DNA Damage and Reactive Nitrogen Species are Barriers to Vibrio cholerae Colonization of the Infant Mouse Intestine

Ingested Vibrio cholerae pass through the stomach and colonize the small intestines of its host. Here, we show that V. cholerae requires at least two types of DNA repair systems to efficiently compete for colonization of the infant mouse intestine. These results show that V. cholerae experiences increased DNA damage in the murine gastrointestinal tract. Agreeing with this, we show that passage through the murine gut increases the mutation frequency of V. cholerae compared to liquid culture passage. Our genetic analysis identifies known and novel defense enzymes required for detoxifying reactive nitrogen species (but not reactive oxygen species) that are also required for V. cholerae to efficiently colonize the infant mouse intestine, pointing to reactive nitrogen species as the potential cause of DNA damage. We demonstrate that potential reactive nitrogen species deleterious for V. cholerae are not generated by host inducible nitric oxide synthase (iNOS) activity and instead may be derived from acidified nitrite in the stomach. Agreeing with this hypothesis, we show that strains deficient in DNA repair or reactive nitrogen species defense that are defective in intestinal colonization have decreased growth or increased mutation frequency in acidified nitrite containing media. Moreover, we demonstrate that neutralizing stomach acid rescues the colonization defect of the DNA repair and reactive nitrogen species defense defective mutants suggesting a common defense pathway for these mutants.

The SOS response is a well studied and conserved stress response in bacteria that is elicited following DNA damage and replication fork arrest [for review [11]]. The SOS response is controlled by positive and negative regulators. RecA positively regulates the SOS response by binding to single-stranded DNA fragments generated by attempted replication past DNA lesions. The RecA/ssDNA nucleoprotein filament induces the autocleavage of the negative regulator LexA, a transcriptional repressor. Cleavage of LexA allows for expression of 57 genes in the E. coli SOS regulon including translesion DNA polymerases that are able to replicate DNA past noncoding base lesion, and proteins involved in the inhibition of cell division [for review [11]].
Base excision repair (BER) is the most common form of repair for single base damage [for review [7]]. In BER, a DNA Nglycosylase first excises the damaged base from the deoxyribose moiety in the DNA strand creating an abasic site. Class II apurinic/apyrimidinic (AP) endonuclease then hydrolyzes the phosphodiester bond immediately 59 to the abasic site [for review (Kornberg andBaker 1992, Friedberg 2005)]. Subsequent actions process this site to prime and repair the abasic site ultimately by DNA synthesis with DNA polymerase I and ligation by DNA ligase.
Normal replication can also introduce errors in the form of mismatched DNA base-pairs. These mismatches can lead to permanent mutations after a subsequent round of DNA replication. Mismatch repair (MMR) specifically identifies and corrects these base pairing errors increasing the fidelity of the replication pathway nearly ,1000-fold [for review [7]].
While extensive work has shown the benefit of maintaining genomic integrity for an invading bacterium, there appear to be instances where lapses in genomic fidelity are beneficial for a pathogenic bacterium [12][13][14]. In order to colonize and thrive in a mammalian host, a bacterium must be able to adapt and respond to the conditions and stresses associated its new environment.
Genomic mutations support this by allowing current gene products to gain or alter their functions. The utility of mutation(s) and a pathogen's ability to grow in the human environment has been a source of discussion for several years [13,15]. Giraud et al. showed that a high mutation rate was initially beneficial for Escherichia coli to colonize the mouse gut, but this benefit became a liability once adaptation had been reached [12]. Oliver et al. demonstrated that Pseudomonas aeruginosa from chronically infected individuals often has an increased mutation frequency, suggesting an increased mutation rate can be beneficial to P. aeruginosa to allow rapid adaptation to the hostile host environment [14]. Thus depending on the pathogen, the mode and duration of the infection, defects in DNA repair may be detrimental or beneficial to the infecting bacterium.
Several studies have indicated that host produced reactive oxygen species (ROS) and reactive nitrogen species (RNS) cause DNA damage to the invading bacterium [4,16,17]. Not surprisingly bacteria have several defense mechanisms to detoxify ROS and RNS. Each enzyme detoxifies a specific type of ROS or RNS. For example catalases/peroxidases decompose H 2 O 2 , superoxide dismutases dismutate superoxide and ferrisiderophore reductase removes nitric oxide [18,19]. As with certain DNA repair systems, loss of ROS and RNS defenses have been shown to attenuate bacterial pathogens [16,20].
Studies supporting the importance of ROS/RNS defenses and DNA repair pathways in bacterial pathogenesis often focus on intracellular pathogens [1,2,4]. To survive, intracellular pathogens engulfed by phagocytic cells are either able to escape the phagosome or have mechanisms to survive within it. Within the phagosome, captured bacteria may be exposed to host production of ROS and RNS in a host defense response called the oxidative burst. It is hypothesized that the oxidative burst is responsible for the DNA damage experienced by engulfed bacteria [4,16,17].
Vibrio cholerae is the causative agent of the severe human diarrheal disease cholera. V. cholerae is a non-invasive pathogen that colonizes the small intestine of its host [21,22]. As a noninvasive pathogen, V. cholerae is not expected to experience the same types of stresses as intracellular pathogens, such as an oxidative burst. However V. cholerae does pass through several hostile environments as the disease progresses. Immediately following ingestion, V. cholerae is exposed to the exceptionally antagonistic environment of the stomach where the pH of gastric acid can reach as low as 1 [23,24]. Furthermore, nitrite from both food sources and the salivary nitrite cycle can enter the stomach creating acidified nitrite [25,26,27]. Acidified nitrite has potent antimicrobial effects on gut pathogens [28,29,30,31]. These studies show that the viability of several pathogenic bacteria decreases rapidly under acidified nitrite conditions. Furthermore, nitrates, which can also be found in the stomach, have been shown to modify of gene expression reducing acid tolerance [32]. The antimicrobial effects of acidified nitrite are thought to be due to the generation of deleterious RNS [33]. However, with the exception of a few studies [34,35,36], the points of action of these RNS as well as the bacterial determinants required for protection against them have remained largely unexamined. After traversing the stomach V. cholerae faces several innate host defenses in the intestine including bile, lysozyme, small antimicrobial peptides and complement [37]. Thus, V. cholerae must overcome several barriers during infection that have the potential to cause DNA damage through a direct or indirect mechanism.
We report here that V. cholerae strain C6706 experiences increased DNA damage during passage through the murine gastrointestinal track. We demonstrate that increased genomic stress is a potential barrier to host colonization by V. cholerae. We found that two important DNA repair pathways are necessary for V. cholerae to efficiently colonize the infant mouse intestine. Furthermore, we show that defense against RNS is also necessary for V. cholerae to colonize the infant mouse. In doing so we identify a novel protein required for defense against RNS in pathogenic bacteria. In vitro we show that all our colonization defective DNA repair and RNS defense mutants share a common sensitivity to acidified nitrite and we further show that neutralizing stomach acid rescues intestinal colonization defect of these mutants.

Results
Mismatch repair and base excision repair pathways are required for V. cholerae colonization of the infant mouse intestine To determine if V. cholerae requires defenses against DNA damage during colonization, we tested a series of transposon mutants that contained insertions in different steps in three important DNA repair pathways for their ability to colonize the infant mouse intestine in competition with the wild type strain. These pathways were nucleotide excision repair (NER), base excision repair (BER) and mismatch repair (MMR) ( Table 1). While the SOS response is an important contributor to genomic integrity we did not test a requirement for SOS since Quiones et al. previously showed that SOS activation is not required for intestinal cholera toxin production or colonization [38]. We used uvrA as a representative gene required for NER since uvrA is obligatory for NER. We found no difference in the ability of the uvrA::Tn containing strain to colonize the infant mouse relative to the parental strain suggesting that NER is dispensable for V. cholerae pathogeneis (Table 1).
Apurinic/apyrimidininc (AP) endonucleases are critical in BER. BER has been most well studied in E. coli. E. coli encodes two class II AP endonucleases, Xth [endo II (endo VI)] and Nfo (endo IV). In E. coli Xth is responsible for ,90% of the AP endonuclease activity in the cell [39,40]. Few phenotypes have been attributed solely to Nfo activity but Nfo is known to contribute to BER [41]. V. cholerae carries close homologs of both Xth and Nfo (VC1860 and VC2360 respectively). Interestingly we found that the xth::Tn mutant was not

Author Summary
Studies on intracellular bacterial pathogens have shown the need for maintaining genomic fidelity to promote colonization. Loss of DNA repair functions often leads to attenuation and rapid clearing of the invading pathogen. However, for some pathogens, an increased mutation rate has been shown to be beneficial for promoting host colonization, presumably by allowing the pathogen to rapidly adapt to adverse host conditions. We asked if the non-invasive pathogen V. cholerae experienced increased DNA damage during infection and if so, how the increased damage influenced host colonization and from where the source of the damage was derived. Our results demonstrate that V. cholerae experiences increased DNA damage during infection in the infant mouse model and that loss of ability to repair this damage results in attenuation of virulence. We specifically show that V. cholerae requires both base excision repair and mismatch repair for efficient intestinal colonization. Furthermore, we present evidence that the source of the DNA damage is derived from reactive nitrogen species (RNS) formed by acidified nitrite in the mouse gut and in doing so we identify a new RNS defense protein found in V. cholerae and several other pathogenic bacteria.
defective in intestinal colonization however the nfo::Tn mutant showed a defect in colonization compared to the parental strains (Table 1). We created a clean deletion of nfo (Dnfo) in V. cholerae and found this mutant also had a colonization defect. In E. coli deletion of xth and nfo leads to a more profound defect in DNA repair than either single mutant. Consistent with this observation, we found that an xth::Tn Dnfo double mutant showed a ,10-fold defect in colonization that appears slightly greater than the ,5-fold defect in colonization of the Dnfo mutant alone (Table 1), although this difference is not statistically significant for this number of replicates tested (p.0.05). Thus, these results suggest that BER is important for V. cholerae to colonize the infant mouse intestine when in competition. These results also show a critical function for Nfo in survival, which has not been apparent under laboratory conditions. Loss of mismatch repair function has been shown to be either beneficial or detrimental depending on the pathogen studied [12,13,14,15]. We found that a transposon mutant in mutS, which encodes the gene product that initially binds to a mismatch, resulted in a decrease in colonization efficiency (Table 1). We constructed a clean deletion of mutS (DmutS) to ensure the defect was not due to the transposon. We found that the clean deletion of mutS was also attenuated in its ability to colonize the intestine suggesting that mismatch repair or at least MutS is important for V. cholerae pathogenesis (Table 1). We also found that a second clean deletion of mutS showed a similar competitive index defect (CI = 0.1860.03) suggesting that the colonization defect was not due to mutations in the first mutS clean deletion strain. We noted that the mutS transposon mutant was more defective than its clean deletion counterpart (Table 1). This difference may be due to a polar effect of the transposon or mutations acquired by the mutS::Tn strain during outgrowth of the original isolate. We also tested the colonization proficiency of a Dnfo DmutS double mutant and found that the colonization defect of this double mutant appears slightly greater than either the Dnfo or DmutS mutant alone (Table 1) although this difference is not statistically significant for this number of replicates tested (p.0.05).
The requirement of BER and MMR for V. cholerae to efficiently colonize the infant mouse intestine suggests that V. cholerae experiences DNA damage in the mouse, and that a reduced ability to repair such damage is detrimental for V. cholerae pathogenesis.
V. cholerae base excision repair and mismatch repair mutants show classic DNA repair defects V. cholerae genes encoding Xth, Nfo and MutS were identified based on sequence similarity with their well-studied E. coli homologs. To ensure the V. cholerae homologs possessed their predicted functions we tested our mutant strains for the well characterized phenotypes described in other bacterial systems. Loss of mismatch repair causes an increase in mutation rate often referred to as a mutator phenotype [42]. We found that our DmutS mutant had a significantly increased mutation frequency compared with the wild type control ( Figure 1A). The wild type phenotype could be restored by expression of mutS from a plasmid but not by the plasmid itself ( Figure S1B). This result indicates that MutS in V. cholerae shares the same activity as its other well studied bacterial homologs in the repair of DNA replication errors.
Loss of Xth activity in E. coli renders the strain sensitive to hydrogen peroxide (H 2 O 2 ) [43]. We found that our xth::Tn strain was also sensitive to H 2 O 2 ( Figure 1B). Loss of nfo activity alone does not greatly sensitize E. coli to H 2 O 2 but loss of xth and nfo creates a strain with increased sensitivity to H 2 O 2 [43]. We found a similar effect in V. cholerae where the xth::Tn Dnfo strain was much more sensitive to H 2 O 2 then the xth::Tn mutant alone ( Figure 1B). Furthermore, high level expression of nfo from a plasmid complemented the H 2 O 2 sensitivity of the xth::Tn Dnfo mutant ( Figure S1C). These results suggest that V. cholerae Nfo acts like its E. coli homolog.
Passage through the mouse increases the mutation frequency of V. cholerae The requirement of BER and mismatch repair (MMR) systems for V. cholerae to efficiently colonize the mouse intestine suggests that V. cholerae experiences increased DNA damage while in the mouse. To address this possibility we measured the mutation frequency of V. cholerae following passage though the mouse as compared to passage in liquid culture. We inoculated five mice and five liquid cultures with the same size inoculums of V. cholerae. The following day we purified bacteria from the mouse intestine (see Materials and Methods). We plated both V. cholerae passaged through the mouse and grown in liquid cultures followed by selection for resistance to two antibiotics we used as an indicator for measuring mutation frequency. The first was a gain of function mutation in rpoB conferring resistance to rifampicin; the second was a loss of function of thyA conferring resistance to trimethoprim. Mutations in rpoB and thyA are well characterized markers for increases in mutation frequency [44,45,46]. We found that following passage of V. cholerae through the mouse there was an ,2 fold increase in rifampicin resistance and ,2.5 fold increase in trimethoprim resistance compared to the liquid culture grown  strains (Figure 2A, B). We sequenced 19 trimethoprim resistance isolates that were passed through the mouse and 20 isolates obtained following growth in liquid culture. We identified 39 unique mutations in thyA (data not shown) suggesting that our results were not influenced by a mutation acquired early on in the procedure. We did not observe a bias in the types of mutation from the two conditions. These results suggest that passage through the mouse results in an increase in mutation rate for V. cholerae suggestive of an increase in DNA damage and the need for repair mechanisms.
Defense against RNS is required for V. cholerae to colonize the infant mouse We have identified two DNA repair mechanisms required by V. cholerae to efficiently colonize the infant mouse, and have shown that V. cholerae passaged through a mouse has an increased mutation frequency. Thus, we sought to identify potential causes of DNA damage for V. cholerae while in the mouse to understand the requirement for BER and MMR in the mouse. A major source of DNA damage for intracellular pathogens is from host produced  ROS and RNS. While V. cholerae is a non-invasive pathogen we considered that it still may experience ROS and RNS at some point during infection. We used a genetic approach to determine if ROS/RNS affected V. cholerae colonization and if so what type(s) of ROS/RNS were most important during this encounter.
Bacteria have several enzymes to detoxify ROS/RNS. Each enzyme detoxifies a specific type of ROS or RNS [for review see [18,19]. For example catalases/peroxidases decompose H 2 O 2 , superoxide dismutases remove superoxide and ferrisiderophore reductases remove nitric oxide. Bacteria can contain multiple proteins capable of dealing with one type of stress. V. cholerae possesses two catalases/peroxidases (KatB/PerA) and one alkyl hydroperoxide reductase (AhpC), three superoxide dismutases (SodA/B/C) but only one ferrisiderophore reductase (HmpA). We tested mutants defective for each of these different types of defense enzymes to identify the type(s) of radicals that may be damaging V. cholerae in the mouse (Table 2).
RNS, including nitric oxide, have been shown to be powerful antimicrobial agents. The most well studied RNS defense enzyme in bacteria is Hmp, a ferrisiderophore reductase that destroys nitric oxide [47]. V. cholerae carries an hmp homolog, hmpA. Both an hmpA::Tn mutant and a DhmpA deletion mutant showed a defect in colonizing the infant mouse intestine ( Table 2). Deletion of hmpA delayed V. cholerae growth in the presence of a nitric oxide donor but not in the absence ( Figure 3A) consistent with previous observations in other bacteria [20,48]. This suggests that V.cholerae may encounter deleterious RNS during passage in the mouse. The growth defect of the DhmpA mutant in the presence of a nitric oxide donor could be complemented by ectopic expression of hmp from the arabinose inducible plasmid pBAD18 ( Figure S1D). In fact, expression of hmp from pBAD18 allowed the DhmpA mutant to recover growth more rapidly than the parental strain in the presence of a nitric oxide donor.
After testing all previously predicted antioxidant enzymes we began to mine the V. cholerae genome for additional putative antioxidant enzymes. We began by searching for putative proteins that belonged to large antioxidant families. Enzymes, such as AhpC, belong to the Peroxiredoxin (PRX) family. Searching for peroxiredoxin family proteins yielded a putative defense enzyme we have called PrxA (VC2637). PrxA, classified is a peroxiredoxin-5 family protein, is found in several pathogenic bacteria and is a distant homolog of a macrophage peroxynitrite detoxification protein [49].
Deletion of prxA did not effect V. cholerae growth in LB alone but significantly delayed V. cholerae growth in the presence of a nitric oxide donor ( Figure 3A). Furthermore, both the prxA::Tn mutant and the DprxA allele we constructed caused a decrease in the ability of V. cholerae to colonize the infant mouse in competition assays ( Table 2). The growth defect of the DprxA mutant in the presence of a nitric oxide donor could be complemented by ectopic expression of prxA from the arabinose inducible plasmid pBAD18 ( Figure S1E). The discovery of a new gene required for both defense against RNS and efficient colonization of the infant mouse further supports our findings that V. cholerae may be exposed to RNS during passage though the mouse.
We tested the sensitivity of a DprxA DhmpA double mutant and found that the growth of the double mutant in the presence of a nitric oxide donor was even more delayed than either the DprxA or DhmpA single mutant alone ( Figure 3A). We also tested the colonization efficiency of a DprxA DhmpA ( Table 2) and found that it was not significantly less than the DprxA mutant alone (p.0.05). Thus hmpA and prxA are both important for colonization but the effects were not additive.
We asked if defects in ROS defense also affect V. cholerae colonization. Disruption of ahpC, katB, perA, sodA or sodC did not affect the ability of V. cholerae to colonize the infant mouse and these deficiencies did not affect the ability of V. cholerae to colonize the infant mouse in competition experiments ( Table 2). We did not test the sodB::Tn mutant since both it and a DsodB deletion strain we constructed had a decreased plating efficiency and grew very poorly compared to the parental strain ( Figure S1F). Thus, while SodB appears to be important for growth of V. cholerae under laboratory conditions we did not pursue the sodB mutant in mouse experiments.
Interestingly, of ahpC, katB and perA only disruption of perA sensitized V. cholerae to H 2 O 2 in vitro ( Figure 3B). Furthermore, of strains disrupted individually for sodA, sodB and sodC only disruption of sodB sensitized V. cholerae to the superoxide generating compound plumbagin ( Figure 3C). We also tested the DprxA mutant but found that it did not show increased sensitivity to either H 2 O 2 or plumbagin ( Figure S2A and data not shown). It is possible that some of these known ROS defense enzymes overlap in function masking the effects of a deficiency in any one gene in vitro or in mouse studies. For other bacterial pathogens and symbionts deletion of several or all catalases and superoxide dismutases has been required before a strong effect on virulence or symbiosis was observed [50,51,52]. Currently, the results from our analysis suggest RNS may pose a significant barrier to V. cholerae in colonizing the infant mouse. ROS may also play a role, however their effect is not immediately evident in our analysis.
XthA appears to be more important than Nfo in protecting V. cholerae against environmental stress in vitro ( Figure 1B), yet Nfo appears to be more important for colonization of the small intestine ( Table 1). The requirement for DhmpA and DprxA for efficient intestinal colonization lead us to ask if Nfo was required for defense against nitric oxide. We monitored the growth of the xth::Tn, Dnfo and xth::Tn Dnfo mutants in the presence of a nitric oxide donor. We found that, at least in vitro, xth:Tn was more important than Nfo for protection against nitric oxide ( Figure 3D). We also found that the double mutant was again more sensitive to the stress than either single mutant alone ( Figure 3D).

Colonization defective mutants have increased sensitivity to acidified nitrite
Our results led us to ask if our DNA repair and RNS defense defective V. cholerae mutants were sensitive to any host defenses. The intestine has several innate defenses [37]. We tested many of these defenses including lysozyme, phospholipase, antimicrobial peptides, complement, bile, changes in osmolarity and pH, however, we did not observe any difference in sensitivity between the parental and the mutant strains (data not shown). RNS have been shown to be generated by macrophages to kill phagocytised bacteria [reviewed in [53]]. The RNS from macrophages is generated by an inducible nitric oxide synthase (iNOS). Inhibition of iNOS activity has been shown to rescue the virulence defects in hmp mutant strains of Salmonella enterica serovar typhimurium [20]. However, our hmpA::Tn V. cholerae mutant showed no difference in it ability to colonize the intestine of a wild type or isogenic iNOS 2/2 infant mouse (Table S1).
Thus, our results suggest that the colonization defect of the DNA repair and RNS defense mutants may occur before V. cholerae is exposed to the host defenses found in the small intestine. In the stomach V. cholerae is exposed to low pH in combination with mM amounts of nitrite from ingested food and the salivary nitrite cycle [25,26,27]. Acidified nitrite produces a variety of toxic RNS. We quantified the amount of nitrite in the infant mouse stomach using the Griess reaction and found that it was 20.060.7 mM, which is similar to that of humans [26]. The pH range of human gastric juice is reported as 1-3 [23,24]. We determined the pH of the infant mouse stomach to be 4.560.1 using a fluorescent pH sensitive dye. This measurement is conservative and the pH of the infant mouse gastric juice may be even less (see Materials and Methods). Thus, the infant mouse stomach is sufficiently acidic to promote the formation of acidified nitrite.
At pH 3 in rich medium we found that V. cholerae had a greater than 99.9% decrease in survival in less than 1 minute (data not shown) agreeing with similar work examining V. cholerae acid tolerance [54]. We did not find a difference in survival between the parental and mutant strains at low pH (1-4) levels (data not shown). We gradually increased pH to identify the lowest level at which V. cholerae could grow. At pH 5.5 V. cholerae and the DNA repair and RNS defense mutants grew with identical kinetics ( Figure 4A). We titrated nitrite into the growth medium and found that nearly all the mutant strains showed a growth defect compared to the wild type at pH 5.5 in the presence of 400 mM nitrite ( Figure 4B). No differences in growth between wild type and mutant strains were observed at pH 7.0 with or without 400 mM nitrite ( Figure S3A, B). Not only did low pH and nitrite slow the growth of our mutants but the DhmpA, DprxA, Dnfo and xth::Tn Dnfo mutants began to show a decrease in optical density after longer exposure ( Figure 4B) suggesting the cells were lysing.
Only the growth of the DmutS strain was unaffected at 400 mM. We considered that while MutS may not be required for survival of acidified nitrite during this time course it may be required to prevent acidified nitrite induced mutations in V. cholerae that are detrimental for colonization. We grew V. cholerae in LB at pH 5.5 over night in the presence or absence of 600 mM nitrite and then plated for rifampicin resistant colonies. We found that V. cholerae grown in the presence of nitrite had a greater than 10-fold increase in mutation frequency compared to the media only control ( Figure 4C). Loss of MutS then increased the mutation frequency of V. cholerae in nitrite at pH 5.5 , an additional 5-fold ( Figure 4C).
Thus, MutS may be important to prevent acidified nitrite induced mutations that could impair the ability of V. cholerae to colonize the infant mouse. To further test this possibility we created a DhmpA DmutS double mutant and tested its colonization proficiency (Table 2). Interestingly, the colonization defect of the DhmpA DmutS double mutant was not significantly different than either of the single mutants alone (p.0.05). This result may suggest that HmpA and MutS may share a similar defense pathway in the infant mouse.
Additionally, E. coli MutS can recognize an O6-methyl-dG:dC base pair, a mutation which can occur by alkylation of G bases [55]. Therefore it is possible that MutS may also be important for protection again some type of alkylation that occurs in the mouse stomach.

Low pH and nitrite induce radical formation in V. cholerae
If acidified nitrite produces RNS that damaged V. cholerae DNA, we reasoned that we should be able to detect increased intracellular radical formation in V. cholerae following nitrite treatment. We grew V. cholerae at pH 5.5 with increasing amounts of nitrite and assayed for radical formation using 29,79-dichlorodihydrofluorescein diacetate (H 2 DCFDA). H 2 DCFDA is a cell permeable dye that fluoresces after reacting with RNS and ROS species. H 2 DCFDA reacts with several ROS and RNS including hydrogen peroxide, nitric oxide and peroxynitrite [56,57,58]. We found that H 2 DCFDA fluorescence correlated with increasing nitrite concentration at pH 5.5 indicating an increase in intracellular radicals ( Figure 4D). For comparison, we found that H 2 DCFDA fluorescence did not increase over the concentrations of nitrate examined at pH 7.0. These results further support the requirement for low pH to induce radical formation from nitrite sources ( Figure 4D).

A colonization defect is observed for the DNA repair and RNS defense mutants early after inoculation
Our results suggested that V. cholerae may experience DNA damage as it passes through the stomach. If so, we were curious if a colonization defect could be observed at an early time point after inoculation. We repeated the competitive colonization assays testing the DmutS, Dnfo, DprxA and DhmpA mutants. We found that at 3 hours post inoculation each mutant strain already showed a 50-60% colonization defect when co-inoculated with the wild type (Table S2) though this defect was not as great as the 5-20% defects reported in Table 1. This suggests that a defect of the mutant strains is detrimental for colonization early after inoculation. Angelichio et al. [59] reported that V. cholerae populations in the small intestine do not show a significant increase in number until between 10-24 h post inoculations. These results suggest that the effects of the DNA damage are ,50% detrimental at the earliest stages of infection and become more apparent as the bacteria replicate to high numbers in the intestine. Such a conclusion is consistent with the concept of damage occurring primarily in the stomach.

The colonization defect of the DNA repair and RNS defense mutants is rescued by neutralizing stomach acid
Our results support a relationship between acidified nitrite sensitivity and colonization defects in our RNS and DNA repair. We reasoned that if acidified nitrite in the stomach was responsible for the colonization defects of our mutants then neutralizing stomach acid in the mouse should relieve, at least in part, the observed colonization defects. We used sodium bicarbonate to neutralize the mouse stomach acid. When we inoculated infant mice with our DNA repair and RNS defense defective mutants in the presence of sodium bicarbonate all four mutant strains (Dnfo, DmutS, DprxA and DhmpA) showed significant improvement in their ability to colonize the intestine in competition with the parental strain (Table 3). In fact the colonization defect of the DmutS and Dnfo mutant was completely rescued. The colonization defect of the DprxA was restored to near wild type levels. The colonization defect of the DhmpA mutant was partially rescued although this difference is not statistically significant for the number of replicates tested (p.0.05).
The nitrite concentration of the infant mouse stomach after sodium bicarbonate treatment remained nearly unchanged at 20.360.8 mM.

Discussion
We have shown that V. cholerae must defend against DNA damage to efficiently colonize the infant mouse intestine and that such damage likely occurs early during infection as V. cholerae enters the stomach. We have demonstrated that V. cholerae specifically requires BER and MMR pathways to efficiently colonize the infant mouse intestine. Furthermore, we have identified one previously known and one novel RNS defense protein that facilitates intestinal colonization of the infant mouse. These DNA repair and RNS defense proteins were also required for V. cholerae to grow or maintain genomic fidelity in the presence of acidified nitrite. Furthermore the colonization defects of each mutant could be partially or fully complemented by neutralizing stomach acid suggesting that RNS defense and DNA repair share a common defensive role in the mouse.
V. cholerae has been shown to be very sensitive to low pH [54]. For this reason, human volunteers have their stomach contents neutralized to promote experimental V. cholerae infection as is done with live attenuated vaccine studies [60]. In the recently developed infant rabbit model for cholera, stomach acid is also neutralized and cimetidine is administered to prevent re-acidification in order for V. cholerae to colonize the infant rabbit intestine [61]. However, our DNA repair and RNS defense mutants did not show increased sensitivity to low pH compared to the parent strain. Agreeing with our observations a large screen used to identify genes necessary for colonization and tolerating low pH in V. cholerae did not identify any of the genes we reported here for influencing colonization [62]. While low pH of the stomach alone is undoubtedly detrimental towards V. cholerae our results suggest that neutralizing stomach acid may also be important to prevent RNS formation by acidified nitrite. We propose that the defect in colonization of the DNA repair and RNS defense mutants is due to RNS formation.
Acidified nitrite is present in the stomach where gastric juice interacts from nitrite sources from the diet or salivary nitrite pathway. The chemistry of acidified nitrite is known to produce several potentially deadly radicals (1).
Both NO N and NO 2 N can directly attack cellular macromolecules, but they can also interact with other radicals to form further species such as peroxynitrite and hydroxyl acids. From equation (1) we can see that a key factor in this process is the pH of the solution. In the human stomach, normal nitrite concentrations range from 10-50 mM [24] but at a pH of 1-3 acidification and radical production can happen rapidly. The low pKa of HNO 2 may explain in part why we required much higher concentrations of nitrite to observe detrimental effects on the mutants we studied. pH 5.5 was the lowest pH level we could successfully grow V. cholerae, a value well above the pKa of HNO 2 . Thus at pH 5.5 the acidification of nitrite would occur more slowly and higher concentrations of nitrite would be necessary for mass action to drive the acidification and radicalization of nitrite. Agreeing with this we did not observe any effect of nitrite on the growth of our mutants or V. cholerae at pH 7.0. While acidified nitrite has been shown to effectively kill several bacterial pathogens [28,29,30,31], the mechanism of its action and bacterial defenses to protect against it have remained unknown. We have now shown that MutS, Nfo, HmpA and PrxA are required for protection of acidified nitrite in V. cholerae.
While our results support a role for acidified nitrite in the stomach acting as a major DNA damaging agent, we have not identified the exact location in the gastrointestinal track where the damage occurs. The most apparent location is the stomach where ingested V. cholerae mixes with gastric juice. However it is possible that DNA damage induced by acidified nitrite radicals occurs, or continues to occur, in the upper intestinal tract. As gastric juice exits the stomach it is neutralized by bile salt, etc. and can reach a pH close to 8 [24]. Since the stability of at least some RNS, such as peroxynitrite, increases with increasing pH [63] the upper intestinal tract may provide a more favorable environment for RNS to reach V. cholerae and induce DNA damage.
Bicarbonate has been shown to induce V. cholerae virulence genes in a ToxT dependent fashion [64]. Abuaita and Withey show that significant upregulation of both cholera toxin and tcpA gene expression are observed 3-4 h after addition of bicarbonate [64]. While our inoculation of V. cholerae occurs on a much shorter time scale (,5-15 min after exposure of bicarbonate) and the majority of V. cholerae has passed into the small intestine before 3 h, it is possible that some bicarbonate induced gene regulation may also aid in the bicarbonate rescue of the colonization defect of our DNA repair mutants.
While the debate over the benefits and detriments of increased mutation frequency for pathogenesis continues, we have shown that increased mutation frequency is detrimental to V. cholerae pathogenesis, at least for the short-term colonization of the infant mouse intestine. However, we cannot exclude the possibility that increased mutation frequency affects long-term survival of V. cholerae in the host. After the initial decrease in competitiveness it is possible that increased mutation frequency in V. cholerae could make it more competitive in later stages of colonization or during release into the environment. It would be interesting to test multiple clinical isolates for a mutator phenotype to address this question.
The Xth and Nfo homologs of V. cholerae have strong sequence similarity to their E. coli counterparts. We have shown that V. cholerae and E. coli deletion mutants of xth and nfo also share a similar pattern of sensitivity to hydrogen peroxide. Nfo and Xth have been most extensively studied in E. coli. In E. coli Xth is responsible for greater than ,90% of all AP endonuclease activity in the cell [39,40]. In E. coli, an xth mutant is very sensitive to a variety of DNA damaging agents whereas nfo mutants generally show milder effects [41]. Interestingly, we have shown that in V. cholerae Nfo is more important for colonization of the infant mouse than Xth. Our nfo xth double mutant suggests that Xth may play a role in colonization when Nfo is absent. However, for whatever damage is occurring, Nfo appears to play a more important role in the mouse. Table 3. Effect of NaHCO 3 on mutant V. cholerae ability to colonize the infant mouse intestine in competition with the parental strain (wild type). It is possible that Nfo and Xth are also used differentially for repair of specific types of lesions. Preferences for specific types of damaged bases between Nfo and Xth from E. coli have been previously reported [65]. If RNS are responsible for DNA damage in the mouse we suggest that Nfo may have enhanced ability to aid in the repair of nitrosylative base damage. Additionally, there may be differential expression of xth and nfo or their preferential glycosylase partners in the host.
In our efforts to identify ROS and RNS defense enzymes required for intestinal colonization, we identified a new protein we have called PrxA that was required for RNS defense. Until now Hmp has been the only bacterial protein identified to detoxify RNS, specifically nitric oxide. We have shown that like HmpA, PrxA protects V. cholerae against the nitric oxide donor spermine NONOate. Similarly, HmpA and PrxA both protect V. cholerae against acidified nitrite. This agrees with previous work showing that Hmp protects Salmonella against nitric oxide and acidified nitrite [34]. While the species(s) produced by acidified nitrite that HmpA and PrxA defend against is not clear, we presume that it is a RNS. PrxA homologs are not as prevalent in bacteria as HmpA homologs, but they are found in several pathogens including Yersina pestis, Haemphilus influenza and Neisseria gonorrhoeae. It will be interesting to determine if PrxA homologs share a similar RNS defense role in other bacteria.
Our observation that HmpA and PrxA are required for colonization lead us to suggest that V. cholerae encounters RNS stress during infection. When studying ROS defense genes we found that deletion of SodB was detrimental for normal V. cholerae growth. This indicated that normal growth of V. cholerae must generate a significant amount of superoxide managed by SodB. The growth defect of the DsodB mutant prevented us from analyzing it by competition in the mouse model. While we did not identify any other single ROS defense enzyme that affected intestinal colonization it is possible that construction of various double mutants may show that V. cholerae must also deal with ROS during disease progression. In bacterial pathogens where SODs have been shown to be necessary for virulence, it is generally the periplasmic SOD that is required as this SOD encounters superoxide entering the cells from the environment [16,66]. However, the V. cholerae periplasmic SOD, SodC, was not required for intestinal colonization suggesting V. cholerae does not experience superoxide stress from the host.

Bacterial strains
Strains and plasmids are listed in Supporting Table S3. V. cholerae El Tor biotype strain C6706 and a spontaneous lacZ2 derivative of C6706, were used as parental (wild type -WT) strains. E. coli DH5a lpir and Sm10 lpir were used for cloning and conjugation, respectively. Antibiotic concentrations used were streptomycin (Sm: 100 mg/ml or 500 mg/ml), kanamycin (Kan: 50 mg/ml), carbenicillin (Carb: 75 mg/ml) and chloramphenicol (Cm: 2.5 mg/ml for C6706 and 10 mg/ml for E. coli DH5a lpir). LB was used for all growth conditions [10 g/liter of tryptone (Bacto), 5 g/liter of yeast extract (Bacto), and 5 g/liter of NaCl] and was supplemented with 16 g/liter of agar (Bacto) for growth on plates. Arabinose was used at 0.1% for complementation assays. All ID numbers/ Accession numbers/for genes highlighted in this study are shown in Table S5.

DNA manipulations
The genomic sequence of C6706 has not been completed. We used the sequence of the close relative, N16961, for clone construction. For in-frame gene deletions of nfo, mutS, hmpA and prxA, genomic DNA surrounding the respective gene was amplified by crossover PCR and cloned into pWM91 for subsequent sacBmediated allelic exchange in V. cholerae, as described [67,68]. For complementation constructs, the respective gene was amplified from chromosomal DNA and cloned into plasmid pBAD18 after digestion with KpnI and SalI. The respective gene was induced by adding arabinose to the growth medium. All cloning products were sequence-verified, and the nucleotide sequence of all primers used is listed in Table S4.

Infant mouse colonization competition assays
A modified version of the protocol of Baselski and Parker [69] was performed for infection and recovery of C6706 derived strains. C6706 or C6706 lac 2 and mutant strains were grown on LB-agar plates with Sm overnight at 37uC. Wild type and mutant strains were mixed together in LB. 50 ml of this competition mixture (,50 000 bacteria) was inoculated into a 5-day-old CD1 mouse pup (Charles River Company). Serial dilutions of the competition mixture were plated in LB+Sm 100 and enumerated to determine the input ratio of wild type and mutant strain. After incubation at 30uC for 3 h or 18 h the mouse pups were sacrificed and small intestines were removed and homogenized in 10 ml of LB. Serial dilutions were plated in LB+Sm 100 and enumerated to determine the output ratio of wild type and mutant strain. The competitive index for each mutant is defined as the input ratio of mutant/wild type strain divided by the output ratio of mutant/ wild type strain. A minimum of four mice were assayed for each mutant strain. The in vivo experiments for the transposon and clean deletion strains were the accumulation of results performed on different days. For ease of communication we reported the average competitive index.

Mutation frequency assays
Rifampicin resistance assays. i) For DmutS mutation frequency and complementation assays cultures were grown to saturation at 37uC in LB Sm 500 or Sm 100 . 500 ml of culture was plated on LB agar+50 mg/mL rifampicin. After overnight growth at 37uC rifampicin resistant colonies were scored. ii) For mouse passaged assays, 5 day old mouse pups were inoculated with 50, 000 cells of wild type V. cholerae. After incubation at 30uC for 18 h the mouse pups were sacrificed and their small intestines removed and homogenized in 10 ml of LB+Sm 500 . The 10 ml of homogenized intestine was passed through cheese cloth and a 3.1mm filter. This filtering retained .90% of V. cholerae and removed the majority of eukaryotic materials as determined by western blot against mouse actin (data not shown). We recovered ,250 000-500 000 V. cholerae cfu per small intestine. The filtrate was grown to saturation. For the control experiment 50 000 wild type V. cholerae were inoculated into 10 mL of LB+Sm 500 and grown to saturation. We then plated an equal number of cfu from both mouse passaged and control cultures on LB agar+50 mg/mL rifampicin and scored resistant colonies. Control mouse samples in which no V. cholerae had been inoculated did not grow in LB+Sm 500 . Primers used for sequencing rpoB are shown in Supporting Table S3.
Trimethoprim resistance assays. A modified version of the Belfort and Pedersen-Lane protocol [44] was used for identified trimethoprim resistant colonies. For mouse passaged trimethoprin assays, 5 day old mouse pups were inoculated with 50 000 cells of wild type V. cholerae. After incubation at 30uC for 18 h the mouse pups was sacrificed and their small intestines removed and homogenized in 10 ml of LB Sm 500 +50 mg/mL thymine. The 10 ml of homogenized intestine was passaged through cheese cloth and a 3.1mm filter. This filtering retained .90% of V. cholerae and removed the majority of eukaryotic materials as determined by western blot against mouse actin (data not shown). We recovered ,250 000-500 000 wild type V. cholerae cfu per small intestine. The filtrate was grown to saturation. For the control experiment 50 000 wild type V. cholerae were inoculated into 10 mL of LB Sm 500 +50 mg/mL thymine and grown to saturation. We then plated an equal number of cfu from both mouse passaged and control cultures on M9 agar+0.1% CAS, 0.2% glucose, 50 mg/mL thymine and 20 mg/mL trimethoprim. After overnight growth at 37uC trimethoprim resistant colonies were scored. Control mouse samples in which no V. cholerae had been inoculated did not grow in LB+Sm 500 . The nucleotide sequences of the primers used for sequencing thyA are shown in Supporting Table S4.
To calculate the relative mutation frequency we plated equal numbers of cfu for both mouse and passaged and control samples. We calculated the average and standard error for the mutation rate for the control samples. Next we normalized the individual mutation frequencies from our 5 mice passaged samples and 5 control samples to the average control sample mutation frequency. This normalized the average control sample mutation frequency to 1 and showed the relative mutation frequency increase in mouse passaged samples.

Stomach pH and nitrite concentration determination
We have developed a fluorescence based assay to determine the pH of the infant mouse stomach. We first determined a standard curve using the fluorescent pH indicator Yellow/Blue DND-160 (Invitrogen) over a range from pH 3-8. We then extracted the gastric juice from 5 individual mice, diluted the sample 1:2 with ddH 2 O (pH 7), added Yellow/Blue DND-160 and determined the fluorescence of the solution. Comparing these fluorescent values to our standard curve we determined the pH of the infant mouse stomach to be 4.560.1. We also note that this is a conservative measurement. In order to obtain enough liquid we diluted the gastric sample ,1:2 with ddH 2 0 that was at ,pH 7. Thus while water is not a buffer, the dilution of the gastric juice likely raised the final pH of our measurements.
Nitrite concentration was determined using the Griess Reagent System (Promega TB229). The concentration shown is the average of 10 mice treated with or without sodium bicarbonate.

Hydrogen peroxide sensitivity assays
Strains were grown to exponential phase in LB with Cm when required. Strains were serial diluted and spotted on LB plates containing increasing concentrations of hydrogen peroxide and incubated at 37uC overnight. For complementation Cm and arabinose were added while strains were growing in liquid, as well as in the LB agar plates.

Nitric oxide sensitivity growth curves
Strains were grown to exponential phase in LB. Strains were then diluted to OD 600 0.01 in LB61.0 mM spermine NONOate and grown at 37uC in a 96 well plate with aeration (SpectraMax Plus 384, Molecular Devices). OD 600 readings were taken every 15 min.

Growth in acidified nitrite
Overnight cultures were diluted into LB and grown to log phase at 37uC with aeration. Cultures were diluted to OD 600 0.05 in 25 mM MES buffered LB of pH 7.0 or 5.5 with or without the addition of 400 mM sodium nitrite (Sigma-Aldrich). The LB media and MES were adjusted to a pH of 7.0 and 5.5 (Corning pH meter 240) with additions of HCl, and filter sterilized (0.22 mm, Corning) prior to use. The growth of strains under various treatments were determined by OD 600 measurement using a 96 well format spectrophotometer (SpectraMax Plus 384, Molecular Devices). Environmental parameters were set to 37uC with shaking and readings were taken every 15 minutes for 16 hours. Studies were conducted in quadruplicate.

Fluoroscein assay
Overnight cultures were diluted into 100 mL LB with Sm 100 and grown to OD 600 ,0.8 (37uC, aeration). 10 mL of culture was dispensed into 15 mL conical and centrifugated at 5,000 RPM for 5 minutes. The supernatant discarded and cells resuspended in an equal volume of 25 mM MES buffered LB of pH 7.0 or 5.5 with or without the addition of sodium nitrite (500 mM, 1 mM, 5mM, or 10 mM). Cells were treated for 1.5 hours at 37uC with aeration then centrifugated at 5,000 RPM for 5 minutes at 4uC. The supernatant was discarded, cells resuspended in 1 mL PBS (LONZA), and transferred to a 1.5 mL eppendorf tube. The cells were centrifugated and washed an additional two times in 16PBS before being resuspended in 1 ml of PBS with 10 mM 29,79dichlorodihydrofluorescein diacetate (Molecular Probes, Invitrogen). The cells were incubated at room temperature for 30 minutes then centrifugated and washed three times to remove all free, extracellular dye. The cells were lysed in 225 mL of lysis buffer (MilliQ water with 0.1M EDTA) via sonication. Cell lysates were centrifugated at 15,000 RPM for 5 minutes, supernatant transferred to another 1.5 mL eppendorf tube and centrifugated again. Fluorescence was measured at 490 nm / 519 nm (excitation/emission) (SpectraMax Gemini XS). Fluorescence was normalized against protein concentrations, as determined by Bradford assay. Studies were conducted in triplicate.

Statistical methods
Statistical significance was assessed for mouse colonization assays and DmutS mutation frequency assays using a one-way analysis of variance (ANOVA) using a Bonferroni post test to determine significant differences in competitive index between all pairs of V. cholerae mutants used in our study. Statistical significance of acidified nitrite, nitric oxide and H 2 O 2 sensitivities was assessed using a mixed model, repeated measures two-way analysis of variance (ANOVA), generating a p value for each pair wise curves over the concentration range of H 2 O 2 to determine the significance of our results. Statistical significance of rifampicin and trimethoprim resistant mutants from LB vs. mouse passaged samples were assessed using a paired t-test. Differences were considered significant at p,0.05. All calculations were performed using Graphpad Prisim version 5.