Circulation of a Quorum-Sensing-Impaired Variant of Vibrio cholerae Strain C6706 Masks Important Phenotypes

Phenotypic diversity between laboratory-domesticated bacterial strains is a common problem and often results in the failed reproduction of published data. However, researchers rarely compare such strains to elucidate the underlying mutation(s). In this study, we tested one of the best-studied V. cholerae isolates, O1 El Tor strain C6706 (a patient isolate from Peru), with respect to two main phenotypes: natural competence for transformation and type VI secretion. We recently demonstrated that the two phenotypes are coregulated and specifically induced upon the growth of pandemic V. cholerae O1 El Tor strains on chitinous surfaces. We provide evidence that of seven C6706 strains collected from different laboratories, four were impaired in the tested phenotypes due to a mutation in a QS gene. Collectively, our data indicate that the circulation of such a mutated wild-type strain of C6706 might have had important consequences for QS-related data.

horizontally acquired integrative and conjugative element (ICE), which rendered such ICE-carrying strains similarly nontransformable (30). However, despite the fact that the well-studied V. cholerae C6706 strain (31), an O1 El Tor patient isolate from Peru, does not contain such an ICE and that this strain has been described as QS proficient and naturally transformable (13), several researchers have reported to us its nontransformability. Here, we followed up on this nontransformability by testing seven C6706 isolates obtained from different laboratories located in North America and Europe. We show that approximately half of these wild-type (WT) strains contain the same compromising mutation within luxO, resulting in impaired QS behavior and, consequently, in low natural transformability and T6SS activity.

RESULTS
The majority of C6706 strains are severely impaired in their natural transformability. In 2005, it was shown for the first time that the human pathogen V. cholerae could enter a state of natural competence and that this phenotype depends on the presence of chitin (13). That study and follow-up studies showed that many patient isolates of V. cholerae, as well as environmental samples, are naturally transformable in a chitin-dependent manner (8,13,32,33). However, frequent concerns exist in the field with respect to the transformability of O1 El Tor pandemic strain C6706 (personal communications from several researchers to M.B.). We therefore asked seven principal investigators working on diverse aspects of V. cholerae to share their C6706 strains with us. First, we tested these seven samples in a well-established chitin-dependent transformation assay (26) and compared the transformation frequencies to those of our main laboratory strain, the QS-and competence-proficient V. cholerae O1 El Tor A1552 strain. As presented in Fig. 2, our data confirmed that four of these seven wild-type C6706 strains were transformable only sparsely compared with the remaining three C6706 samples and the A1552 control strain. This bipartite response indicates that the nontransformability of such C6706 strains was not caused by an improper following of published protocols but rather by genetic differences between circulating C6706 strains. The majority of C6706 strains produce lowered hapR transcript levels and changed levels of expression of HapR-regulated genes. To elucidate whether the impairment of natural transformation was caused by a lack of chitin induction or by a problem in the QS circuit, we tested the seven C6706 strains for the expression of several QS-related and QS-unrelated genes, first in the absence of chitin (at HCD). In particular, we first monitored the transcript levels of hapR by quantitative reverse transcription-PCR (qRT-PCR) because the gene encodes an important coregulator of natural transformation and T6SS (Fig. 1). Interestingly, the hapR transcript levels were reduced in the same C6706 isolates (isolates 2 to 5) that also showed low transformability (Fig. 3A). These low hapR transcript levels were mirrored in lowered expression of hapA (for which HapR acts as an activator [34]) ( Fig. 1) and in higher expression of the Vibrio polysaccharide synthesis gene vpsA (for which HapR acts as a repressor [35,36]) ( Fig. 1) than in the three highly transformable C6706 samples (sample 1, sample 6, and sample 7) and the A1552 control strain (Fig. 3A). The expression levels of competence-related genes did not differ between the samples under such chitinindependent HCD conditions ( Fig. 3A), consistent with the fact that the competence regulon is not induced in LB medium in pandemic O1 El Tor strains.
When the same strains underwent a chitin-induced expression analysis, a similar QS-dependent pattern became obvious. As highlighted in Fig. 3B, the same differences in the hapR transcript levels were observed upon growth on chitinous surfaces, as described for LB-grown bacteria (Fig. 3A). The low levels of the HapR regulator were again reflected in the changed levels of expression of hapA and vpsA (Fig. 3B). Notably, while expression of the pilA QS-independent competence gene ( Fig. 1) was induced in all of the tested strains upon growth on chitin (Fig. 3), the expression of the chitin-(TfoX-) and QS-coregulated competence genes was severely reduced in the same subset of QS-impaired C6706 strains, which were almost nontransformable ( Fig. 2 and 3B). The affected genes were qstR, which itself requires induction by TfoX and HapR (22) (Fig. 1), and all of the tested QstR-dependent genes that encode either parts of the DNA-uptake machinery (e.g., comEA) or components of the T6SS (e.g., vasD, clpV, and vasK) (Fig. 3B). Notably, this striking difference in competence gene expression was observed even though we have previously demonstrated that chitin-attached bacteria show heterogeneity with respect to competence expression (18). We therefore conclude that the low HapR level produced in 4 of 7 of the C6706 isolates is not sufficient to properly induce the competence regulon, which includes the T6SS.

QS-impaired wild-type C6706 strains show reduced TfoX-induced interbacterial killing.
Because we found decreased HapR levels in approximately half of the seven C6706 strains compared with the rest of the C6706 strains and the A1552 control, we considered whether this decrease was also reflected in a reduced ability to kill other Gram-negative bacteria, such as Escherichia coli, by means of T6SS-mediated attack. To answer this question, we introduced a transposon carrying an arabinose-inducible copy of tfoX (TntfoX-strep [18,37]) into the chromosome of a representative set of C6706 strains, which allowed us to induce TfoX through the provision of the inducer (e.g., in a chitin-independent manner, as the latter condition does not support the growth of E. coli). We mixed these V. cholerae strains with an arabinose-non-degrading E. coli strain (37) and scored the recoverability of the E. coli prey after 4 h of coincubation. As shown in Fig. 4, QS-impaired C6706 strains 2 to 5 reduced the E. coli numbers only slightly, whereas the QS-proficient C6706 sample (sample 6) significantly reduced the prey population. This phenotype therefore reflects the expression analysis.

QS-impaired wild-type C6706 strains possess a mutation in the QS gene luxO.
To elucidate the cause of the QS impairment and the low hapR transcript (and HapR protein) levels in a subset of the C6706 samples but not in the second half or in other O1 El Tor isolates, we sequenced two important QS genes: hapR itself and the gene encoding LuxO, the upstream-acting regulator of the hapR transcript (19). The rationale behind performing the latter was that at a low cell density (LCD), LuxO is in its phosphorylated and therefore active form. Phosphorylated LuxO (LuxO~P) indirectly (via the regulation of small RNAs) leads to the degradation of hapR mRNA (19) (Fig. 1). Importantly, and as shown in Table 1, all four of the QS-impaired C6706 strains contained the same G-to-A mutation in luxO, which resulted in an amino acid change annotated as glycine 319 initially [29]). None of the other three QS-proficient C6706 samples harbored this mutation, nor did any of the other O1 El Tor, O1 classical, or O37 strains (seven, two, and three isolates, respectively) which were tested at the same time (Table 1).

DISCUSSION
Here, we provide evidence for the circulation of a QS-impaired wild-type version of V. cholerae strain C6706. Four of seven tested C6706 samples obtained from international laboratories were severely affected with respect to their natural transformability and, accordingly, also with respect to their T6SS production and activity. Using qRT-PCR, we showed that the low transformability perfectly correlated with low levels of hapR transcripts and a changed expression pattern of several HapR-regulated genes. Accordingly, HapR-/QstR-coregulated competence genes (e.g., those encoding the DNA-uptake machinery and the T6SS) were not induced upon growth on chitin, explaining the compromised transformation and T6SS responses in these C6706 samples. Sequencing the hapR and luxO genes of the C6706 isolates and other O1 and non-O1 V. cholerae strains showed that the QS-impaired C6706 strains contained a mutation in luxO. Interestingly, the exact same mutation was recently described for a luxU mutant derivative of strain C6706 (38). In their study, Jung et al. provided evidence that the G333S amino acid change mimics the active form of LuxO (38). This change therefore explains why the mutated C6706 strains consistently had lower hapR transcript and protein levels than the nonmutated C6706 strains and non-C6706 control strain A1552. An earlier study reported frequent mutations in the hapR gene of V. cholerae (39), and we speculated that such mutations are overrepresented in culture collections due to a sampling bias (40). Importantly, none of these O1 El Tor and classical isolates of V. cholerae contained the exact same mutation in hapR (39), excluding the clonal expansion of one successful mutant strain. Interestingly, however, Joelsson and col-  character (a, b, or c) are not significantly different. Differences between data labeled with different characters are significant (P <

for comparisons between data labeled with b and data labeled with c).
leagues mentioned in their study that the HapR protein level was reduced in strain C6706 compared with that in several other O1 serogroup strains (39), possibly caused by the here-described mutation of several wild-type C6706 strains. In this context, it should be noted that a recent study on V. fischeri showed that luxO mutations are frequently isolated from cultures in prolonged stationary phase (41). Importantly, the authors describe the isolation of a plethora of different luxO mutant alleles, all of which mimic the gene encoding a constitutively active LuxO protein (41). In our study, however, we found the exact same mutation in four different C6706 strains obtained from different laboratories, and this mutant allele of luxO exactly matches a previously reported mutation in a C6706-derived luxU mutant (38). Thus, it appears rather unlikely that the mutation arose independently in those five different strains. Instead, it can be assumed that the mutated C6706 strain was circulated among different laboratories. It is therefore of prime importance for any group studying QS-related phenotypes, such as the QS network, virulence expression, biofilm formation, natural competence for transformation, and T6SS production, in strain C6706 to ensure that the wild-type laboratory stock(s) (some laboratories seem to have more than one stock of the wild-type C6706 strain) and mutants thereof from other laboratories do not contain the previously reported (38) and here-described (for the WT) luxO mutation. Indeed, as we show in this study, for natural transformation and T6SS production/interbacterial killing, this mutation masks important QS-dependent phenotypes and therefore leads to the irreproducibility of such features. It is unnecessary to mention that mutations can occur at any time. It is important that the presence of such mutations in commonly used strains is communicated within the scientific community in a timely manner to avoid unnecessary investments of time and resources using flawed reagents. Indeed, we have received numerous inquiries with respect to the nontransformability of C6706. The current study solved this mystery, as we revealed a QS-impairing mutation in a circulating C6706 strain. Our recommen- dation for researchers who work with V. cholerae C6706 is therefore to take measures to ensure that the luxO gene is not mutated in their laboratory stock.

MATERIALS AND METHODS
Bacterial strains and growth conditions. The V. cholerae strains used in this study are listed in Table 2. E. coli strain TOP10-TnKan (37) served as prey in the interbacterial killing assay (see below). Bacteria were grown in liquid LB medium under shaking conditions or on LB agar plates (1.5% agar) unless otherwise stated. The temperature was kept at room temperature, 30°C, or 37°C. Half-concentrated defined artificial seawater (0.5ϫ DASW [13]) was used for the chitin-induced natural transformation experiments. Antibiotics and other supplements were added at the following concentrations: kanamycin at 75 g/ml and L-arabinose at 0.2%. For the selection of V. cholerae after triparental mating performed with E. coli donor strains, thiosulfate citrate bile salts sucrose (TCBS) agar plates were used. The plates were prepared following the standard protocol provided by the manufacturer (Sigma-Aldrich/Fluka, Buchs, Switzerland).
Natural transformation assays on chitin surfaces. The natural transformability of the diverse V. cholerae strains was tested through an established transformation assay performed using chitin flakes (26,27). The genomic DNA of strain A1552-lacZ-Kan (26)   independent biological replicates are indicated in the figure. For calculation purposes, the value was set to the detection limit for the experiments that resulted in the absence of transformants (e.g., values below the detection limit of the assay). Significant differences between V. cholerae strain A1552 and the seven isolates of strain C6706 were evaluated with Student's t test on log-transformed data (42).

Gene expression analysis by quantitative reverse transcription-PCR (qRT-PCR).
For the LB growth conditions, the bacteria were grown at 30°C for 6 h under shaking conditions in liquid LB medium to reach a high cell density and processed as previously reported (8,18). RNA that was extracted from chitin-grown bacteria was obtained by growing V. cholerae on chitin flakes (Sigma-Aldrich, Switzerland) (26). After 22 h of static incubation on chitin surfaces (for each strain in quadruplicate), the samples were centrifuged for 3 min, the supernatant was removed, and the pellet was resuspended in 1 ml of Tri Reagent (Sigma-Aldrich, Switzerland). After vortex mixing was performed to ensure homogenization, the samples were again centrifuged to remove residual chitin flakes. The supernatant was transferred to a new tube, shock-frozen on dry ice, and stored at Ϫ80°C.
The expression of representative genes was analyzed by quantitative reverse transcription-PCR (qRT-PCR) as previously described (18). The transcript levels of the indicated genes were normalized to the expression of gyrA to obtain the relative expression values. All of the experiments were performed three independent times, and averages (ϮSD) of results of all of the biological replicates are provided. Statistical analyses were based on two-way analysis of variance (ANOVA), which was performed using GraphPad Prism version 7 for Mac (GraphPad Software, San Diego, CA, USA).
Interbacterial killing assay. The interbacterial killing assay was performed as previously described (8) using E. coli strain TOP10-Kan (37) as the prey and the indicated V. cholerae strains as predators. The bacteria were grown in the absence (Ϫara) or presence (ϩara) of 0.2% arabinose to induce the chromosomal copy on tfoX (harbored on transposon TntfoX-strep [18,37]). Coincubation occurred at 37°C for 4 h with a ratio of predator to prey of 10:1. Recovered E. coli cells were enumerated through serial dilution followed by the counting of CFU per milliliter. Three independent experiments were performed, and averages (ϮSD) of results of these biological replicates are given in the figure. Significant differences were determined using two-way ANOVA (GraphPad Prism).