A toxin-antitoxin system associated transcription factor of Caulobacter crescentus can influence cell cycle-regulated gene expression during the SOS response

Toxin-antitoxin (TA) systems are widespread in bacterial chromosomes but their functions remain enigmatic. Although many are transcriptionally upregulated by stress conditions, it is unclear what role they play in cellular responses to stress and to what extent the role of a given TA system homologue varies between different bacterial species. In this work we investigate the role of the DNA damage-inducible TA system HigBA of Caulobacter crescentus in the SOS response and discover that in addition to the toxin HigB affecting cell cycle gene expression through inhibition of the master regulator CtrA, HigBA possesses a transcription factor third component, HigC, which both auto-regulates the TA system and acts independently of it. Through HigC, the system exerts downstream effects on antibiotic (ciprofloxacin) resistance and cell cycle gene expression. HigB and HigC had inverse effects on cell cycle gene regulation, with HigB reducing and HigC increasing the expression of CtrA-dependent promoters. Neither HigBA nor HigC had any effect on formation of persister cells in response to ciprofloxacin. Rather, their role in the SOS response appears to be as transcriptional and post-transcriptional regulators of cell cycle-dependent gene expression, transmitting the status of the SOS response as a regulatory input into the cell cycle control network via CtrA. Importance Almost all bacteria respond to DNA damage by upregulating a set of genes that helps them to repair and recover from the damage, known as the SOS response. The set of genes induced during the SOS response varies between species, but frequently includes toxin-antitoxin systems. However, it is unknown what the consequence of inducing these systems is, and whether they provide any benefit to the cells. We show here that the DNA damage-induced TA system HigBA of the asymmetrically dividing bacterium Caulobacter crescentus affects the cell cycle regulation of this bacterium. HigBA also has a transcription factor encoded immediately downstream of it, named here HigC, which controls expression of the TA system and potentially other genes as well. Therefore, this work identifies a new role for TA systems in the DNA damage response, distinct from non-specific stress tolerance mechanisms which had been proposed previously.


Importance 27
Almost all bacteria respond to DNA damage by upregulating a set of genes that helps them to repair 28 and recover from the damage, known as the SOS response. The set of genes induced during the SOS 29 response varies between species, but frequently includes toxin-antitoxin systems. However, it is 30 unknown what the consequence of inducing these systems is, and whether they provide any benefit 31 to the cells. We show here that the DNA damage-induced TA system HigBA of the asymmetrically 32 dividing bacterium Caulobacter crescentus affects the cell cycle regulation of this bacterium. HigBA 33 also has a transcription factor encoded immediately downstream of it, named here HigC, which 34 controls expression of the TA system and potentially other genes as well. Therefore, this work 35 Introduction the TA system through a DNA binding domain found in the antitoxin and transcriptionally represses 48 it (7). This repressive activity can be modulated by the abundance of the toxin protein by 49 "conditional cooperativity," a mechanism by which low levels of toxin protein promote TA complex 50 binding to the TA system promoter, while high levels of toxin destabilise TA complex binding 51 resulting in derepression of transcription (8). The mechanisms of toxins are variable but fall into two 52 broad groups: inhibition of DNA replication by inhibiting the activity of DNA gyrase, and inhibition of 53 translation at various levels including cleavage or modification of mRNA, tRNA or rRNA, or 54 phosphorylation of aminoacyl-tRNA synthetases or EF-Tu (9). 55 However, the physiological relevance of chromosomally encoded TA systems is still an open 56 and hotly debated question. TA systems encoded on plasmids can certainly function as plasmid 57 maintenance systems by elimination of plasmid free cells through post-segregational killing (PSK), 58 where the antitoxin component of the TA complexes in a plasmid-free cell is degraded and not 59 replaced, resulting in elimination of plasmid free cells from the population because they are 60 poisoned by excess toxin (10). It has also been noted that chromosomal TA systems are associated 61 with cryptic prophages (11), superintegrons (12) and transposons (13) and have been suggested to 62 promote stability and/or propagation via horizontal gene transfer of these mobile elements. Other 63 TA systems, of types III and IV, have been shown to have roles in bacteriophage resistance (14,15). A 64 Table 1. Products were purified by agarose gel electrophoresis. Cloning of the correct region was 137 confirmed by sequencing and plasmid stocks were maintained in E. coli EC100 or TOP10. Plasmids 138 used in this work are listed in Table 2. Replicating plasmids (pMT335 and plac290 derivatives) were 139 transferred into Caulobacter strains by electroporation while suicide plasmids for generation of 140 deletion mutants (pNPTS138 derivatives) were transferred by conjugation from E. coli S17-1 λpir 141 (34). After integration of suicide vectors by recombination through one of the homologous flanking 142 regions, secondary recombination events were induced by counterselection on PYE agar containing 143 3% sucrose and mutants carrying resulting in-frame deletions were screened for by PCR. Double 144 mutants were made by introducing the ΔhigC or ΔhigBAC alleles into the ΔlexA mutant strain. Strain 145 numbers and genotypes are listed in Table 3. 146 The overexpression plasmid for WT HigC was constructed by amplification of full length higC 147 with primers cc3036_nde and cc3036_eco, digested with NdeI and EcoRI and ligated into 148 correspondingly digested pMT335 to form pMT335-higC. The overexpression plasmid for truncated 149 HigC was constructed by amplification of the C-terminal half of higC with primers cc3036_noTM_nde 150 and cc3036_eco. The forward primer in this reaction (cc3036_noTM_nde) contains a NdeI site 151 (CATATG) which overlaps amino acid 133 (GTG-Val) and replaces it with ATG-Met. This was digested 152 with NdeI and EcoRI and ligated into correspondingly digested pMT335 to form pMT335-higC-noTM. 153 To construct the higC knockout plasmid pNPTSΔhigC, the flanking regions of higC were amplified 154 with primer pairs 3036_up_bam and 3036_up_hind (606 bp upstream region including the first 6 155 amino acids of HigC) and 3036_down_bam and 3036_down_eco (556 bp downstream region 156 including the last 6 amino acids and stop codon of HigC). These were digested with BamHI/ HindIII 157 and BamHI/EcoRI respectively and ligated simultaneously into EcoRI/HindIII-digested pNPTS138. To 158 construct the higBAC operon knockout plasmid, the pNPTSΔhigBA plasmid backbone (23) was used. 159 This was digested with EcoRI/BamHI to remove the fragment corresponding to the higA downstream 160 region. The vector (including the upstream flanking region of higB) was purified by agarose gel 161 electrophoresis and ligated to the EcoRI/BamHI-digested PCR product of 3036_down_bam and 162 3036_down_eco (downstream flanking region of higC) to give pNPTSΔhigBAC. 163 β-galactosidase assay 164 β-galactosidase assays were performed on strains carrying low copy plasmid-borne transcriptional 165 fusions of the promoters of interest to lacZ. Cultures were grown to early exponential phase (OD 600 = 166 0.1 -0.4) with exposure to ciprofloxacin or vanillate as described in the main text or figure legends, 167 followed by β-galactosidase assays using the method of Miller (35) on three independent biological 168 replicates. 169

RNA extraction and quantitative RT-PCR 170
RNA was extracted from 4 ml mid-exponential phase cultures which were treated with 2500 U 171 Ready-Lyse (Bionordika) and homogenized using QiaShredder columns, prior to RNA extraction with 172 the RNEasy Mini Kit (Qiagen) according to the manufacturers' instructions, including on-column 173 DNase digestion. RNA quality was assessed by agarose gel electrophoresis and concentration was 174 measured in a Nanodrop spectrophotometer. cDNA was prepared using the SuperScript IV Reverse 175 Transcriptase Kit (Thermo Fisher) according to the manufacturers' instructions, on 400 ng RNA 176 template using random hexamer primers. Quantitative RT-PCR was performed in technical and 177 biological triplicates on a 96-well LightCycler real-time PCR system (Roche) using SYBR Green 178 (Roche). The higB transcript was amplified with primers higB_qrt_fwd and higB_qrt_rev, the higC 179 transcript was amplified with primers CC_3036_qrt_fwd and CC_3036_qrt_rev, and the reference 180 gene rpoD was amplified with primers rpoDfow and rpoDrev. Quantification was by the standard 181 curve method and higB and higC transcript levels were normalized to rpoD. 182

Quantitative PCR-chromatin immunoprecipitation (qChIP) 183
Chromatin immunoprecipitation experiments followed by quantitative PCR were performed using an 184 anti-CtrA polyclonal antibody (26) as described previously (36). Quantitative PCR was performed in 185 technical duplicates on two biological replicates using primers pilA_chip_f and pilA_chip_r to amplify 186 the promoter of pilA and primers sciP_chip_f and sciP_chip_r to amplify the promoter of sciP. 187 Quantification was by the standard curve method and results are expressed as fold enrichment of a 188 given product in the ChIP sample relative to the input DNA. 189

Efficiency of plating assay 190
Resistance to ciprofloxacin and chloramphenicol was assessed by dilution spot plating. Cultures of 191 strains to be tested were grown overnight to stationary phase, then inoculated into new medium to 192 grow to mid-exponential phase. Culture density was measured, normalised to the OD600 of the least 193 dense culture (OD600=0.5 or less), serially diluted in PYE to 10 -6 and 5 µl spotted onto plates 194 containing PYE medium with sub-inhibitory concentrations of ciprofloxacin (1 μg/ml) or 195 chloramphenicol (0.02 μg/ml). For higC overexpression experiments, the plates contained 196 gentamicin in addition to ciprofloxacin or chloramphenicol in order to maintain selection on the 197 plasmid, and all plates contained vanillate (50 μM) to induce expression of higC. Plates were imaged 198 after 3 days growth at 30°C. Images are representative of three independent biological replicates. 199

Persister assay 200
Overnight cultures were diluted into new PYE medium and grown to mid-exponential phase (OD 600 = 201 0.4 -0.6). Ciprofloxacin was added to a final concentration of 10 μg/ml and a 100 μl sample of the 202 culture was immediately taken out for quantification of cfu/ml at zero time. Further 100 μl samples 203 were taken at 2, 4, 6, 24 and 48 hours after ciprofloxacin addition. Immediately after sampling, cells 204 were washed in 1 ml PYE followed by centrifugation at 8000g for 5 minutes, repeated 3 times. 205 Washed cells were serially diluted to 10 -6 and plated in technical duplicates as described in the figure  206 legends, then incubated at 30°C for 3 days. Data are reported as fraction surviving cfu/ml relative to 207 zero time for each time point, for three independent biological replicates. 208

Statistical analysis 209
All numerical data are reported as mean of all biological replicates performed and error bars indicate 210 the standard deviation unless otherwise stated. Statistical significance was analysed by non-paired 211 equal variance 2-tailed Student's T test for comparisons between strains or treatment conditions. * 212 signifies p < 0.05 and ** signifies p < 0.01 throughout. 213 214

Results 215
The putative transcription factor CCNA_03131 (higC) is associated with and regulates the higBA 216

toxin-antitoxin system 217
We previously observed that loss of the HigB toxin in the ΔlexA background improved viability of the 218 cells, likely because production or activation of this toxin is increased in the absence of LexA. 219 Unexpectedly, we did not observe this improvement in the ΔlexA ΔhigBA strain even though this also 220 lacks HigB. Moreover, both ΔlexA ΔhigBA and ΔhigBA strains were sensitized to ciprofloxacin relative 221 to the ΔlexA and wild type parent strains, respectively, while the ΔhigB and ΔlexA ΔhigB mutants 222 were not (23). To investigate the reason for this difference, we first measured activity of a P higBA 223 promoter-reporter construct in wild type, ΔlexA and ΔhigBA single mutants compared to ΔlexA 224 ΔhigBA. Although we observed the anticipated increased promoter activity in the absence of LexA or 225 HigA repressors, the absence of both of these in the ΔlexA ΔhigBA double mutant surprisingly led to 226 lower, rather than higher, promoter activity ( Fig 1A). Investigating the genomic context of the higBA 227 TA system, we noted that a putative transcription factor gene, CCNA_03131, lies 42 bp downstream 228 of higBA (Fig. 1B). Bioinformatic analysis of promoter and terminator locations (using the bprom (37)  229 and ARNold software (38, 39), respectively) failed to find any putative promoter or terminator 230 sequences between the 5' end of higB and the 5' end of CCNA_03131, while the same programs 231 identified a promoter upstream of higBA and a putative rho-independent terminator downstream of 232 CCNA_03131, suggesting that CCNA_03131 is a member of the higBA operon and potentially a third 233 component of this TA system. 234 Since this gene had been annotated as a LytTR-family transcription factor (40) based on 235 sequence homology, and some TA systems are known to have third components that act as 236 transcription factors (11), we investigated whether it could also regulate higBA. Overexpressing 237 CCNA_03131 from the vanillate-inducible promoter reduced P higBA activity in WT, ΔlexA and ΔhigBA 238 strains relative to the empty vector control, with the most significant effect seen in the ΔhigBA 239 background ( Fig 1C). Hence, the product of CCNA_03131 can repress the P higBA promoter, albeit 240 weakly, and seems to have stronger repressive activity when HigA is absent. Due to the likely co-241 regulation of CCNA_03131 with higBA, and its ability to repress transcription from the higBA 242 promoter, we now consider CCNA_03131 as a part of the higBA TA system operon and name it higC. 243 We then measured the steady-state mRNA levels for higB and higC in WT, ΔlexA, ΔhigA, ΔhigBA and 244 ΔlexA ΔhigBA strains to confirm whether their expression levels were consistent with the promoter 245 activity measurements ( Fig 1D). Both mRNAs were detectable, but higC appeared to be expressed at 246 a much lower level than higB and no obvious induction of higC in the ΔlexA, ΔhigA, and ΔhigBA 247 strains relative to WT was seen. However, higC was very strongly expressed in the ΔlexA ΔhigBA 248 mutant, in which the in-frame deletion of higBA has placed the higC coding sequence immediately 249 downstream of the higBA promoter, and the LexA and HigA repressors are missing. Taken together, 250 these data show that the product of higC functions as a repressor of the higBAC promoter and is 251 strongly overproduced in the ΔlexA ΔhigBA mutant, providing a plausible explanation for why the 252 P higBA promoter-reporter activity in this strain was unexpectedly low. 253 254

The N-terminal helical domain of HigC is required for promoter regulatory activity 255
Based on sequence homology, HigC belongs to the LytTR family of DNA binding proteins 256 (pfam04397, COG3279), but in addition to the DNA binding domain that is typical of this family, it 257 was previously proposed to contain four transmembrane helices (40). Analysis of the HigC protein 258 sequence by the Dense Alignment Surface (DAS) program (41) agreed with this study, suggesting 259 that the four transmembrane helices were in the N-terminal half of the protein sequence, preceding 260 the DNA binding domain that is predicted to start at amino acid 171 (Fig 2A). Since the existence of 261 transmembrane helices seemed counter-intuitive in a transcription factor, which should be able to 262 localize to the nucleoid rather than the membrane, we constructed a truncated version of HigC in 263 which the transmembrane helices were removed (HigC-noTM). Placing this construct under the 264 control of the vanillate-inducible promoter allowed us to compare its effect on the higBA promoter 265 to wild type HigC or the empty vector control ( Fig 2B). In both WT and ΔhigBA strains, the wild type 266 HigC repressed the promoter as before, but the truncated HigC lacking the N-terminal helical domain 267 was completely inactive as a repressor, showing that this domain is required for activity. 268 We further analysed the HigC protein sequence for the presence or absence of a signal 269 peptide, reasoning that if this N-terminal helical domain is genuinely a four-helix transmembrane 270 domain, it should be preceded by a signal peptide to direct it to the membrane for co-translational 271 insertion. However, using the SignalP software (42), no signal peptide for either the Sec or Tat 272 secretion pathways was seen ( Fig 2C). It is unlikely that this is a false negative, because the same 273 program could detect the signal peptide of the Caulobacter outer membrane protein ChvT with high 274 probability ( Fig 2D). Therefore, it is possible that this domain was annotated as transmembrane 275 helices simply because it shares the same helical secondary structure and hydrophobicity of genuine 276 transmembrane helices, but is not actually targeted to the membrane. Since the domain was 277 required for promoter repression activity, and bacterial DNA binding proteins frequently function as 278 dimers or other multimers (43)

HigC affects ciprofloxacin resistance independently of the toxin HigB 283
We then investigated whether HigC overproduction had other phenotypic effects than P higBA 284 repression, by overexpressing it from the vanillate-inducible promoter in WT, ΔlexA and ΔlexA ΔhigB 285 strains and testing its effect on viability in the presence of antibiotics (Fig 3A). At a sub-inhibitory (for 286 WT) concentration of ciprofloxacin, the viability of the ΔlexA mutant was reduced relative to the WT 287 and ΔlexA ΔhigB strains (all containing empty vector), but the viability of the three strains was 288 unchanged on the control plate (containing gentamicin to maintain selection of the pMT335 vector) 289 and on a sub-inhibitory concentration of chloramphenicol. However, on mild overexpression of HigC, 290 the viability of the ΔlexA strain in the presence of ciprofloxacin was reduced even further, and 291 strikingly the improved resistance of the ΔlexA ΔhigB strain to ciprofloxacin was completely 292 reversed. This effect was unique to ciprofloxacin, as it was not seen in the control condition or on 293 chloramphenicol. Viability of a ΔlexA ΔhigC strain was slightly improved relative to the ΔlexA parent 294 strain on ciprofloxacin (Fig 3B), showing that the negative effect of HigC overexpression was not 295 likely due to non-specific intolerance of producing this protein at higher levels than the cell normally 296 experiences. Therefore, HigC negatively influences survival in the presence of DNA damaging 297 antibiotics, especially in the context of constitutively activated SOS response of the ΔlexA mutant. 298 Moreover, since this effect was observed in a ΔlexA ΔhigB mutant, this effect cannot be ascribed to 299 HigC altering P higBA promoter activity and HigB acting as the effector of the response. Rather, HigC 300 must be a direct effector of the ciprofloxacin sensitivity, potentially through regulatory activity on 301 other promoters than P higBA . 302 303 HigBAC has no effect on formation of persister cells 304 Since TA systems had been previously implicated in persister cell formation, we next investigated 305 whether the effect of HigB or HigC on viability in the presence of ciprofloxacin was associated with 306 any change in frequency of persister cell formation. Exposure of WT, ΔhigA, ΔhigBA, ΔhigBAC and 307 ΔhigC cells to a bactericidal concentration of ciprofloxacin followed by dilution spot plating showed 308 that all strains exhibited a biphasic killing curve typical of persister cell formation with the initial 309 rapid killing phase from 0 to 6 hours and with persister cells detectable after 24 and 48 hours (Fig  310   4A), similar to recent work in which persistence to streptomycin and vancomycin was quantified 311 (44). This timecourse experiment showed that these strains displayed very similar biphasic curve 312 profiles to each other with no difference in the rate of the rapid killing phase or the fraction of 313 persisters recovered at 24 or 48 hours. However, we were unable to consistently recover persisters 314 at 48 hours from the ΔhigBA cultures using the spot dilution plate method, so we repeated the 315 experiment measuring only 48-hour persisters but from a larger number of cells. This showed that all 316 strains reproducibly had a fraction of 10 -4 to 10 -5 surviving persister cells after 48 hours ciprofloxacin, 317 and that there was no significant difference in fraction of surviving persisters between any of these 318 strains (Fig 4B). We also performed the timecourse experiment for the ΔlexA, ΔlexA ΔhigBA and 319 ΔlexA ΔhigB strains relative to WT and found that these strains had similar biphasic curve kinetics 320 and a similar fraction of surviving persister cells at 24 and 48 hours, and again no significant 321 difference between any of the strains was seen (Supplementary Fig S1). Therefore, while (ciprofloxacin) treatment, this process is not influenced by the toxin HigB, the transcription factor 324 HigC, or the LexA repressor which controls their expression, and the viability differences observed in 325 our efficiency of plating assays are unrelated to persistence. We compared the activity of CtrA-dependent promoters that are expressed in the G1 phase 341 (swarmer cells) and subject to repression by the co-repressors MucR1/2, with CtrA-dependent 342 promoters that are expressed in the late S-phase and in G2 (stalked and pre-divisional cells) and 343 subject to repression by the regulatory protein SciP, which includes the promoter of CtrA itself. In 344 the CtrA-DN-overexpressing but otherwise WT cells, we anticipated that SciP protein levels should 345 be high and that the CtrA-SciP-dependent promoters should be inactive or weakly active compared 346 to the mixed population/empty vector control. Meanwhile, the CtrA-MucR-dependent promoters 347 should be active in both conditions (possibly increased upon CtrA-DN overexpression). Then, any 348 further differences in promoter activity in the ΔhigA or ΔhigBA backgrounds relative to WT should be 349 accounted for by increased or decreased activity of the toxin HigB against ctrA mRNA. Consistent 350 with our previous result that the loss of HigA had a reduced swarmer cell fraction in a mixed 351 population but no difference in other cell types (23), as observed by FACS, we found reduced activity 352 of the MucR-dependent G1-phase promoter P pilA in ΔhigA relative to WT both with and without CtrA-353 DN overexpression (Fig 5A). We did not observe the same effect for P sciP , suggesting that promoters 354 controlling structural genes are better proxies for this effect than promoters controlling regulatory 355 factors. 356 Surprisingly, the CtrA/SciP-dependent S/G2-phase promoters were not significantly 357 repressed in the WT background when the CtrA-DN allele was overexpressed. However, loss of 358 HigBA appeared to promote this repression, since the ΔhigBA strain had significantly lower activity 359 of both promoters during CtrA-DN overexpression compared to empty vector. Surprisingly, upon 360 CtrA-DN overexpression in the ΔhigA strain, we observed much stronger repression of P ctrA than in 361 ΔhigBA or WT (Fig 5B). The bipartite ctrA promoter is subject to complex multi-level regulation by 362 SciP, CtrA itself, the S-phase associated transcription factor GcrA and the methylation state of the 363 promoter DNA (Fig 5C), so the contribution of the multiple regulatory inputs cannot be inferred from 364 the promoter activity measurement alone. However, we can nonetheless conclude that the ΔhigA 365 genetic background influences cell cycle dependent gene expression, in a manner which is consistent 366 with its cognate toxin HigB negatively regulating ctrA at the post-transcriptional level. In support of 367 this function for HigB, we also found by anti-CtrA ChIP followed by quantitative PCR that the 368 CtrA/MucR-dependent promoters P pilA and P sciP had much less CtrA bound to them in non-369 synchronized populations of the ΔhigA strain compared to WT and ΔhigBA (Fig 6), despite the 370 modest effects observed at the level of promoter activity (Fig 5A). HigB is therefore capable of 371 negatively influencing CtrA binding to and activating its target promoters, regardless of the cell cycle 372 phase they are associated with. 373

HigC influences cell cycle gene expression during the SOS response independently of HigB 375
Since we had observed that HigC could negatively affect survival in the presence of DNA damaging 376 antibiotics in a HigB-independent manner, we then investigated whether this was associated with 377 cell cycle gene expression by using the P pilA -lacZ construct as a reporter for CtrA-dependent 378 promoter activity in the presence and absence of ciprofloxacin, in strains lacking higBA, higC or lexA 379 separately or together (Fig 7A). There was no difference in P pilA activity between WT and ΔhigBA in 380 the control condition, but its activity was increased in ΔhigBA cells treated with ciprofloxacin. 381 However, this effect was not due to increased HigB activity in ciprofloxacin-treated WT, because the 382 activity in a ΔhigBAC mutant strain treated with ciprofloxacin was reduced down to WT levels again. 383 Hence, HigC must have been responsible for the elevated P pilA activity in the ΔhigBA mutant upon 384 induction of the DNA damage response with ciprofloxacin. In the ΔlexA strain, which has the DNA 385 damage response constitutively activated, we observed similar results. Here, the baseline activity of 386 P pilA was lower, probably because of the LexA-induced cell division block (46, 47) that would prevent 387 normal progression through the cell cycle and the associated pulse of pilA transcription in G1 phase. 388 We did not observe any ciprofloxacin-induced increase in P pilA activity in a ΔlexA ΔhigBA mutant 389 compared to the ΔlexA strain. However, the ΔlexA ΔhigBAC quadruple mutant had decreased activity 390 of this promoter compared to ΔlexA ΔhigBA, both with and without ciprofloxacin. Therefore, in 391 conditions where the SOS response is induced but the HigBA TA system inactive, HigC can promote 392 expression of this CtrA-dependent promoter. This activity must be functionally independent of the 393 HigBA TA system, in the sense that it is not mediated by HigB toxin activity against CtrA via HigC 394 regulation of the higBAC promoter. Overexpression of HigC in WT cells from the vanillate-inducible 395 promoter, under the same conditions in which we saw HigC repression of P higBA , did not result in any 396 alteration of P pilA activity relative to the empty vector ( Supplementary Fig S2), suggesting that the 397 effect of HigC on this promoter is either indirect, or undetectable if HigBA is present. 398 399 Discussion 400 In the present study we report that the HigBA toxin-antitoxin system of Caulobacter possesses a 401 transcription factor third component HigC, which participates in auto-regulation of the higBA 402 promoter but which also acts independently of HigBA (Fig 7B). We confirm our findings from 403 previous work that the HigB toxin can target the cell cycle regulator CtrA at the post-transcriptional 404 level, resulting in decreased CtrA-promoter binding and lower CtrA-dependent promoter activity. 405 Moreover, we find that under conditions of SOS response induction, HigC decreases cell viability and 406 can also influence expression of CtrA target genes. The decrease in viability was independent of the 407 higB toxin gene, while the expression of the CtrA-dependent pilA promoter was increased by HigC 408 specifically in the absence of HigBA. Deletion of neither higBA nor higC had any effect on formation 409 of persister cells in the presence of ciprofloxacin, suggesting that the HigBA-and HigC-dependent 410 phenotypes that we observe are unrelated to the persistence phenomenon and instead indicate that 411 HigBAC is acting as a regulatory coupling factor linking regulation of cell cycle genes to the SOS 412 response. 413 While the close proximity of higC to higBA initially suggested that these genes may be in the 414 same operon and therefore co-regulated, some aspects of whether higC is regulated identically to 415 higBA still remain unknown. Mindful of the recent observation that in the E. coli mqsRA TA system, 416 the antitoxin mqsA is transcribed from promoters internal to the mqsR coding sequence (18), we 417 searched for promoters not only in the higA-higC intergenic region but in the entire higBA coding 418 sequence. This bioinformatic analysis did not uncover any cryptic internal promoters. However, 419 Caulobacter -10 and -35 promoter sequences do not closely match the canonical -10 and -35 boxes 420 characterised in E. coli and other Gram negative bacteria, so it is also possible that this is a false 421 negative. Indeed, in a previous global analysis of genome-wide transcription start sites over the cell 422 cycle (48), it was found that there was a low-frequency transcription start site which corresponded 423 to the A of the start codon of higC, in addition to the high-frequency transcription start site 4 bp 424 upstream of the start codon of higB. Therefore, it is possible to infer that higBA and higC are 425 transcribed from different promoters, with the higC promoter being much weaker, which would 426 explain our result that the higC transcript is apparently present at much lower levels than higB in 427 WT, ΔhigA or ΔlexA strains (Fig 1D). This is more difficult to reconcile with the similar fold changes 428 (between ΔlexA mutant and WT) for all three genes observed by qRT-PCR by da Rocha et al (30), 429 since the region upstream of the putative higC transcription start site has no LexA binding site. 430 However, if transcriptional readthrough occurred during the high levels of transcription from the 431 higB promoter that would be expected during the SOS response, this could account for the lexA-432 dependent higC induction. Interestingly, the higC transcription start site was suggested to be cell 433 cycle regulated while the higB transcription start site was not, with RNA-Seq reads corresponding to 434 the higC site peaking at 80 to 100 minutes after synchronisation (48). This correlates closely with the 435 peak time of the transcription start site of the ctrA P2 promoter, but higC is unlikely to be a 436 candidate for direct regulation by CtrA since there is no CtrA binding motif (TTAA-N 7 -TTAA) in the 437 higA -higC intergenic region and it has not been identified as a CtrA target in any genome-wide 438 analysis (26,49). 439 The role of HigC in repressing the HigBAC promoter is consistent with that observed for 440 other three-component type II TA systems encoding a transcription factor (11, 31), but we also 441 observe some unique differences. In those studies, the transcription factor was primarily responsible 442 for repression of the system, either alone or together with the cognate TA complex acting as co-443 repressor. Meanwhile, for HigBAC, the HigC repression appears much less important than the 444 repression provided by LexA and HigA. Based on our β-galactosidase data, HigC seems to exert the 445 strongest repressive effect when HigA is absent, suggesting that it might act as a negative feedback 446 mechanism to bring higBAC transcription back under control during the late SOS response, if LexA 447 and/or HigA have been absent from the promoter. It will be intriguing to investigate how the 4-helix 448 N-terminal domain of HigC is involved in promoter regulation, since removal of this domain 449 completely abolished its activity as repressor of the higBA promoter. While we find that it is unlikely 450 to be a true membrane protein, on account of the lack of signal peptide, one possibility is that it 451 could mediate protein-protein interactions between HigC monomers or between HigC and other 452 proteins. Pull-down assays of WT and truncated HigC could identify binding partners of this protein 453 and differentiate between ones that depend on the presence of the 4-helix domain and ones that do 454 not. Moreover, since these helical domains were identified in proteins of this family from other 455 alpha-proteobacteria, not only Caulobacter (40), this domain may represent a novel conserved 456 mediator of DNA binding protein interaction in this class of bacteria. It will also be important to 457 define the regulon, either direct or indirect, of HigC in order to fully characterise the role of HigC in 458 the SOS response based on which other genes it regulates in addition to higBA. 459 Our genetic approach, in which we have characterised the effect of HigB in the absence of 460 the antitoxin, and HigC in the absence of the HigBA TA system, has allowed us to gain valuable 461 insight into the functions of these two proteins. However, it is also important not to infer too much 462 from studies of mutant strains about the physiological roles of these proteins in wild type cells. A 463 criticism which is often levelled at studies of TA systems is that phenotypes of antitoxin mutant 464 strains are not equivalent to phenotypes of wild type cells experiencing high levels of toxin 465 production and therefore not physiologically relevant (2,19,21), and therefore a phenotype 466 associated with a given TA system should only be postulated if a phenotype can be observed for a 467 toxin or whole TA system mutant. We do indeed observe such a phenotype for HigBA, since the loss 468 of the toxin in the ΔlexA background substantially improved its resistance to ciprofloxacin. However, 469 in this work we have also found that the difference in this ciprofloxacin resistance phenotype 470 between our ΔhigB and ΔhigBA strains was due to the polar effect of the higBA deletion on higC 471 (specifically, placing it immediately downstream of the strong higBA promoter leading to much 472 stronger higC expression than would normally occur). This underscores the importance of taking 473 genetic context into account and not assuming that in-frame deletions are free of polar effects. 474 Nonetheless, we can still conclude that the HigB toxin is likely to be active to some extent during the 475 SOS response, based on the ciprofloxacin resistance phenotype of the ΔlexA ΔhigB strain, and that 476 when active it should inhibit CtrA at the post-transcriptional level resulting in lower expression levels 477 of CtrA-activated genes. In addition, the effect of HigC overexpression or deletion on ciprofloxacin 478 resistance of the ΔlexA strain shows that it can exert its effect when expressed at relatively low 479 levels and when higBA is still present. Taken together, our data show that the TA system HigBAC of 480 Caulobacter crescentus is a uniquely acting three-