Genetic and Biochemical Analysis of Anaerobic Respiration in Bacteroides fragilis and Its Importance In Vivo.

Bacteroides species are abundant in the human intestine and provide numerous beneficial properties to their hosts. The ability of Bacteroides species to convert host and dietary glycans and polysaccharides to energy is paramount to their success in the human gut. We know a great deal about the molecules that these bacteria extract from the human gut but much less about how they convert those molecules into energy. Here, we show that B. fragilis has a complex respiratory pathway with two different enzymes that transfer electrons from NADH to quinone and a third enzyme complex that may use an electron donor other than NADH. Although fermentation has generally been believed to be the main mechanism of energy generation in Bacteroides, we found that a mutant lacking one of the NADH:quinone oxidoreductases was unable to compete with the wild type in the mammalian gut, revealing the importance of respiration to these abundant gut symbionts.

acceptor, with fumarate reductase (FRD) transferring electrons from the reduced quinone, producing succinate (47,48,50). To our knowledge, no studies have analyzed NADH:quinone oxidoreductase activity in Bacteroides species, and the complexes involved in this important energy generation step in Bacteroides species are unknown. Phylogenetic studies have identified the genes for both NUO (51) and NQR (39) in the Bacteroides genome, but no detailed biochemical or mutational analyses have demonstrated the involvement of these complexes in respiration. NQR activity has been identified in the related organism Prevotella copri, where a study of central carbon metabolism found NADH:quinone oxidoreductase activity in isolated cell membranes (52). This activity was attributed to NQR, as several of the genes coding for NUO subunits are absent in Prevotella copri. Similar results have also been reported for Prevotella bryantii (53). In the oral bacterium Porphyromonas gingivalis, the RprY response regulator positively activates NQR by interacting with the promoter upstream of the nqrA gene (54), and under conditions of oxidative stress, the production of RprY decreases, as does the expression of nqrA (55). In addition, the RprY regulator of P. gingivalis was shown to bind the promoter region of the nqr operon of Bacteroides fragilis, suggesting that the regulation of this operon is similar in these two bacteria.
In addition to anaerobic respiration, Bacteroides species are also capable of aerobic respiration. When Bacteroides strains are grown under nanaerobic conditions (0.05 to 0.15% oxygen), similar to the conditions that may be present at the mucosal lining of the gut (56), oxygen can serve as the terminal electron acceptor. Bacteroides species contain cytochrome bd oxidase, a high-affinity oxidase that functions under lowoxygen conditions, transferring electrons from reduced quinone to oxygen, producing water (21,57). During nanaerobic respiration in Bacteroides, cytochrome bd oxidase contributes to the proton motive force during electron transfer, thereby conserving energy (58).
Here, we show that B. fragilis has orthologs of all three diverse NADH:quinone oxidoreductases, NUO, NQR, and NDH2. Through the creation and analysis of single and double deletion mutants, combined with activity measurements of isolated membrane fractions, we show that both NQR and NDH2 have functional NADH dehydrogenase activity but that NUO does not. We found that a mutant lacking NQR was unable to compete with wild-type (WT) bacteria in a gnotobiotic mouse competitive colonization model. In contrast, a mutant lacking NUO showed no competitive colonization defect, while a mutant lacking NDH2 had a modest colonization defect. This is the first study demonstrating the importance of NQR to bacterial fitness in the colonization of a mammalian host. On the basis of these results, we propose a new paradigm of respiration in Bacteroides species that is more complex than previously appreciated, with NQR playing a critical role in energy generation.

RESULTS
Identification of putative NADH:quinone oxidoreductase genes in B. fragilis. Analysis of the genome sequence of B. fragilis strain 638R revealed that it contains each of three described types of NADH:quinone oxidoreductases: NUO (BF638R_0850-0841), NQR (BF638R_2136-2141), and NDH2 (BF638R_1612). Figure 1 shows the organization of each of these genes/operons compared to their organization in the corresponding regions of Escherichia coli for nuo and ndh2 and of Vibrio cholerae for nqr, as E. coli lacks nqr. The three genes shown in green in the E. coli nuo operon but missing in the corresponding B. fragilis operon code for the soluble portion of the enzyme complex that includes the NADH binding site. The presence of genes for these three different types of electron transfer complexes suggests that B. fragilis has a respiratory chain more complicated than previously appreciated, containing enzymes that have been shown in other organisms to pump H ϩ (NUO) or Na ϩ (NQR), as well as NDH2, which was not previously reported in the Bacteroidales.
Measurement of NADH:quinone oxidoreductase activity of isolated membranes. To study the role of each of these enzymes in the respiration of B. fragilis, we made mutants with an internal deletion in each gene/operon, performed biochemical Analysis of Anaerobic Respiration in B. fragilis ® activity measurements, and compared the NADH:quinone oxidoreductase activity in each of the mutants to that in the WT strain. Figure 2 shows measurements of NADH:quinone oxidoreductase activity measured in two ways: by the steady-state rates of NADH oxidation and by the reduction of menadione. For the nuo deletion mutant, there was no statistically significant difference in the activities in the mutant compared to that in the WT. This was the result that we predicted, based on the lack of nuoEFG in the nuo operon. In contrast, in the ndh2 and nqr deletion mutants, the activities were significantly decreased and were approximately 78% and 42% of the WT values, respectively. These results strongly suggest that under the growth conditions used in these experiments, the NADH:quinone oxidoreductase activity of B. fragilis is a function of NQR and NDH2 in a roughly 2:1 ratio, with no contribution from NUO, as predicted based on the lack of nuoEFG. Full NADH:quinone oxidoreductase activity was restored in the nqr and ndh2 complemented strains ( Fig. 2A), indicating that the activity lost was due to these deletions.
Three double deletion mutants were created to provide more specific information about the contribution of each individual enzyme complex. As illustrated in Fig. 2B, each double deletion mutant could synthesize only one of the three enzymes, allowing measurements to focus on the activity of the one remaining enzyme (Fig. 2C). The activity in the Δndh2 Δnuo strain was approximately 61% of that in the WT, which was similar to that of the ndh2 single deletion mutant, as expected, since we showed that the nuo single deletion mutant had full NADH:quinone oxidoreductase activity ( Fig. 2A). Activity in the Δnuo Δnqr strain was approximately 36%, which was similar to that in the nqr single deletion mutant. These results strengthen our conclusion that NUO does not contribute to NADH:quinone oxidoreductase activity, as predicted. Furthermore, the Δnqr Δndh2 deletion mutant exhibited essentially no measurable activity (Fig. 2B), supporting the suggestion that NUO is not an NADH:quinone oxidoreductase in B. fragilis.
The Δnqr Δndh2 mutant was created by making an ndh2 deletion in the Δnqr strain. During resolution of the ndh2 cointegrate, we noticed that both normal and small-size colonies were produced and determined that the small colonies were those of the Δnqr Δndh2 mutant (Fig. 2D). This growth defect likely results from the greatly reduced ability of this mutant to recycle NADH. In addition, this mutant likely obtains most of its energy by substrate-level phosphorylation, which yields less ATP per molecule of glucose, which may also contribute to the slower growth. When the ndh2 operon was added back to the Δnqr Δndh2 mutant, the NADH:quinone oxidoreductase activity was recovered (Fig. 2B). fragilis, we utilized deamino-NADH (d-NADH), a substrate analog of NADH (Fig. 3). d-NADH has been widely reported in other species to donate electrons to NUO and NQR but not to NDH2 (59)(60)(61). This specificity makes it possible to discriminate between NDH2 activity and the other NADH dehydrogenase activities by comparing the activities with the two different substrates. To confirm that d-NADH has the same specificity in B. fragilis, we first measured the activities with each of the substrates in the two double deletion mutants that retained only one type of NADH:quinone oxidoreductase , the single deletion mutant strains (the Δnuo, Δndh2, and Δnqr strains) and complemented strains (the Δnqr attB2::pnqr ϩ and Δndh2 attB2::pndh2 ϩ strains) (A), and double deletion mutant strains (the Δndh2 Δnuo, Δnuo Δnqr, Δnqr Δndh2, and ΔnqrΔ ndh2 attB2::pndh2 ϩ strains) (B). The rates of NADH oxidation (NADH ox.; orange) and menadione (vitamin K 3 ) reduction (Q red.; gray) using NADH were measured. Mean values (n ϭ 3 to 5 experiments) are presented, and error bars show the standard errors. Data were analyzed by Dunnett's test. *, P Ͻ 0.05 between the wild type and each mutant strain. (C) Diagram showing the enzyme that is present in each double mutant, i.e., the enzyme whose activity is measured. The letters P and C at the left-hand side of the membrane indicate the periplasm and cytoplasm, respectively. (D) Colony size comparison between the Δnqr Δndh2 mutant (small colonies) and the Δnqr single-deletion-mutant parent strain (large colonies). The photograph was taken after 5 days of growth.
Analysis of Anaerobic Respiration in B. fragilis ® activity, the Δndh2 Δnuo mutant, which contains only nqr, and the Δnuo Δnqr mutant, which contains only ndh2, as well as the WT. In the WT and the mutant containing only nqr, the activities with d-NADH were slightly lower than those with NADH, consistent with reports that d-NADH is a slightly less efficient substrate. However, in the mutant containing only ndh2, the activity was almost completely absent. Thus, d-NADH can be used in B. fragilis to discriminate the NDH2 activity in the presence of other NADH: quinone oxidoreductases.
Based on this finding, we then used NADH and d-NADH to analyze single deletion mutants. The activity of the Δndh2 mutant was approximately the same with either NADH or d-NADH, whereas the Δnqr mutant was active with NADH but had very low activity with d-NADH. These results are another confirmation that B. fragilis NUO does not use NADH as an electron donor.
In vitro growth analysis of single and double deletion mutants. We investigated the growth characteristics of the single and double deletions in rich medium under anaerobic conditions ( Fig. 4; see also Table S1 in the supplemental material). With the exception of the Δnqr Δndh2 mutant, all of the strains had doubling times within 13% of the doubling time of the WT and reached similar maximum values of the optical density at 600 nm (OD 600 ) at the stationary-phase plateau. The Δnqr Δndh2 strain had a much longer doubling time (176 min) than the WT and an OD 600 at the plateau of 70% of that of the WT. These changes were largely reversed when ndh2 was reintroduced into the strain, yielding a much faster doubling time of 75 min ( Fig. 4; Table S1).
Competitive colonization assays in gnotobiotic mice. As the nqr, ndh2, and nuo single deletion mutants did not demonstrate any significant in vitro growth defect compared to the growth of the WT, we tested their importance to bacterial fitness in vivo. For each mutant, competitive colonization assays (competing WT with the mutant) were performed using gnotobiotic mice ( Fig. 5; Table S2). After a week of competitive colonization, we found that the Δnqr mutant was outcompeted and was present as less than 1% of the total bacteria in the feces, as no Δnqr mutants were detected in the more than 100 colonies analyzed for each mouse. The ability of this mutant to competitively colonize was reestablished to a large extent, although not completely, when the nqr operon was restored to the Δnqr mutant. In contrast, the Δnuo mutant showed no significant difference in colonization from that of the WT in the competitive colonization assay, while the Δndh2 mutant showed a modest colonization defect, with the mutant colonization being reduced from 57.3% in the starting inoculum to 41.1% Ϯ 4.1% (standard deviation) in the feces at 1 week (P ϭ 0.0002). This colonization defect was also reversed when ndh2 was added back to this mutant (Fig. 5). These data support the findings of the activity assays showing that NQR is the most active NADH:quinone oxidoreductase and the most important enzyme complex of the three studied for bacterial colonization in this model system.

DISCUSSION
In this study, we defined the enzymes involved in the crucial first step of respiration in B. fragilis, the transfer of electrons from NADH to quinone (Fig. 6). While all three typical classes of NADH dehydrogenases are present in B. fragilis, we found that only two have this activity. Under the growth conditions used in this study, NQR was the most active and the most important NADH:quinone oxidoreductase. Deletion of nqr resulted in a mutant that was severely attenuated for competitive colonization of the mouse intestine. A second enzyme, NDH2, could also catalyze this step and similarly had a competitive colonization defect, albeit not as severe as that of the Δnqr mutant. In addition, B. fragilis has a homolog of complex I (NUO) for which we could not demonstrate NADH:quinone oxidoreductase activity, nor could we demonstrate a fitness defect of the Δnuo mutant in the mouse colonization model.

FIG 4
Growth curves for the WT, the single and double deletion mutants, and the Δnqr Δndh2 attB2::pndh2 ϩ strain in rich medium (BHIS medium) under anaerobic conditions. The growth of each strain (in duplicate or triplicate) was followed as the change in the absorbance at 600 nm in a BioTek plate reader. Doubling times were calculated as outlined in the Materials and Methods section, with all growth data being supplied in Table S1 in the supplemental material.
Analysis of Anaerobic Respiration in B. fragilis ® The lack of NADH:quinone oxidoreductase activity of NUO may be attributable to poor expression of the nuo operon under the conditions of our assay. Our analysis of published transcriptome sequencing data (62) revealed that the nuo operon is in fact poorly expressed during anaerobic growth in rich medium. However, the lack of NADH ) and feces (day [D7]) of the samples is shown as the mean and standard error of the mean (when applicable). The differences in abundance of the Δnqr and the Δndh2 mutants from days 1 to 7 were statistically significant (****, P ϭ 1EϪ40; **, P ϭ 0.0002), but they were not for the Δnuo mutant (P Ͼ 0.05 [n.s., not significant]). All experiments were performed with 3 male and 3 female germfree Swiss Webster mice. The phenotype was replicated in both the males and the females, and the breakdown by sex is shown in Fig. S1 in the supplemental material. shown as solid and broken arrows, respectively. NQR and NDH2 oxidize NADH to NAD ϩ and donate electrons to the menaquinone pool (Q/QH 2 ). Fumarate reductase (FRD) accepts electrons from reduced menaquinone and transfers them to the terminal electron acceptor, fumarate, which is reduced to succinate. B. fragilis NUO, lacking the NADH binding domain (NuoEFG subunits), likely oxidizes an unknown electron donor (X red. ) to transfer electrons via the menaquinone pool to FRD. In this scheme, energy is conserved when ion gradients are generated by NQR and NUO.

Ito et al.
® oxidation activity is likely also explained by NUO's unusual subunit composition. In comparison to the nuo operons of most bacteria, the nuoE, nuoF, and nuoG genes are missing in B. fragilis (Fig. 1). In a typical NUO complex, NuoE, NuoF, and NuoG make up a hydrophilic domain that includes the conserved NADH binding site. A NUO lacking this soluble domain should not be able to use NADH as a substrate. We analyzed 144 B. fragilis genomes and found that all have similar nuo operons lacking these three genes. The conservation of this operon in the species suggests that it has a different yet likely still important function in respiration. NUO complexes lacking these subunits have been described. The enzyme complex from Prevotella copri has the same 11subunit organization as that from B. fragilis (52), while the enzyme from Campylobacter jejuni does not have NuoE and NuoF but retains NuoG, giving it a total of 12 subunits (52,63). The more typical form of NUO, which uses NADH as a substrate, is exemplified by the complex from E. coli, with 14 subunits, including NuoE, NuoF, and NuoG (51). We predict that the B. fragilis NUO receives electrons from a distinct donor and transfers them to menaquinone while generating an H ϩ motive force (Fig. 6).
NQR is found only in prokaryotes and is present in many marine and pathogenic bacteria. In most organisms where NQR has been studied, the enzyme operates as a primary Na ϩ pump (64,65). Our findings suggest that Bacteroides generates an Na ϩ gradient directly through respiration by NQR, rather than depending completely on Na ϩ /H ϩ antiporters to generate an Na ϩ gradient. The primacy of NQR as the most active NADH:quinone oxidoreductase in B. fragilis and as the most important one for competitive colonization suggests that the Na ϩ gradient likely has a key role in the production and distribution of energy in these bacteria. In fact, more than 10 genes annotated as encoding Na ϩ -dependent transporters are present in the B. fragilis 638R genome.
B. fragilis is also capable of aerobic respiration, employing the terminal oxidase cytochrome bd, which is expected to conserve energy in the form of an H ϩ electrochemical membrane gradient (49,57). We predict that nanaerobic respiration confers a fitness advantage in the oxygenated environment near the colonic mucus layer. Thus, to fully understand the respiratory pathways in B. fragilis and their importance in energy conservation under different environmental conditions in the gut, it will be important to also better define the properties and enzymes of this nanaerobic respiratory chain.

MATERIALS AND METHODS
Bacterial strains and growth conditions. For all studies, we used a strain of B. fragilis 638R (TM4000) and its thy-negative mutant (ADB77). The strains were grown anaerobically at 37°C in brain heart infusion supplemented (BHIS) medium containing 0.5% (wt/vol) yeast extract, 5 g/ml hemin, and, when necessary, 50 g/ml thymine, as well as antibiotics (5 g/ml erythromycin, 200 g/ml gentamicin, or 40 ng/ml anhydrotetracycline [aTC]). Escherichia coli strain S17 pir and derivatives were grown in LB broth or plates, with 100 g/ml ampicillin or carbenicillin being added where appropriate. All strains and plasmids used are listed in Table S3 in the supplemental material. (BF638R_2136-2141) and nuo (BF638R_0841-0850) operon deletion. Upstream and downstream flanking regions of the nqr and nuo operons were amplified by PCR using the primers listed in Table S4 containing BamHI and NcoI sites for the upstream fragment and NcoI and HindIII sites for the downstream portion. These PCR products were digested and cloned by three-way ligation into BamHI-HindIII-digested pTY102 (66). The resulting plasmids were introduced into B. fragilis ADB77 by conjugation. Cointegrates were selected with erythromycin and gentamicin, and the cointegrate was passaged in nonselective medium and then plated on plates with 80 to 100 g/ml trimethoprim and 50 g/ml thymine to select against cointegrates (66). The resulting colonies were tested by PCR to differentiate mutant and WT resolvants. For animal experiments, the thy mutation in the background strain was restored to the WT by allelic exchange with the thy-positive suicide vector pYT102 (66).

Construction of deletion mutant strains. (i) nqr
(ii) Creation of pLGB36, an inducible counterselection vector for allelic deletions and replacements in B. fragilis 638R. Plasmid pLGB36 was made specifically to make allelic deletions and replacements in B. fragilis 638R and other B. fragilis strains that have the same type VI secretion system (T6SS) locus as strain 638R (67). This vector was created using the backbone of plasmid pLGB13 (68), which was designed for allelic replacements in Bacteroides and Parabacteroides species using the type VI secretion effector Bfe1 of strain 638R. As pLEC13 uses the 638R effector, strains with this T6SS region are immune to this toxin. In pLGB36, we used an effector from the T6SS region of B. fragilis strain S36_L11 (M136_1999), which we designated bfe-3, and replaced bfe-1 and its engineered signal sequence of Analysis of Anaerobic Respiration in B. fragilis ® pLGB13 with bfe-3 with the same signal sequence to target the effector to the periplasm, where it is toxic upon induction with aTC. We have not found any resistance to this toxin, nor have we encountered the growth of any colonies that are not double-cross-out resolvants, demonstrating the utility of this vector for allelic deletions and replacements in strain 638R and others encoding the Bfe1 effector/immunity pair.
(iii) ndh2 (BF638R_1612) deletion. The ndh2 deletion was created using pLGB36. Upstream and downstream flanking regions were cloned into the BamHI site of pLGB36 using the NEBuilder assembly tool (New England Biolabs) with the primers listed in Table S4 and transformed into E. coli S17 pir. The construct was verified by whole-plasmid sequencing and transferred by conjugation into TM4000. Cointegrates were selected on BHIS medium plates with gentamicin and erythromycin. A cointegrate was grown in basal medium (69) for 5 h and then plated with 40 ng/l aTC. Resolvants were screened by PCR and mutants were selected.
Complementation of mutant strains. (i) nqr operon complementation. The complete nqr operon with its native promoter (6,563 bp) was amplified by PCR using primers with a NotI restriction site at the 5= end and a BamHI restriction site at the 3= end (Table S4). The digested PCR product was cloned into NotI-and BamHI-digested pNBU2-bla-ermGb, which integrates into the chromosome and is tranformed into S17 pir. The construct was sequenced to confirm that no mutations were introduced. The plasmid was transferred by conjugation into the nqr deletion strain. Transconjugants with integration of the plasmid at the attB2 site of tRNA-Ser were selected by PCR using the primers listed in Table S4.
(ii) ndh2 complementation. The ndh2 gene with its promoter was cloned into pNBU2-bla-ermGb using the NEBuilder assembly tool with the primers listed in Table S4 and transformed into S17 pir. The construct was verified by whole-plasmid sequencing. Transconjugants with integration of the plasmid at the attB2 site of tRNA-Ser were selected by PCR, as described above.
Cell membrane preparation. All bacterial cultures were harvested in mid-logarithmic growth phase and washed with KPi buffer, containing 40 mM KH 2 PO 4 , pH 7, and 5 mM dithiothreitol. The cells were broken using a French press (24,000 lb/in 2 , 2 cycles) under anaerobic conditions in KPi buffer containing DNase and phenylmethylsulfonyl fluoride (PMSF) protease inhibitor at 4°C. Broken cells were removed by centrifugation (3,800 ϫ g), and the remaining supernatant was then centrifuged at 185,500 ϫ g overnight to separate the inner membranes. The inner membrane preparations were washed and resuspended in KPi buffer under anaerobic conditions and stored at Ϫ80°C. NADH:quinone oxidoreductase activity. The NADH:quinone oxidoreductase activity was followed spectrophotometrically at room temperature under an argon atmosphere in a buffer containing 50 mM Tris-HCl, pH 7, 1 mM EDTA, 100 mM NaCl, 5% (vol/vol) glycerol, 0.05% (wt/vol) n-dodecyl ␤-maltoside, 100 M NADH, and 50 M menadione (vitamin K 3 ). As it has been reported that menaquinone is the quinone produced by B. fragilis (70), menadione was used as the electron acceptor from NADH for all activity studies.
The oxidation of K 2 -NADH or nicotinamide hypoxanthine dinucleotide, reduced-form Na ϩ salt (deamino-NADH), was measured at 343 nm ( ϭ 6.22 mM Ϫ1 cm Ϫ1 ), where menadione and menadiol are isosbestic (71). The reduction of menadione was measured at 262 nm ( ϭ 14.0 mM Ϫ1 cm Ϫ1 ), and the results were adjusted by subtracting the calculated contribution of the accumulating NAD ϩ from the total absorbance at 262 nm. The absorbance of NAD ϩ at 262 nm ( ϭ 17.8 mM Ϫ1 cm Ϫ1 ) was calculated from the absorbance change at 343 nm. Membrane preparations from the WT or mutant strains were used at a protein concentration of 60 g/ml. The reaction was initiated by adding membranes to the buffer containing 100 M NADH and 50 M vitamin K 3 under an argon atmosphere. All activity measurements were performed at least three times.
Growth analysis. Cultures of the wild-type and mutant strains described in this paper were grown in BHIS broth overnight, diluted 1 to 10 in fresh broth, and grown to mid-exponential phase. A suitable volume (usually 10 l) was then used to inoculate 1 ml of prereduced BHIS broth in wells of a 24-well tissue culture plate. All incubations were at 37°C under anaerobic conditions. After 1 min of shaking, OD 600 readings were recorded every 10 min for 12 h using a BioTek Power Wave plate reader (all OD 600 data and doubling time calculations are provided in Table S1).
Doubling times were calculated for each strain on a segment of the logarithmic growth phase using the exponential growth equation [y ϭ y 0 ·exp(k·x)], with y being OD 600 , y 0 being OD 600 at time zero, and k being rate constant expressed in inversed minutes, with a least-squares fit, as implemented in Prism software (64-bit version 8.2.0 for Windows; GraphPad Software, Inc., San Diego, CA), and are presented as the average from at least two experiments (see the details in Table S1). The OD 600 data were graphed using Prism software after the data were transformed by averaging the values for 5 neighbors and using zero-order smoothing.
Competitive colonization assays in gnotobiotic mice. Mouse studies were approved by the Institutional Animal Care and Use Committee (IACUC) of Brigham and Women's Hospital and complied with all relevant ethical regulations for animal testing and research. Gnotobiotic mice were obtained from and maintained in the Harvard Digestive Diseases Center gnotobiotic core facility at Brigham and Women's Hospital and housed in sterile OptiMICE cages (Animal Care Systems, Centennial, CO). All experiments were performed using Swiss Webster germfree mice that were 5 to 8 weeks old. For each experiment, both male and female mice (groups of three male and three female mice) were used. To differentiate the strains for quantification, the WT B. fragilis TM4000 strain was made tetracycline resistant by introduction of pNBU2-bla-tetQ into the attB2 tRNA-Ser site, and the mutant strains were made erythromycin resistant by integration of pNBU2-bla-ermGb. For experiments with complemented mutant strains, the pNBU2-bla-ermGb constructs containing the cloned genes for complementation were similarly integrated into the attB2 tRNA-Ser site. Strains were grown to an OD 600 of approximately 0.6 and mixed at a 1:1 ratio, with the final percentage of each strain in the inoculum being determined by plating, as reported in Fig. 5 and Table S2. Mice were gavaged with 200 l of the bacterial mixtures, and 7 days later, fresh fecal samples were collected, diluted in sterile phosphate-buffered saline, and plated on BHIS medium plates. After colony growth, the bacteria were replica plated onto two plates containing either tetracycline or erythromycin to determine the exact percentage of each strain in the feces ( Fig. 5; Table S2). A one-sample t test was performed using the arcsine-transformed values of the proportions of the WT and mutant strains present in the samples on days 1 and 7. The Δnqr mutant competition data set had no standard error since there were no Δnqr mutant colonies detected, so the value 1EϪ13 was added to one sample so that the values of the t test could be computed.
Data availability. Plasmid pLGB36 has been deposited in Addgene under accession no. 135621 for distribution to the scientific community.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only. FIG S1, TIF file, 0.9 MB.