Enterococcus faecalis Sex Pheromone cCF10 Enhances Conjugative Plasmid Transfer In Vivo

ABSTRACT Cell-cell communication mediated by peptide pheromones (cCF10 [CF]) is essential for high-frequency plasmid transfer in vitro in Enterococcus faecalis. To examine the role of pheromone signaling in vivo, we established either a CF-producing (CF+) recipient or a recipient producing a biologically inactive variant of CF (CF− recipient) in a germfree mouse model 3 days before donor inoculation and determined transfer frequencies of the pheromone-inducible plasmid pCF10. Plasmid transfer was detected in the upper and middle sections of the intestinal tract 5 h after donor inoculation and was highly efficient in the absence of antibiotic selection. The transconjugant/donor ratio reached a maximum level approaching 1 on day 4 in the upper intestinal tract. Plasmid transfer was significantly lower with the CF− recipient. While rescue of the CF− mating defect by coculture with CF+ recipients is easily accomplished in vitro, no extracellular complementation occurred in vivo. This suggests that most pheromone signaling in the gut occurs between recipient and donor cells in very close proximity. Plasmid-bearing cells (donors plus transconjugants) steadily increased in the population from 0.1% after donor inoculation to about 10% at the conclusion of the experiments. This suggests a selective advantage of pCF10 carriage distinct from antibiotic resistance or bacteriocin production. Our results demonstrate that pheromone signaling is required for efficient pCF10 transfer in vivo. In the absence of CF+ recipients, a low level of transfer to CF− recipients occurred in the gut. This may result from low-level host-mediated induction of the donors in the gastrointestinal (GI) tract, similar to that previously observed in serum.

IMPORTANCE Horizontal gene transfer is a major factor in the biology of Enterococcus faecalis, an important nosocomial pathogen. Previous studies showing efficient conjugative plasmid transfer in the gastrointestinal (GI) tracts of experimental animals did not examine how the enterococcal sex pheromone response impacts the efficiency of transfer. Our study demonstrates for the first time pheromoneenhanced, high-frequency plasmid transfer of E. faecalis plasmid pCF10 in a mouse model in the absence of antibiotic or bacteriocin selection. Pheromone production by recipients dramatically increased plasmid transfer in germfree mice colonized initially with recipients, followed by donors. The presence of a coresident community of common gut microbes did not significantly reduce in vivo plasmid transfer between enterococcal donors and recipients. In mice colonized with enterococcal recipients, we detected plasmid transfer in the intestinal tract within 5 h of addition of donors, before transconjugants could be cultured from feces. Surprisingly, pCF10 carriage provided a competitive fitness advantage unrelated to antibiotic resistance or bacteriocin production. KEYWORDS antibiotic resistance transfer, cell-cell signaling, competitive fitness, gut microbiota, mobile genetic element H orizontal gene transfer by conjugation is a major contributor to genome plasticity and evolution in Enterococcus faecalis (1). This Gram-positive intestinal commensal bacterium is also a major nosocomial pathogen, with infections often following dysbiosis triggered by antimicrobial therapy. E. faecalis has high intrinsic resistance to several antibiotics, and a multitude of acquired antibiotic resistance genes on mobile genetic elements further complicate antimicrobial therapy (2). Determination of the genome sequence of V583, the first vancomycin-resistant enterococcal isolate in North America (3), provided a paramount example of the propensity of this organism to acquire multiple mobile elements in the clinical setting, including conjugative pheromone-responsive plasmids (4).
Pheromone-responsive plasmids in E. faecalis are the most extensively characterized conjugative plasmids among Gram-positive bacteria. The complex regulatory circuits controlling transfer of the pheromone-responsive plasmids pAD1 and pCF10 have been the major focus of these studies (5). Pheromone-responsive plasmids carry genes that can encode antibiotic resistance and hemolysin/bacteriocin production, but these plasmids can also be devoid of genes encoding detectable phenotypic markers (6,7). Conjugative transfer of the tetracycline resistance plasmid pCF10 is initiated in the donor by sensing cCF10 (CF), a 7-amino-acid hydrophobic peptide encoded within the chromosome. A pCF10-encoded inhibitory peptide iCF10 (iCF) modulates the pheromone response by direct competition with CF (Fig. 1A); the induction state of an individual donor cell depends on the intracellular molar ratio of the two peptides (5). Expression of aggregation substance and other proteins on the cell surface of the donor follows pheromone uptake, with subsequent formation of macroscopic aggregates and plasmid transfer at an efficiency of up to 10 Ϫ1 transconjugants/donor (T/D) in liquid medium in vitro. In donors carrying pCF10, a low level of induction (in the absence of recipients) can also be triggered by host factors in serum, via a mechanism involving differential sequestration of donor-produced iCF and CF peptides (8,9).
While the pheromone response has been well characterized in vitro, the role of this form of cell-cell communication in the natural habitat of E. faecalis has not been determined. Efficient transfer of pheromone-responsive plasmids in the gastrointestinal (GI) tracts of hamsters and pigs following antibiotic-induced suppression of the normal gut microbiota has been observed (10,11). More recently, transfer was demonstrated in conventional mice following long-term administration of donors and recipients in drinking water to enhance stable colonization (12). None of these studies directly examined the effects of pheromone signaling on transfer, and interpretation of the results was affected by either pretreatment of animals with antibiotics, by the use of plasmids encoding bacteriocin production (which could select against plasmid-free cells), or both.
Our goal was to answer several basic questions related to plasmid transfer of pCF10 in vivo. We established a germfree mouse model to investigate the kinetics and localization of pCF10 plasmid transfer without interference by resident microbiota or antibiotics. Plasmid transfer in the intestinal tract was detected a few hours after donor inoculation in the intestinal tract, before either donors or transconjugants appeared in feces. Introduction of a competing defined microflora did not impact plasmid transfer. Plasmid transfer also occurred in the absence of pheromone production by the recipient, but at significantly lower rates, suggesting a basal level of donor induction by the host. Efficient transfer into recipients producing a biologically inactive variant of CF (CFϪ recipients) was not restored in the intestine by coresident CF-producing recipients (CFϩ recipients), suggesting that natural pheromone signaling occurred in mi-croniches where mating cells were colocalized. The data also suggest that pCF10 may contribute to competitive fitness of its host cell in the gut by an unknown mechanism.

RESULTS
Efficient plasmid transfer in the germfree mouse. An overview and summary of the experimental setup and time line are depicted in Fig. 1B. These protocols were based on preliminary experiments analogous to experiments 1 and 3 below, and where very similar results were obtained (see Tables S1 and S2 in the supplemental material). All mice were initially monocolonized with recipients that were allowed to establish for 3 or 4 days prior to addition of donors. In experiment 2, mice were inoculated with a defined microbial community in the interim between inoculation of recipients and donors. Preliminary experiments indicated that plasmid transfer could be detected by plating feces 1 day after donor inoculation, but not 5 h after donor inoculation (Table S1). On the basis of these results, we plated both feces and homogenized intestinal tissue segments 1, 4, and 7 days after donor inoculation. In addition, we sacrificed and plated homogenized intestinal segments from two mice 5 h after donor inoculation. This enabled us to assess plasmid transfer in the intestinal tract at an early time point when feces were still negative for transconjugants, confirming transfer within the GI tract as opposed to transfer in expelled feces and subsequent coprophagic reingestion. Experiment 1 used wild-type recipient and wild-type donors. Recipients, donors, and transconjugants established at the highest cell density in the lower intestinal tract, For experiments 1 and 2, a CFϩ recipient initially established, while in experiment 3, a CFϪ recipient was first established. For all three experiments, two mice were sacrificed 5 h after the inoculation with the donor, the isolated intestine was divided into three sections, and donors, recipients, and transconjugants were enumerated by plating on selective media. Four mice from one cage were sacrificed on days 1, 4, and 7 following donor inoculation, and enterococci associated with intestinal tissue were enumerated. Fecal samples were taken for all individual animals on days 1, 4, and 7. For experiment 2 with the competing microbial community, the CFϩ recipient strain was inoculated on day Ϫ4, with a defined multispecies community inoculated on day Ϫ3, and donors on day 0.
Pheromone-Induced Plasmid Transfer in the GI Tract When we assessed the efficiency of plasmid transfer by enumerating the transconjugants per donor (T/D) ratio (Fig. 2D), we detected T/D ratios in the intestinal tracts between 10 Ϫ2 to 10 Ϫ1 5 h after donor inoculation. These ratios remained at that level in all sections of the intestinal tract throughout the experiment, suggesting that even at the low densities found at the earliest time point, the donor populations had good access to recipients. A significantly higher T/D ratio was observed on day 4 in the upper intestinal tract (P ϭ 0.0148 for upper and middle sections; P ϭ 0.0104 for upper and lower sections), whereas the difference in the T/D ratio for the middle and lower sections of the intestinal tract was not significant.
Cell numbers determined for donors, recipients, and transconjugants collected from feces mirrored trends seen in intestinal tract sections. At later time points, recipient populations decreased slightly, while both donors and transconjugants increased (Fig. 3). The expansion of the donor cell and transconjugant populations despite competition by a large excess of plasmid-free resident recipients suggests that pCF10 provides its bacterial host with a competitive advantage in the absence of antibiotic selection or a plasmid-encoded bacteriocin ( Fig. 4A and B). Due to their earlier establishment in the mice, recipient numbers at the time of introduction of the donor outnumbered the donors from 10 3 to 10 4 to one 5 h after donor inoculation. This ratio shifted down by the end of the experiment to approximately 80:1 recipient to donor Effects of a defined microbial community on plasmid transfer. Experiment 2 investigated the effect of competing microflora on plasmid transfer; mice colonized with cCF10-producing strains (CFϩ) were inoculated with a defined community (Bacteroides ovatus, Bacteroides vulgaris, Clostridium sporogenes, Akkermansia muciniphila, Faecalibacterum prausnitzii, and Roseburia faecis) 3 days before introduction of the donor (Fig. 5).
Five hours following the introduction of donors into mice precolonized with both recipients and a defined community, transconjugants were recovered from the upper intestinal tract from only one of two mice and from the lower intestinal tract from none of the two mice. All mice sacrificed 1 day after donor addition harbored high levels of transconjugants (T/D ratio of 10 Ϫ2 in all intestinal sections) with a peak T/D ratio (10 Ϫ1 )  reached in the upper intestinal tract at day 4 and a slight decrease at day 7 (Fig. 5D). Transfer in the upper intestine appeared to be slightly more efficient, but T/D were only significantly higher compared to the lower intestinal tract (P ϭ 0.0286 for the upper/ lower sections; P ϭ 0.2 for the upper/middle sections). A comparison of T/D ratios from experiments with and without a competing community is shown in Fig. 6A to C. The T/D ratio on day 4 was more affected by the community in the lower intestine (P ϭ 0.0162), with no significant impact in the upper and middle intestinal sections (P ϭ 0.08 and P ϭ 0.28, respectively). Modest interference of the competing flora with plasmid transfer efficiency was also observed in the feces (P ϭ 0.02). A comparison of the overall cell numbers of recipients, donors, and transconjugants in both experiments shows that the total enterococcal population densities were barely affected (Fig. 6D). Analysis of the microbial community in feces by 16S rRNA gene sequencing revealed that two members of the introduced competing community (F. prausnitzii and R. faecis) did not establish in our system. At the end of the experiment, A. muciniphila comprised 30% of the total population, while B. ovatus, B. vulgatus, and E. faecalis each comprised about a fifth of the total population. C. sporogenes was present at approximately 3%. Over the course of the experiment, the number of C. sporogenes showed the least variation, suggesting a well-defined niche ( Fig. S1 and S2). F. prausnitzii and R. faecis, the least aerotolerant species studied, were not detected in the communities, suggesting a possible loss of viability during preparation of the initial inoculum or gavage.
Effects of cCF10 production by recipients on plasmid transfer in the intestinal tract suggest that pheromone signaling is highly localized in vivo. Figure 7 shows a very high level of pCF10 transfer into two isogenic pheromone-producing strains (CFϩ) with different selective antibiotic resistance markers in 4-h broth matings in vitro. In contrast, an isogenic recipient with an engineered mutation resulting in production of a V2A biologically inactive variant of mature cCF10 (CFϪ), acquired pCF10 at Ͻ1/10,000 the wild-type level. Transfer into the CFϪ recipient strain was substantially rescued during in vitro matings containing mixtures of CFϩ and CFϪ recipients if the CFϩ strain comprised at least 10% of the total recipient population. A partial rescue of the CFϪ phenotype was also obtained by the addition of 10% human serum to the mating mixture, increasing the T/D ratio to 10 Ϫ4 (Fig. 7).
To compare the effects of pheromone production by recipients in the intestine to those observed in vitro (Fig. 7), we carried out experiment 3, which also examined the ability of a CFϩ recipient to increase transfer into a CFϪ recipient by extracellular complementation. Transconjugants were recovered as early as 5 h after introduction of  Pheromone-Induced Plasmid Transfer in the GI Tract ® the donor (Fig. 8A), although the numbers were lower compared to the wild-type recipient ( Fig. 2A) (P ϭ 0.0286 by Mann-Whitney test for all intestinal sections). In two mice sacrificed 5 h after the addition of donors, the T/D ratio was 10 Ϫ3 (Fig. 8A), about 10-to 100-fold lower than the ratio with CFϩ donors (Fig. 2A). For the reminder of the experiment, the T/D ratio was highest on day 4 but without an apparent preference in location (P ϭ 0.7984 for upper/middle sections; P ϭ 0.0535 for upper/lower sections). Although the numbers of transconjugants increased over time as in experiment 1, the T/D ratio remained in the range of 10 Ϫ5 to 10 Ϫ4 . Levels of plasmid transfer into the CFϪ recipient were substantially lower at all time points; for example, both total transconjugants and transconjugants/donor values were reduced by 2 to 3 orders of magnitude on day 4, which showed the highest overall frequencies with either recipient ( Fig. 2A and D versus Fig. 8A and D). Results of the in vivo experiment depicted in Fig. 8A and D indicate higher frequencies of transfer (10-to 100-fold) into the CFϪ recipient in the intestine than in the in vitro experiment (Fig. 7). This is likely due to host factormediated low-level induction of donors in the intestinal tract, as observed in vitro in the presence of serum (Fig. 7), although other mechanisms are also possible.
To test whether the presence of pheromone-producing recipients in the intestinal tract could increase plasmid transfer into coresident pheromone-less recipients, as observed in vitro (Fig. 7), we gavaged some mice from experiments 1 and 3 that had harbored both a recipient strain (CFϩ in experiment 1 and CFϪ in experiment 3) and a wild-type donor strain for 4 days with a differentially marked recipient with the opposite pheromone production phenotype. In both experiments, the recipient added at day 4 never achieved population densities more than 2.5% (upper intestinal tract, experiment 1) of the total recipient population at day 7, while averaging 0.75% Ϯ 0.17% in all other sections and animals. Likewise, transfer into the CFϪ recipient was much less frequent than into the wild type when transconjugants/recipient were determined  Fig. 9A and B). In both experiments, plasmid transfer into the CFϪ recipient was not rescued by the presence of the CFϩ strain, regardless of whether the CFϪ recipient was added before or after the CFϩ recipient. This suggests that in the GI tract, effective pheromone signaling is restricted to donor and recipient cells in close proximity. Since we have previously shown production of microcolony biofilms on the mucosal surface throughout the length of the intestine (13), it is tempting to speculate that these adherent communities could be important ecologic niches for pheromone signaling and plasmid transfer in vivo.

DISCUSSION
The occurrence of highly efficient transfer of the E. faecalis pheromone-responsive plasmid pCF10 in a germfree mouse model was expected, since high transfer ratios had been reported in several mammalian model systems before (10)(11)(12). However, the in vivo role of pheromones in plasmid transfer has not been examined. We also note that all previous studies of enterococcal plasmid transfer in the GI tract used either animals treated with antibiotics, plasmids carrying genes that encoded markers that could select against recipients during the course of the experiment, or both. These factors limit the ability to determine the intrinsic efficiency of transfer. The use of a germfree model allowed us to examine pheromone signaling and plasmid transfer under defined conditions without antibiotic selection and to begin to assess how multispecies communities impact this process. We detected high-level plasmid transfer in the intestinal tract 5 h after inoculation of the donor, the earliest time point thus far reported for transfer in vivo. The fact that transconjugants and donors could not be detected 5 h after donor inoculation in feces strongly supports the hypothesis that plasmid transfer occurs in the intestinal tract and is not a result of extracorporeal plasmid transfer in the feces followed by coprophagy. Plasmid transfer was detected in all three experiments in five of six mice in the upper intestinal tract, all six mice in the middle section, and two of six mice in the lower intestinal tract. These observations suggest that the upper and middle sections of the intestinal tract are favorable, if not preferred, locations for transfer of pCF10. Lack of detection of early plasmid transfer in the lower intestinal tract in the mice with the competing microflora supports the view that the competing effect of the defined intestinal community is stronger in the environment with the greatest microbial density.
Plasmids are often considered to be a metabolic burden for the cell that may be selected for under certain conditions, such as the presence of an antibiotic. Interestingly, the percentage of plasmid-carrying cells (donors plus transconjugants) in the total enterococcal population increased during the experiments reported here and in preliminary experiments (data not shown). This suggests that pCF10 provides one or more factors that confer a selective advantage to plasmid-carrying cells in the absence of antibiotic or bacteriocin selection. A candidate for this selective advantage is Pheromone-Induced Plasmid Transfer in the GI Tract aggregation substance (AS), encoded by the prgB gene on pCF10. A previous study using an endocarditis model showed that a strain carrying a pCF10 derivative deleted of prgB was dramatically attenuated relative to both wild-type pCF10-carrying cells and to isogenic plasmid-free cells, suggesting that the prgB-free plasmid comprised a burden to the host strain (14). The selective advantage provided by AS in the endocarditis model has been attributed to the increased resistance to polymorphonuclear leukocyte (PMN)-mediated killing (15) provided by two RGD sequences in the protein and to increased colonization ability (14,16). The presence of AS could also provide a selective advantage for the donor in the intestinal tract. It is conceivable that AS allows the donors to expand the ecologic niche for E. faecalis in the intestinal tract by serving as an adhesin to host cells (17,18) and mucosal proteins and by attaching to the already established recipient microcolonies on the intestinal wall (13) with subsequent plasmid transfer. Aggregation substance is not the only gene product that might affect fitness. Since pCF10 carries many other genes (19), including uvr orthologs, and genes encoding numerous secreted cell envelope proteins that could impact competitive fitness in the GI tract, we are currently examining this issue comprehensively using a transposon sequencing approach. The introduction of a defined community after the recipient was inoculated but before the donor was administered had little effect on plasmid transfer in the upper and middle sections of the intestine but modestly reduced transfer in the lower intestinal tract. This might be expected, since the members of the defined community are normal inhabitants of the large intestine, and thus, it is likely that the reduced T/D rate is due to spatial interference with the interaction between donor and recipient. E. faecalis still comprised about 20% of the community in the large intestine based on 16S rRNA gene sequencing of the feces, consistent with the only marginally reduced numbers of enterococci recovered in comparison to the monospecies-colonized mice.
A major aspect of pheromone signaling in E. faecalis that has not been addressed is the role of the conjugation-inducing pheromones such as cCF10 in vivo. Licht et al. showed that cCF10 had biological activity in intestinal mucus in vitro (20); the high plasmid transfer rates in the previously described in vivo experiments would suggest that cCF10 induction was active under in vivo conditions, but direct evidence for the role of cCF10 had not been obtained. We used a strain deficient in the production of a functional cCF10 peptide as a recipient to investigate contributions of the peptide in vivo. With the mean T/D ratio 10 2 -to 10 3 -fold lower compared to the CFϩ recipient (day 4), the results from experiment 3 demonstrated the requirement of cCF10 for high-frequency plasmid transfer in vivo. However, plasmid transfer to the CFϪ recipient still occurred with a T/D ratio of 10 Ϫ4 , which was considerably higher than that observed in vitro with the same donor and recipient (Fig. 7). The observed level of plasmid transfer with a CFϪ recipient in vivo is most likely due to the previously described autoinduction of the donor by the donor cell's own cCF10 (9). Autoinduction is triggered by sequestration of the inhibitory peptide iCF10 by a host factor, possibly albumin/lipid complexes (9). We could indeed show that the addition of serum in vitro could partially restore plasmid transfer to the CFϪ recipient (Fig. 7). Currently, the identity of a host factor(s) that would cause donor induction in the intestinal tract is not known and will be the focus of future studies. Our attempt to restore efficient plasmid transfer to the CFϪ strain by addition of the CFϩ wild-type recipient to the mouse GI tract did not significantly increase the T/D ratio for the CFϪ recipient. This was most clearly demonstrated when the CFϩ recipient was established first and CFϪ was added at day 4, with a CFϩ/CFϪ ratio between 40 and 150:1 (experiment 1; Fig. 9A). This result indicates that recipient-produced cCF10 is not freely available throughout the intestine for donor induction. Plasmid transfer to CFϪ recipients could be initiated by autoinduction of the donor with subsequent expression of AS and binding to the resident recipient microcolonies. Subsequent plasmid transfer within the colony could be dependent on the production of CF by the other recipients in the colony. A recent study in germfree mice with a defined community indicated low bacterial density in the lumen of the intestinal tract particularly in the small intestine (21), suggesting that donors and recipients in our study may be too far apart for direct cell-cell communication in the lumen. In addition, these investigators showed that even in the large intestine, bacterial cells are concentrated on the intestinal wall (20). This would support our hypothesis that plasmid transfer most likely occurs in established microcolonies.
In all three experiments, plasmid transfer occurred at a high level in the intestinal tract as early as 5 h after donor inoculation, and before either donors or transconjugants could be detected in feces. In feces, where individual mice could be monitored over time, the rates of increase of recipients and donors between day 1 and day 4 were remarkably similar in all three experiments (see Fig. S3A in the supplemental material). Interestingly, while the number of recipients showed a tendency to decrease following donor colonization in experiments 1 and 3, the presence of the competing community appeared to abrogate this effect. The spread in rates was wider for transconjugants, also reflected in T/D ratio rates, indicating variability of plasmid transfer and proliferation in the recipient population between individual mice (Fig. S3B). The rate of increase of transconjugants or T/D ratio between day 1 and day 4 was, however, not significantly different in the three experiments. The influence of spatial structure in restricting plasmid transfer in biofilms has been observed before (22), and our results indicate that a similar mechanism could be occurring here. In samples enumerated directly from the intestinal tract, where T/D ratios were generally higher than in feces, the number of samples available did not provide sufficient statistical power for definitive conclusions, but the average T/D ratios on day 4 compared to the T/D ratios on day 1 suggested that the upper intestinal tract could be an equally, if not more favorable, niche than the lower intestine for pheromone-inducible plasmid transfer.
In summary, we have shown that pCF10 plasmid transfer occurs within 5 h of donor introduction in a germfree mouse model in the intestinal tract. Early sampling and results from the day 4 time points suggest that the upper intestinal tract provides the most favorable environment for plasmid transfer. The introduction of a simple multispecies community did not substantially impact E. faecalis cell numbers and plasmid transfer rate, except in the lower intestinal tract. The failure to complement the pheromone-negative strain with a wild-type recipient suggests the need for close contact between recipient and donor in vivo. These studies represent the initial investigation into how plasmid carriage and cell-cell signaling may impact the ecology of E. faecalis in the GI tract. The ability to introduce a variety of genetic constructs into this well-defined model and to vary the overall complexity of the community will ultimately enhance our understanding of important questions related to how antibiotics and other inducers of GI tract dysbiosis impact the transition of the enterococci from harmless commensal organisms to lethal nosocomial pathogens.

MATERIALS AND METHODS
Bacterial strains. In vivo transfer experiments employed either a CFϩ wild-type recipient OG1RF (23) and OG1ES (24) or an isogenic CFϪ strain producing the peptide LATLVFV rather than the native CF peptide LVTLVFV (Chandler et al. [9]). The Enterococcus faecalis plasmid donor for all experiments was the wild-type strain OG1Sp(pCF10) (C. J. Kristich, unpublished). Stocks of E. faecalis were made and frozen at Ϫ80°C; isolates were grown from stocks before each experiment. A defined community of Bacteroides ovatus, Bacteroides vulgatus, Clostridium sporogenes, Akkermansia muciniphila ATCC BAA-835, Faecalibacterum prausnitzii ATCC 27755, and Roseburia faecis DSMZ 16840 was used as competing GI microflora. For these bacteria, stock solutions were made of each strain in prereduced medium inside an anaerobic chamber and frozen at Ϫ80°C. Selective plates. CFϩ and CFϪ recipients were grown and enumerated on tryptic soy agar (TSA) plates with 50 g/ml rifampin, OG1ES on TSA with 10 g/ml erythromycin, and OG1Sp(pCF10) on TSA with 630 g/ml spectinomycin plus 10 g/ml tetracycline. Transconjugants were selected with 50 g/ml rifampin plus 10 g/ml tetracycline for OG1RF(pCF10) and 10 g/ml erythromycin plus 10 g/ml tetracycline for OG1ES(pCF10).
Gnotobiotic mice. All animal experiments were approved by the institutional IACUCs at both the Mayo Clinic, Rochester, MN (25,26), and at the University of Minnesota, with the latter giving final overall approval of the experiments (protocol 1511-33206A). Gnotobiotic experiments were done using germfree (GF) 5-to 6-week-old Swiss Webster mice (male and female) maintained and bred in sterile isolators (Class Biologically Clean, Ltd., Madison, WI) at the Mayo Clinic gnotobiotic facility. GF status of isolators and mice was confirmed using a combination of culture (Sabouraud dextrose medium, brain heart infusion medium, and nutrient broth medium at 37°C for 7 days under aerobic and anaerobic conditions) Pheromone-Induced Plasmid Transfer in the GI Tract ® and PCR (using universal [27,28] and genus-specific primers [29] to screen for bacterial DNA encoding 16S rRNAs prior to initiating experiments). Mice were fed an autoclaved standard diet (Purina LabDiet 5K67). Individual mouse identity was tracked using ear punches and cage number.
Mouse inoculation for in vivo transfer experiments. The germfree mice described above were used to examine pCF10 transfer in the intestinal tract; for the experiments described in this paper, recipients were first introduced, followed by donors (Fig. 1B). For experiment 1, mice received 4.3 ϫ 10 7 CFU of wild-type recipient OG1ES/mouse on day Ϫ3 and 2.9 ϫ 10 7 CFU of mutant CFϪ recipient (OG1RF background)/mouse on day 4 (Fig. 1B). For experiment 2, mice received 6.3 ϫ 10 7 CFU of mutant CFϪ recipient (OG1RF background)/mouse on day Ϫ3 and 1.0 ϫ 10 7 CFU of wild-type recipient OG1ES/mouse on day 4 (Fig. 1B). For experiment 3, mice received 4.3 ϫ 10 7 CFU of wild-type recipient OG1ES/mouse on day Ϫ4, and the defined community as competing GI microflora was gavaged on day Ϫ3 (Fig. 1B). For the defined community gavage, immediately after thawing the stocks, 350 l of each stock was mixed, and 100 l of the mixture was gavaged into the mice within 30 min. Mice received oral gavage with 2.0 ϫ 10 8 to 4.4 ϫ 10 8 CFU of E. faecalis plasmid donors/mouse on day 0 for all experiments (Fig. 1B).
Enumeration of donors, recipients, and transconjugants. Fourteen mice were used for each of three experiments; donors, recipients, and transconjugants were enumerated from fecal samples on days 1, 4, and 7 and from GI tissue (an upper section, a middle section, and a lower section) at 5 h (n ϭ 2) and at 1 (n ϭ 4), 4 (n ϭ 4), and 7 (n ϭ 4) days by selective plate quantitative culture. The upper GI section includes the duodenum plus the proximal jejunum, the middle GI section includes the distal jejunum and the ileum, and the lower GI section consists of the colon. Fecal samples were collected from individual mice and placed in 1 ml sterile saline, homogenized, quantitatively cultured, and reported as log 10 CFU/ milliliter. The GI tissue sections were aseptically dissected, and the fecal contents were removed, sliced open longitudinally, weighed, and placed into 1 ml sterile saline. They were vortexed for 30 s, sonicated for 5 min (frequency, 40 Ϯ 2 kHz; power density, 0.22 Ϯ 0.04 W/cm 2 ; Zenith Ultrasonic Inc., Norwood, NJ), and vortexed an additional 30 s. The sonicate fluid was quantitatively cultured, and log 10 CFU/gram of tissue was calculated. The frequency of plasmid transfer is reported as transconjugant/donor (T/D) ratio.
In vitro plasmid transfer. Recipients were inoculated 1:10 and donors were inoculated 1:100 in 2 ml of fresh Todd-Hewitt broth (Becton, Dickinson and Company, Sparks, MD). Plasmid transfer was assessed after 4 h coculture of donor and recipients by enumeration of donors and transconjugants by plating on selective plates as described above.
Characterization of the microbial community. The defined microbial community used was characterized by partial 16S rRNA gene sequencing of the communities found on days 1, 4, and 7 in mouse feces and also in an aliquot of the original gavage input. Fecal DNA was isolated by previously described methods (30,31). Library generation and sequencing were performed at the University of Minnesota Genomics Center. The V4 region of 16S rRNA gene was amplified, and 100,000 reads/sample were performed on an Illumina MiSeq platform with two 300-bp paired-end reads.
Statistical analysis. Analysis was performed by the in-program suite of the Prism7 software package (GraphPad Software, Inc.).