changes of gene expression for LVX (FC-LVX) and CRISPR (FC-CRISPR) are indicated 666 for all genes that were differentially regulated as described in the transcriptomics analysis section. 667 Also included are sheets which categorize genes up-and down-regulated by CRISPR or LVX. 668 For these sheets, if the fold change of gene expression was >2 but the P-Value was >0.05, a fold 669 change of 1 was manually entered; the true fold change value can be found in the "master" sheet. 670 671 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; Table S1. Fluoroquinolone minimum inhibitory concentrations. Single transconjugant colonies were suspended in 5 mL BHI and used as inocula in broth microdilution antibiotic susceptibility assays. Units of concentrations are μg/ml. V583 pGR-tetM V583 pGR-IS256 V649 pGR-tetM V649 pGR-IS256 Levofloxacin 1 1 1 0.5 Ciprofloxacin 1 1 1 0.5 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017;  . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; Figure S3. Relationship among changes in gene expression between LVX treatment and CRISPR self-targeting. A) All genes (except those with fold changes of infinity) that were significantly (see SI materials and methods) differentially regulated either by LVX or CRISPR were plotted, irrespective of individual P-value. The horizontal axis represents the fold change of gene expression caused by LVX, and the vertical axis represents the corresponding fold change of gene expression caused by CRISPR self-targeting. Green regions indicate genes that were similarly differentially regulated by CRISPR and LVX. Red regions indicate genes that were oppositely differentially regulated by CRISPR and LVX. Yellow regions indicate genes that were differentially regulated by either CRISPR or LVX, but not both. B) is the same as A) except lacking genes located on prophage elements. Representative CRISPR editing in vancomycin-sensitive clones after passaging and counterselection. Edited products are 100 bp smaller than unedited products. CRISPR2 was amplified as a control to verify that edited clones are not donor strains, which possess a longer CRISPR2 array than V117 and V649 (17).
. CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; Figure S7. Deletion of EF3217 increases conjugation frequency. pGR-tetM (control), pGR-vanB (targets chromosome), and pCE-pstSCAB (used for CRISPR editing) were conjugated into V117 (V583 + PbacA-cas9) or V200 (V583 + PbacA-cas9 ΔEF3217) and conjugation frequencies are shown as transconjugants per donor (n=3). The limit of detection was 100 CFU/ml. **P<0.01, ***P<0.001 . CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; Figure S8. CRISPR targeting does not induce unintended single nucleotide polymorphisms (SNPs) but drives large scale recombination events. A) Strain construction is shown. Red arrows indicate CRISPR editing, with the corresponding edits located adjacent to the arrows. The complete genomes of the underlined strains were sequenced. *pCE-vanB was not removed in V117 for this experiment. B) Relevant mutations are shown for the four sequenced strains relative to V117 pCE-vanB, since V117 pCE-vanB possessed the fewest mutations. V200 (V583 + PbacA-. CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; cas9 ΔEF3217) and V204 (V583 + PbacA-cas9 ΔEF3217, erm S , gent S ) differ from V117 pCE-vanB by two SNPs, and V202 (V583 + PbacA-cas9 ΔEF3217, erm S ) differs from V117 pCE-vanB by three SNPs. Mutations that were supposed to occur because of CRISPR editing and large scale recombination events in the pTEF plasmids are not represented in this table, but were confirmed by whole genome sequencing. C) Regions of deletion in pTEF1 and pTEF3 of V204 (V583 + PbacA-cas9 ΔEF3217, erm S , gent S ) are shown as a red line. Relevant genes are indicated as shown, and transposases that were found flanking the deleted region are underlined. Each graph represents the number of reads (from 0-2000) as a function of the nucleotide position of each plasmid. The three lines at each position represent the minimum, mean, and maximum number of reads for each 1000 nt or 100 nt grouping for pTEF1 and pTEF3, respectively. This grouping was automatically performed by CLC Genomics Workbench to display the data effectively when representing the entirety of the plasmid. Reads that mapped within the deleted regions were only those that mapped to multiple locations in the genome. V202 (V583 + PbacA-cas9 ΔEF3217, erm S ), which is not shown in this figure, contains a deletion of only ermB mediated by recombination between the adjacent IS1216 transposases.
. CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; Figure S9. Plasmid construction scheme. The general plasmid workflow is shown (components not to scale). CRISPR repeats are depicted by thin, light-blue rectangles; the colored rectangles adjacent to the repeats represent various spacers. All CRISPR editing plasmids can be derived from pGR-ermB as either one-step or two-step assemblies. Generic primer schematic for generating CRISPR editing deletion plasmids from a single step is shown as arrows indicating 5'-3' directionality. The primer pairs used in each reaction are colored identically (i.e., the two red arrows represent the primers that are used in the same reaction to amplify one fragment).
Homologous overhangs for subsequent Gibson assembly are shown. 30 bp overhangs were used in all cloning procedures.
. CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017; Figure S10. CRISPR-Cas genome editing protocol for E. faecalis. A workflow for achieving CRISPR-assisted genome editing in E. faecalis is shown. Media are color coded. BHI, BHI + chloramphenicol, and MM9YEG + p-Cl-Phe are shown in red, brown, and green, respectively. The bacteria present at each step of the process are also indicated. The appropriate number of colonies to screen (n) is dependent on each experiment, but we find that screening 6 transconjugants is sufficient.
. CC-BY-NC-ND 4.0 International license peer-reviewed) is the author/funder. It is made available under a The copyright holder for this preprint (which was not . http://dx.doi.org/10.1101/232322 doi: bioRxiv preprint first posted online Dec. 11, 2017;