Adaptation in a Mouse Colony Monoassociated with Escherichia coli K ...

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, July 2010, p. 4655–4663 0099-2240/10/$12.00 doi:10.1128/AEM.00358-10 Copyright © 2010, American Society for Microbiology. All Rights Reserved.

Vol. 76, No. 14

Adaptation in a Mouse Colony Monoassociated with Escherichia coli K-12 for More than 1,000 Days䌤 Sean M. Lee,1 Aaron Wyse,1 Aaron Lesher,1 Mary Lou Everett,1 Linda Lou,1 Zoie E. Holzknecht,1 John F. Whitesides,2 Patricia A. Spears,3 Dawn E. Bowles,1 Shu S. Lin,1,4 Susan L. Tonkonogy,3 Paul E. Orndorff,3 R. Randal Bollinger,1,4 and William Parker1* Departments of Surgery1 and Immunology4 and Duke Human Vaccine Institute Flow Cytometry Core Facility,2 Duke University Medical Center, Durham, North Carolina 27710, and Department of Population Health and Pathobiology, College of Veterinary Medicine, North Carolina State University, Raleigh, North Carolina 276063 Received 9 February 2010/Accepted 9 May 2010

Although mice associated with a single bacterial species have been used to provide a simple model for analysis of host-bacteria relationships, bacteria have been shown to display adaptability when grown in a variety of novel environments. In this study, changes associated with the host-bacterium relationship in mice monoassociated with Escherichia coli K-12 over a period of 1,031 days were evaluated. After 80 days, phenotypic diversification of E. coli was observed, with the colonizing bacteria having a broader distribution of growth rates in the laboratory than the parent E. coli. After 1,031 days, which included three generations of mice and an estimated 20,000 generations of E. coli, the initially homogeneous bacteria colonizing the mice had evolved to have widely different growth rates on agar, a potential decrease in tendency for spontaneous lysis in vivo, and an increased tendency for spontaneous lysis in vitro. Importantly, mice at the end of the experiment were colonized at an average density of bacteria that was more than 3-fold greater than mice colonized on day 80. Evaluation of selected isolates on day 1,031 revealed unique restriction endonuclease patterns and differences between isolates in expression of more than 10% of the proteins identified by two-dimensional electrophoresis, suggesting complex changes underlying the evolution of diversity during the experiment. These results suggest that monoassociated mice might be used as a tool for characterizing niches occupied by the intestinal flora and potentially as a method of targeting the evolution of bacteria for applications in biotechnology. ability in global regulation, surface properties, and nutrient permeability in their homogeneous experimental evolution culture (14). Other experimental setups have utilized the fact that effective adaptive radiation depends on niche diversity (13) and also utilized a heterogeneous experimental evolution culture in an effort to drive the evolution of substantially more diversity than would a homogeneous culture. For example, Rainey and Travisano (19) showed that Pseudomonas fluorescens grown in unstirred liquid cultures for 7 days led to the evolution of distinct clones adapted to grow either at the bottom of the culture, in suspension, or at the top of the culture. This diversity could be reversed, in part, by returning the bacteria to a homogeneous (stirred) liquid culture, confirming the effectiveness of a heterogenous environment for the evolution of diversity in bacterial populations. With this in mind, it seems likely that introduction of bacteria into the large array of niches associated with previously germfree mice will lead to extensive adaptation of those bacteria. In the initial use of monoassociated mice to study the evolution of bacteria, Giraud et al. probed the benefits and costs of high mutation rates in evolving bacteria by inoculating previously germfree mice with E. coli that carried either a high or a low mutation rate (8). Their study suggested that E. coli with a rapid mutation rate evolved more rapidly than did E. coli with a lower mutation rate during the first 42 days of evolution but that the slowly mutating E. coli “caught up” with the rapidly mutating E. coli after 400 days (about 7,500 generations of E. coli) in the mouse gut. In a second study, Giraud and colleagues identified some of the key mutations that occur in

Mice associated with either a single bacterial species or a limited number of bacterial species have been widely used for biomedical research (9, 24, 25, 27), providing a simplified model for host-bacteria relationships in both health and disease. However, numerous studies have demonstrated the ability of bacteria to rapidly adapt to particular laboratory conditions (5), and recent work (7, 8) has probed changes in bacteria that occur within days of their introduction into previously germfree mice. Although this observation increases the complexity of the monoassociated mouse model, it also introduces a wide range of potential uses of monoassociated mice for the study of evolutionary biology and the directed evolution of bacteria for a variety of purposes. The highly adaptive nature of bacteria is often observed in liquid cultures, in which bacteria compete for a limited amount of glucose or another metabolite(s) for weeks, months, or even years (6, 22). Such experiments generally produce extensive evolutionary changes in the bacterial population, and a degree of diversity is sometimes observed despite the homogeneous nature of the culture (13). For example, Manche´ et al. (15) as well as Maharjan et al. (14) observed several clones coexisting in a stable fashion in homogeneous experimental evolution cultures by the end of their experiments (14, 15). Further, Maharjan et al. found stable bacterial populations with vari* Corresponding author. Mailing address: Department of Surgery, Duke University Medical Center, Box 2605, Durham, NC 27710. Phone: (919) 681-3886. Fax: (919) 684-7263. E-mail: bparker@duke .edu. 䌤 Published ahead of print on 14 May 2010. 4655

4656

LEE ET AL.

E. coli within the first week of evolution in the mouse gut (7). However, the extent of diversity that might be expected following extended periods of evolution in the mouse gut has not been evaluated, and the changes in the colonization of the mice were not evaluated. In this study, we initially set out to evaluate the effect of type 1 pilus expression by E. coli on the colonization of monoassociated mice. With that in mind, two germfree isolators were each inoculated, one with E. coli not expressing the type 1 pilus and the other with E. coli constitutively expressing the type 1 pilus. However, changes in the phenotype of the colonizing microbes were apparent within a few weeks, including the loss of the type 1 pilus from mice inoculated with bacteria that initially expressed that factor. Given this observation, the study was extended to probe some of the potential adaptations that the microbes might make, given time, in response to the highly competitive environment of the mouse gut. Here we report some initial measures of the diversity obtained during the adaptive radiation of E. coli during approximately 20,000 generations in monoassociated mice and evaluate changes in the density of colonization. MATERIALS AND METHODS Study design. The mice in two isolators, each containing three breeding pairs of germfree (sterile) mice, were each inoculated with a variant of E. coli K-12 strain MG1655 (2). Mice (strain 129S6/SvEv) were bred and removed from the colony as needed to maintain 20 to 50 mice per isolator for the duration of the experiment. An average of three breeding pairs of mice was maintained in each of the isolators until day 457, at which time males were separated from females and no further breeding was allowed to occur. Because it was hypothesized that the type 1 pilus might affect bacterial colonization, one isolator (isolator 1) was inoculated on day zero with a mutant lacking the ability to express type 1 pili (3) and the second isolator (isolator 2) was inoculated on day 35 with a mutant constitutively expressing type 1 pili (18). However, within 79 days of inoculating isolator 2 (day 114 of the experiment), no examples of piliated isolates were observed as determined by the lack of observable antigen in an immunoassay and as determined by the loss of the ability to undergo pilus-dependent, IgA-mediated aggregation (18). Thus, the offspring of the mutant used to inoculate isolator 2 apparently lost pilus expression during the first 79 days of the experiment. The experiment was carried out as described above using the two separate isolators until day 457, at which time a longer-term study was initiated using only one isolator and starting with the bacteria colonizing the mice on day 457. For that purpose, in lieu of arbitrarily eliminating one-half of the experiment, the two isolators were combined by moving the mice in isolator 2 into isolator 1, using appropriate care to avoid accidental microbial contamination. To effectively combine the isolators, all mice from both isolators were inoculated using oral and rectal swabs with fecal bacteria obtained from mice housed in the other isolator. This longer-term study was extended from day 457 until day 1,031, approximately 2 years and 10 months after the initiation of the experiment on day 0. Three generations of mice were housed in the isolators during the course of the experiment. The first generation of mice housed in the isolators was born 69 days (isolator 1) or 88 days (isolator 2) prior to the initiation of the experiment. In both isolators, the first mice in the second generation were born 27 days after initiation of the experiment (day 27 of the experiment in isolator 1 and day 62 of the experiment in isolator 2). The last of the mice bred in the isolators were third-generation mice born on day 464 of the experiment, 7 days after the two isolators were combined on day 457. Procurement of bacterial samples from monoassociated mice. To provide insight into the evolution of function during the experiment, changes in colonization density were evaluated as a function of time. For this purpose, the concentration of bacteria in the cecum of the mice was utilized as a measure of colonization density, and the concentration of bacteria present in the ceca of mice 80 days (isolator 1) and 79 days (isolator 2) after inoculation of their respective isolators were compared with the concentration of bacteria present in the ceca of mice at the end of the experiment on day 1,031. To obtain cecal contents, mice were removed from the isolator, immediately euthanized by CO2

APPL. ENVIRON. MICROBIOL. inhalation, and dissected with sterile surgical instruments. Samples were collected in air-tight containers, flash frozen in liquid nitrogen, and stored at ⫺145°C until use. Colonization levels were evaluated only in mice born and weaned after initial introduction of bacteria into the isolators. To evaluate the potential effect of a freeze-thaw cycle on the results, we performed the flow cytometric analyses (see below) after one freeze-thaw cycle and after three freeze-thaw cycles. The relative results obtained were the same, indicating that the approach used was not sensitive to freeze-thaw cycles. Measurement of colony size on agar as a relative indicator of growth rate. The size of bacterial colonies after growth for approximately 17 h at 37°C on agar was used as a relative indicator of the growth rate of bacteria. This approach made possible high-throughput screening of large populations of bacteria, and it was confirmed that colony size on agar plates of representative bacteria correlated with actual growth rates measured at 37°C in liquid medium (minimal essential medium with 1.0 mM HEPES, 0.1 mM nonessential amino acids, and 1 mM sodium pyruvate). However, the colony size formed by a given strain (e.g., a single culture separated, aliquoted, and flash frozen) varied depending on the agar preparation, perhaps due in part to differences in the dryness of the plates after cooling. With that in mind, internal standards were run for each experiment. For evaluation of colony size, samples of bacteria grown for approximately 17 h at 37°C in liquid medium (described above), or cecal contents were appropriately diluted in a phosphate-buffered saline solution and spread on 10-cm plates containing approximately 25 ml of agar (15.2 g/liter agar with 25.2 g/liter low-salt LB broth medium, pH 7.5; obtained from Teknova, Hollister, CA). After growth for approximately 17 h, plates were photographed and the photographs were printed so that the image diameter was 15 to 16 cm. The sizes of the colonies on the photos were then measured using a 5⫻ Viewcraft Lupe (Dot Line Corp, Chatsworth, CA), and the colony sizes were calculated by multiplying the measurements by the magnification factor (size of actual plate/size of photo). Restriction fragment length polymorphism and pulsed-field gel electrophoresis. Bacterial samples were submitted to the Laboratory of Molecular Typing at Cornell University for assessment of restriction fragment length polymorphisms. The Centers for Disease Control and Prevention (CDC) PulseNet protocol was used to perform pulsed-field gel electrophoresis (PFGE) for the isolates, and the program BioNumerics was used to analyze the results. 2D-DIGE. Two-dimensional differential in-gel electrophoresis (2D-DIGE) was run on 20- by 24-cm (second-dimension) gels by the Systems-Proteomics Center at the University of North Carolina, Chapel Hill. Cy2 was used as an internal control for all gels, with Cy3 and Cy5 used to label the isolates in three independent experiments. The gels were scanned using a Typhoon 9400 imager and analyzed using the differential in-gel analysis component of the DeCyder software. For each of three experiments, an independent culture of each bacterium was grown and assessed. All spots identified using the software were categorized as either protein peaks or artifacts (e.g., dust, scratches, edges, etc.) by a visual inspection by the core facility personnel. The data were assessed with a one-way analysis of variance using the DeCyder software. Evaluation of cecal contents and bacterial cell lysis by flow cytometry. To quantify the bacteria and small fiber particles in cecal contents of mice, samples of cecal contents were thawed and diluted to a concentration of 10 mg/ml in phosphate-buffered saline with 10 mg/ml bovine serum albumin and 0.2 g/liter NaN3. Large fiber particles were allowed to settle, and 15 ␮l of sample was suspended in an additional 3 ml of the same buffer (phosphate-buffered saline with 10 mg/ml bovine serum albumin and 0.2 g/liter NaN3). Any remaining large particles were then removed from the suspension using a 35-␮m strainer (BD Biosciences, San Jose, CA) to avoid clogging of the 70-␮m nozzle on the flow cytometer. Bacteria were cultured and lysed according to the following protocols. A 10- to 20-mg sample of cecal contents or 20 ␮l of cultured bacteria (the same strains used to inoculate mice in the isolators) was cultured at 37°C in 10 ml of liquid medium (minimal essential medium with 1.0 mM HEPES, 0.1 mM nonessential amino acids, and 1 mM sodium pyruvate) with constant shaking for 17 h. After this initial incubation, samples were used for flow cytometry or lysed using the following methods. (i) Some samples were centrifuged at 5,000 ⫻ g for 12 min, and the pellets were then submitted to chemical lysis using 200 ␮l B-PER reagent (Thermo Scientific, Rockford, IL) for 15 min at 22°C with vigorous mixing. (ii) Other samples were repeatedly placed at ⫺80°C until frozen and then thawed at 22°C until five freeze-thaw cycles had been completed. (iii) Other samples were incubated at 4°C for 14 days and then at 22°C for an additional 29 days with no change in medium (without addition of nutrition for the 43-day time period). To evaluate the autolysis of bacterial cells over the course of time, a 10- to 20-mg sample of cecal contents or 20 ␮l of cultured bacteria (the same strains used to inoculate mice in the isolators) was cultured at 37°C in 10 ml of liquid

VOL. 76, 2010

MICROBIAL ADAPTATION IN MONOASSOCIATED MICE

4657

FIG. 1. Variation in colony size on agar plates after approximately 17 h of growth at 37°C as described in Materials and Methods. The colony sizes on agar plates correlated with growth rates in liquid glucose-based medium and were taken as a convenient measure of the relative growth rate in vitro. (A) Colony sizes of E. coli K-12 cells used to inoculate mice in isolator 2 were compared with the relative colony sizes of bacteria isolated from the cecum of mice in isolator 2 at various times postinoculation. Samples from three mice at each time point were evaluated, and all data for a given time point are plotted together. (B) Bacteria from mice in isolator 1 at day 457 of the experiment were grown on agar plates, followed by selection of individual colonies of various sizes (isolates A through F), which were in turn replated on agar. Each isolate maintained the growth rate that was initially observed. The bar indicates the average relative colony size of each isolate following replating. In this particular experiment, the size of bacterial colonies formed by the strain used to inoculate isolator 1 (day zero) was 3.65 mm ⫾ 0.04 mm (mean ⫾ standard error).

medium as described above. Next, 50 ␮l of this initial 17-h culture was transferred to fresh liquid medium and incubated at 37°C for an additional 24 or 72 h. This second transfer was performed to dilute soluble factors in the cecal contents which could affect cell lysis over time (e.g., mucin, particular enzymes, etc.). All samples were diluted in buffer and filtered for large particles by using a 35-␮m strainer prior to evaluation by flow cytometry. The bacteria counting kit (Invitrogen Corp., Carlsbad, CA) was utilized to quantify the bacteria. Samples were divided three ways and assessed by flow cytometry as either unstained, stained, or stained with counting beads (6-␮m beads provided with the kit). Filtered buffer only (background), buffer with counting beads only, and cultured bacteria were used as controls. Staining using the SYTO BC bacteria stain was assessed using the fluorescein isothiocyanate (FITC; 515BP/20) channel. For staining, 0.5 ␮l SYTO BC and 2.5 ␮l beads were added per 500-␮l sample. Beads were sonicated as directed before use. The suspension was allowed to incubate for 5 min or longer in the dark before assessment by flow cytometry. Flow cytometry was performed in the Duke University Center for AIDS Research (CFAR) Flow Cytometry Core Facility, an NIH-funded program (P30 AI 64518). The machine used for flow cytometry was a BD FACSAriaII (FACSDiva 6.1.1), and all parameters were collected on a log scale with a time duration of 60 s. Calibration was performed with quality control particles of 0.3 ␮m, 0.5 ␮m, 0.8 ␮m, and 1.0 ␮m diameters, which were a generous gift of Jeff Ware and Larry Duckett at BD Biosciences (San Jose, CA). Data were assessed using FlowJo 8.6.3. The flow cytometric analysis was not 100% efficient in the detection of bacteria, and the efficiency varied depending on the settings and the tuning of the instrument on a given day. Thus, this approach probably underestimates levels of bacteria in the cecal samples, although the relative values were determined to be repeatable on different days using different settings of the instrument. In other words, although the absolute concentration of bacteria may have been underestimated on a given day, the relative concentrations determined on that day were very reliable. The concentration of bacteria in each mouse was assessed separately, with the exception of four mice taken from isolator 2, which were combined into two samples, each containing the cecal contents from two mice.

RESULTS Diversification of E. coli colonizing the monoassociated mouse colony measured by the increasing diversity of colony sizes on agar. Over the course of the experiment, flora colonizing all mice were found to have evolved diverse growth rates on agar as indicated by variations in the sizes of colonies

formed after approximately 17 h of growth at 37°C on agar (Fig. 1). For example, bacteria colonizing the mice in isolator 2 at day 114 of the experiment demonstrated a marked variation of colony sizes on agar (Fig. 1A) compared to the isolate used initially to colonize the mice. Within an additional 64 days, on day 178 of the experiment, this diversification in isolator 2 had increased markedly, forming a bimodal distribution of colony sizes formed on agar (Fig. 1A). Importantly, evaluation of the colony sizes formed on agar by bacteria colonizing mice in isolator 1 also revealed substantial diversification (Fig. 1B), indicating that the diversification was reproducible in the two independent experiments. Replating of both large-colonyforming isolates and small-colony-forming isolates in the laboratory revealed that the changes that occurred in the isolators were stable in the laboratory (Fig. 1B). Further, culture of a small-colony-forming isolate, which grew severalfold slower in liquid culture than large-colony-forming isolates in up to 750 ml of liquid culture, never produced a reversion with faster growth, despite selection pressure that would presumably favor such a reversion. Finally, the diversity of growth on agar was maintained in the bacteria associated with the mouse colony, with both fast-growing and slow-growing isolates present at day 1,031, the end of the experiment. Assessment of bacterial cell lysis by flow cytometry. While evaluating the bacteria present in the ceca of mice during the course of the experiment, it became apparent that particles with a size comparable to that of intact E. coli cells (between about 0.5 ␮m and 0.8 ␮m), as well as particles substantially smaller than typical E. coli cells (ⱕ0.5 ␮m), were bound by fluorescent dyes specific for bacterial cells (see below). To assess the possibility that the small particles represented lysed bacterial cells, a series of experiments was conducted with the cultured E. coli strains that were used to inoculate the isolators. As shown in Fig. 2, the lysis of E. coli cells using either a lysis buffer (Fig. 2B to D) or repeated freeze-thaw cycles (Fig.

4658

LEE ET AL.

APPL. ENVIRON. MICROBIOL.

FIG. 2. Assessment of bacterial cell lysis by flow cytometry. (A to C) Light scattering determined with size calibration beads is shown (A). In addition, the fluorescence versus forward light scattering is shown for cultured bacteria used to inoculate mice in isolator 1 either before (B) or after (C) lysis using a commercially available lytic solution. (D to F) Histogram showing forward light scattering of the cultured bacteria before (solid line) and after (dotted line) lysis with lytic solution (D), a series of five freeze-thaw cycles (E), or aging for over a month in the laboratory without nutrition (F). Protocols for culture and lysis of the bacteria are described in Materials and Methods, and similar results were obtained for cultured bacteria used to inoculate mice in isolator 2 (data not shown).

2E) resulted in the loss of particles with a size consistent with intact bacterial cells and an increase in particles that were smaller than expected for intact E. coli cells. Similar results were obtained with heat treatment (70°C for 10 min) of E. coli (data not shown). Allowing the bacteria to sit for 43 days (14 days at 4°C plus 29 days at about 22°C) without addition of nutrients also resulted in a loss of particles with a size consistent with intact cells and an increase in particles smaller than intact cells (Fig. 2F). These data indicate that lysis of E. coli can be monitored by flow cytometry, with the loss of intact cells corroborated by the appearance of “cell debris,” although the exact nature of the cell debris is unknown. Evaluation of sorted samples by fluorescence light microscopy confirmed the presence of intact cells where expected, but the smaller particles were more difficult to conclusively identify using the fluorescence microscope, suggesting that further work using an electron microscope may be required to probe the nature of the smaller particles. Changes in colonization density. To provide insight into potential changes in the interaction between the bacteria and host during the course of the experiment, the colonization densities of bacteria present in the ceca of mice 79 and 80 days after their inoculation in the two isolators were compared with the colonization density of bacteria present in the ceca of mice at the end of the experiment on day 1,031. In addition, the colonization density of bacteria present in the ceca of mice maintained under specific-pathogen-free (SPF) conditions was assessed, providing an example of the characteristics of a typical, diverse, and well-adapted microbial flora for the sake of

comparison. Presumably this microbial population represents the “optimal” colonization level, since it has coevolved with the host species over hundreds of millions of years. The number of bacterial cells per gram of dry cecal contents as determined by flow cytometry was taken as a measure of colonization density. As shown in Fig. 3, the colonization of the cecum by bacteria increased in a highly significant fashion when comparing results on day 80 in isolator 1 with day 1,031 results (P ⫽ 0.0001) and when comparing day 114 (79 days postinoculation) results in isolator 2 with day 1,031 results (P ⫽ 0.0001). This change corresponded to mean colonization levels 3.7-fold greater (average for all mice in both isolators) at the end of the experiment than were found 79 to 80 days after inoculation of the isolators. If the degree of colonization of SPF mice were taken as a level of “complete” colonization, the change observed in this experiment would entail an improvement from about 7% colonization early in the experiment to about 28% colonization at the end of the experiment. To what extent this change reflects evolution of the bacteria versus potential nongenetic factors associated with changes in mouse physiology during the course of the experiment will necessarily be the subject of future experiments, as elaborated upon in the Discussion section. However, the change in colonization density does indicate that profound changes in host-bacteria relationships do occur over the course of time of this experiment in monoassociated mice. Changes in susceptibility to cell lysis during the course of the experiment. Not only did the degree of cecal colonization change during the course of the experiment, but the relative

VOL. 76, 2010


4659

FIG. 3. Bacterial colonization in the ceca of mice early and at the end of the experiment on day 1,031. Forward light scattering as a measure of relative bacterial cell size was assessed in bacteria isolated from the ceca of mice early (n ⫽ 10 on day 80 in isolator 1 and n ⫽ 8 on day 114 in isolator 2) and at the end (n ⫽ 8) of the experiment. The results from SPF mice, fully colonized with a normal bacterial flora (n ⫽ 4) are shown for comparison. Bacteria were distinguished from other particulate matter in the cecal contents by staining with SYTO BC green nucleic acid stain, which was detected using the FITC channel. (A) Flow cytometry profiles of a blank sample containing no cecal contents, with regions corresponding to intact bacterial cells, cell debris (cd), and fiber labeled. The region just under the area marked cell debris is possibly a combination of cell debris and fiber and thus was not evaluated in the study. (B to D) Flow cytometry profiles of a representative sample of cecal contents from a mouse on day 80 in isolator 1 (B), a representative sample of cecal contents from a mouse on day 1,031 of the experiment (C), and a representative sample of cecal contents from an SPF mouse (D). (E to G) Each point in the scatter plots represents the result obtained using the cecal contents of one mouse, with the exception of two data points from isolator 2, each of which represents results obtained using the combined cecal contents from two mice. The lines indicate the means of the data at each time point. The bacteria/gram dry cecal weight (E), ratio of intact bacterial cells to cell debris (F), and microparticulate fiber/gram dry cecal weight (G) are shown. When comparing results from samples taken at day 80 or day 114 with those at day 1,031, the differences in bacteria concentration (E) and cell integrity (F) were highly significant (P ⬍ 0.0001) in all cases. The difference in average microparticulate fiber at day 80 was significantly (P ⫽ 0.0015) different that that observed at day 1,031, although the difference at day 114 and day 1,031 was not significant (P ⫽ 0.0752). A two-way unpaired t test was used to calculate P values.

intact bacteria/bacterial cell debris (observed as particles of ⱕ0.5 ␮m in size) ratios changed substantially during the course of the experiment (P ⬍ 0.0001) (Fig. 3F). The average ratio of bacterial cells to cell debris present early in the experiment was relatively low, with 1 bacterial cell for every 5.0 and 3.3 smaller particles in isolators 1 and 2, respectively. In contrast, on day 1,031, 1 bacterial cell was associated with only about 1.4 smaller particles, approaching values found in the ceca of SPF mice, which averaged slightly less than 1 smaller particle per intact cell. To evaluate the extent to which the bacteria may have evolved resistance to cell lysis during the course of the experiment, the spontaneous lysis of bacterial cells over time was evaluated, using the ratio of intact cells to cell debris as a measure of lysis. For this purpose, the bacterial strains used to inoculate mice in isolator 1 (day 0) and in isolator 2 (day 35), as well as the cecal contents from mice at day 1,031, were

grown in liquid culture as described in Materials and Methods. Bacteria cultured from the cecal contents of SPF mice were evaluated for comparison. As shown in Fig. 4, substantial spontaneous lysis of bacteria cultured from mice on day 1,031 was observed after 72 h at 37°C, but considerably less lysis was observed in the strains used to inoculate the isolators (P ⬍ 0.0001). For example, after 72 h at 37°C, each intact bacterial cell of the strain used to inoculate isolator 1 was associated with only about 0.13 smaller fragments, whereas each intact bacterial cell from mice on day 1,031 of the experiments was associated with 0.85 smaller fragments. Bacteria cultured from the SPF mice showed even greater spontaneous lysis, with 2.4 smaller fragments for every intact cell after 72 h at 37°C (Fig. 4C). Phenotypic and genotypic changes in selected isolates. To provide some initial view of the degree of variation that might be involved in the evolution of the bacteria, two isolates ob-

4660

LEE ET AL.


was a primary concern, and a number of results indicated that contamination of the isolators did not take place. For example, the gradual change in phenotype observed (Fig. 1A) strongly argued against contamination as the cause of diversity, and the diversity of phenotypes observed (Fig. 1B) was not consistent with rare contamination events as the cause of the diversity. The observation that similar changes occurred in both isolators (Fig. 1) also argues against rare contamination events as a cause of the observed diversity. Further, repeated cultures throughout the duration of the experiment failed to identify bacteria possessing a colony morphology distinct from that of E. coli. To further ensure that the observed results were not due to contamination, the 16S sequences of two isolates obtained on day 1,031 with widely different growth rates on agar (the same two isolates described above and characterized in Fig. 5) were compared to the sequence of the parent E. coli strain, MG1655. The sequences of bacteria harvested from the mice on day 1,031 were not different from strain MG1655, with the exception of a 1-bp variation that was also found in the type 1 pilus-containing construct used to inoculate isolator 2. Thus, despite substantial phenotypic and genotypic differences among the isolates, we found no evidence for the presence of bacteria other than those we initially introduced. FIG. 4. Lysis of cultured bacterial cells over time. The lysis of bacteria used to inoculate mice in isolator 1 (Iso 1) and isolator 2 (Iso 2), bacteria obtained from the cecum of mice at day 1,031 of the experiment, and bacteria obtained from SPF mice was assessed by flow cytometry. Cultured, frozen bacteria or cecal contents, where appropriate, were used to inoculate medium and allowed to grow for 17 h at 37°C, at which time 50 ␮l of liquid culture was used to inoculate an additional 10 ml of liquid medium, as described in Materials and Methods. The lysis of bacteria was then assessed by flow cytometry 24 h and 72 h after the second inoculation. (A and B) Representative histograms created from the forward scatter profile of bacteria used to inoculate mice in isolator 1 (A) and bacteria from the cecum of a mouse at day 1,031 (B). (C) Each point represents the result obtained using either one of the strains used to inoculate the isolators or the cecal contents of one mouse. The lines indicate the means of the data at each time point.

tained on day 1,031 of the experiment, one very fast-growing on agar and the other very slow-growing, were selected for study. The generation times of the fast-growing and the slowgrowing clones in liquid culture (as described in Materials and Methods) were approximately 1.1 h and 6.7 h, respectively. As shown in Fig. 5A, the proteomes of these two isolates were radically different, with 105 proteins out of 1,042 resolved proteins (10%) being expressed at significantly (P ⬍ 0.05) different levels between the two isolates. Further, as shown in Fig. 5B, assessment of fragment lengths following digestion of DNA with the restriction endonuclease XbaI revealed six points of difference (indicated by the red dots) between the slow-growing isolate and the fast-growing isolate, with both isolates being distinct from the strains used to inoculate the isolators. These findings indicate that the changes underpinning the evolution of morphological and functional diversity observed in this experiment are likely complex and are due to multiple, stable mutations in the DNA. Assessment of the integrity of the isolators. Avoiding contamination of the isolators during the course of the experiment

DISCUSSION A wide range of factors, including the diversity of unoccupied niches in the germfree mouse gut and the inability of a laboratory strain of E. coli to effectively utilize those niches, likely contributed to the evolution of bacteria observed in this study and in previous experiments (7, 8). Based on 75 min as the previously measured generation time of E. coli in monoassociated mice (20), we estimate that approximately 20,000 generations of E. coli elapsed during the course of the experiment. This number of generations, in combination with the very large number of bacteria present in each mouse (Fig. 3; from an average of about 8 ⫻ 107 bacteria/gram of dry cecal contents at the beginning of the experiment to about 35 ⫻ 107 bacteria/gram at the end of the experiment), is expected to facilitate adaptive radiation to a substantially greater extent than might be possible in a relatively homogeneous experimental evolution culture using limited amounts of simple sugars as the driving force for evolution. Although more complex than a homogeneous culture, this model has great potential as a tool to explore a wide range of medical and industrial applications. The mouse gut has been described as a complex natural habitat useful in experimental evolution studies (8), and it certainly appears to have driven a complex adaptive radiation based on the results presented herein. However, the changes in colonization density as assessed in this study may be influenced by factors other than evolution of the bacteria. For example, although colonization levels were evaluated only in mice born and weaned after initial introduction of bacteria into the isolators, it is possible that epigenetic factors passed down through the three generations of mice present in the experiment may have altered the ability of the E. coli cells to colonize the gut. Another factor that can affect colonization density is the age

VOL. 76, 2010


4661

FIG. 5. Comparison of the phenotypes and genotypes of a fast-growing isolate and a slow-growing isolate obtained on day 1,031 of the experiment. These two isolates were confirmed to have identical 16S sequences to the parent strain that was used to inoculate isolator 2. (A) 2D-DIGE analysis shows the proteomes of two isolates obtained from the mice at day 1,031. Proteins produced predominantly by the slow-growing isolate are indicated by red fluorescence, proteins produced by the fast-growing isolate are indicated by green fluorescence, and proteins produced in similar amounts by both isolates are indicated by yellow fluorescence. Isolates were grown to 26% to 41% of maximum density in glucose-rich medium (Dulbecco’s modified Eagle’s medium) before extraction of the protein and analysis, and the experiment was conducted in triplicate. (B) Assessment of restriction fragment length polymorphisms shows unique cleavage patterns (21) of XbaI for both isolates from day 1,031, which are distinct from the patterns observed in the clones used to inoculate the isolators. Bands or the absence of bands in the isolates that are distinct from each other and from the pattern observed in the clones used to inoculate the isolators are indicated by a red dot. The brightness and contrast of each lane were adjusted independently for clarity of illustration, although the raw data, without manual adjustment, were analyzed using BioNumerics to identify the unique bands indicated by the red dots.

of the mice. Given the experimental design, the mice evaluated at the end of the experiment were substantially older (561 to 710 days old, mean, 615 days old) than those mice present 79 to 80 days after inoculation of the isolators (52 to 53 days old). Although the age of the mice would presumably have no impact on any in vitro assessments of bacterial evolution, the aging of the mice might have an impact on bacterial colonization of the cecum. Previous studies (4) demonstrated that aging of mice can affect the concentration of bacteria in the ceca, although the concentration of bacteria in the ceca of mice was found to decrease by about a factor of 2 as mice of aged from 84 to 168 days (4). In addition, in the present study, no significant difference was observed (P ⫽ 0.59) between the colonization of mice aged 702 to 710 days (n ⫽ 3) at the end of the experiment and mice aged 561 days (n ⫽ 5) at the end of the experiment. Further, one of the oldest mice (702 days) at the end of the study was found to have the lowest colonization level of any mouse evaluated at the end of the experiment. Thus, it is expected that the observed increase in colonization density over time is not likely due to factors associated with the age of the mice.

The results suggest that the potential for spontaneous cell lysis of bacteria in the mouse gut may have diminished greatly during the 1,031 days of the experiment. The tendency for spontaneous cell lysis in the colon would likely affect the survival of bacteria in the colon, and changes in this phenotype are therefore consistent with the rate of evolution that might be expected to occur. However, during the course of the 1,031 days of the experiment, the E. coli clearly acquired an increased tendency for spontaneous lysis in liquid culture. This finding suggests that improved fitness in the intestinal environment occurs concomitantly with a decreased fitness under laboratory conditions. However, whether the changes affecting the potential gain of fitness in vivo are related to the loss of fitness in vitro remains unknown. To isolate the effects of bacterial evolution on colonization density, alternative approaches to assessing the colonization of mice in future experiments might be considered. For example, periodic introduction of germfree animals into ongoing experiments and evaluation of bacterial colonization in those animals should prove an effective means of assessing colonization efficiency of evolving microbial communities that would be

4662

LEE ET AL.

independent of changes in the animal hosts. However, the results presented herein are important, as they reflect the overall changes in host-bacteria interactions as influenced by genetic changes in the bacteria as well as epigenetic changes in the mice. The evolution observed during the 2 years and 10 months of this study encompasses diversification in metabolism and other properties of the bacterial community and may potentially lead to improved colonization of the mice. These findings indicate that animal housing facilities designed as barriers to prevent unintentional introduction of bacteria can be utilized as powerful tools for assessing the mechanisms that create and maintain biological diversity, a topic of great interest (10, 23). Such facilities, first developed in the 1950s, are currently operating at several institutions worldwide and thus offer access to this technology for a variety of investigators with different interests. Future studies will likely include a detailed assessment of the molecular mechanisms underlying the evolution of new traits during adaptive radiation and might include assessment of bacterial evolution on a variety of epithelial surfaces, such as the skin and the oral epithelium. Adaptive radiation of bacteria in previously germfree mice represents a potential use of experimental evolution as a tool for advancing medical science. Since the intestinal flora of all vertebrates share many features (12), and since mice and humans share an omnivorous diet, which strongly influences the flora (12), it seems likely that the mouse will serve as an excellent model for humans in this regard, as has been suggested (27). Further, the ability of particular E. coli strains to colonize the human gut is reflected in the efficiency by which those strains colonize the mouse gut (17). Thus, elucidating the details of adaptive radiation in previously germfree mice may provide substantial insight into the nature and diversity of niches normally occupied by the human intestinal flora. The understanding of that flora is of substantial medical importance and is currently the subject of intense focus in the scientific community (11). In addition to an improved understanding of the human flora, experimental evolution utilizing a monoassociated rodent model has the potential to provide technological and medical applications. For example, adaptive radiation in a previously germfree gut could be used to generate probiotics with an improved ability to persist in the bowel of humans. Probiotics, defined as live microorganisms which confer a health benefit on the host when administered in adequate amounts, have been used to treat a variety of medical problems, including ileal pouchitis, postoperative infections, certain diarrheal illnesses, and inflammatory bowel disease, among others. Further, probiotics are currently in widespread use for the maintenance of general health in individuals without disease (16) and also have substantial potential for use in veterinary medicine (26). While specific genetic modifications designed to increase the persistence of probiotics in the gut may be worthwhile (1), it seems likely that natural selection may provide a much more effective approach, both in terms of cost and the actual ability to improve persistence of probiotics in the bowel. Given the ability to genetically manipulate mice and, in addition, the wide range of diet which might be used to influence evolution of bacteria in the gut, the potential applications of this approach for both basic biology and biotechnology are


extensive and add substantially to the practical applications previously identified for experimental evolution. ACKNOWLEDGMENTS This work was supported in part by awards AI-51445 (Duke Human Vaccine Institute Flow Cytometry Core Facility) and P30 DK34987 (Center for Gastrointestinal Biology and Disease) from the National Institutes of Health. We thank Patrice McDermott for assistance with the flow cytometry and Donna Kronstadt for masterful maintenance of the gnotobiotic mice at the Gnotobiotic Animal Core, Center of Gastrointestinal Biology and Disease, College of Veterinary Medicine, North Carolina State University. REFERENCES 1. Barbas, A. S., A. P. Lesher, A. D. Thomas, A. Wyse, A. P. Devalapalli, Y.-H. Lee, H.-E. Tan, P. E. Orndorff, R. B. Bollinger, and W. Parker. 2009. Altering and assessing persistence of genetically modified E. coli MG1655 in the large bowel. Exp. Biol. Med. 234:1174–1185. 2. Blattner, F. R., G. Plunkett III, C. A. Bloch, N. T. Perna, V. Burland, M. Riley, J. Collado-Vides, J. D. Glasner, C. K. Rode, G. F. Mayhew, J. Gregor, N. W. Davis, H. A. Kirkpatrick, M. A. Goeden, D. J. Rose, B. Mau, and Y. Shao. 1997. The complete genome sequence of Escherichia coli K-12. Science 277:1453–1474. 3. Bloch, C. A., B. A. Stocker, and P. E. Orndorff. 1992. A key role for type 1 pili in enterobacterial communicability. Mol. Microbiol. 6:697–701. 4. Brennan-Craddock, W. E., A. K. Mallett, I. R. Rowland, and S. Neale. 1992. Developmental changes to gut microflora metabolism in mice. J. Appl. Bacteriol. 73:163–167. 5. Buckling, A., R. Craig Maclean, M. A. Brockhurst, and N. Colegrave. 2009. The Beagle in a bottle. Nature 457:824–829. 6. Cooper, T. F., D. E. Rozen, and R. E. Lenski. 2003. Parallel changes in gene expression after 20,000 generations of evolution in Escherichia coli. Proc. Natl. Acad. Sci. U. S. A. 100:1072–1077. 7. Giraud, A., S. Arous, M. De Paepe, V. Gaboriau-Routhiau, J. C. Bambou, S. Rakotobe, A. B. Lindner, F. Taddei, and N. Cerf-Bensussan. 2008. Dissecting the genetic components of adaptation of Escherichia coli to the mouse gut. PLoS Genet. 4:e2. 8. Giraud, A., I. Matic, O. Tenaillon, A. Clara, M. Radman, M. Fons, and F. Taddei. 2001. Costs and benefits of high mutation rates: adaptive evolution of bacteria in the mouse gut. Science 291:2606–2608. 9. Gordon, H. A., and L. Pesti. 1971. The gnotobiotic animal as a tool in the study of host microbial relationships. Bacteriol. Rev. 35:390–429. 10. Grunbaum, D. 2009. Ecology. Peter principle packs a peck of phytoplankton. Science 323:1022–1023. 11. Hattori, M., and T. D. Taylor. 2009. The human intestinal microbiome: a new frontier of human biology. DNA Res. 16:1–12. 12. Ley, R. E., M. Hamady, C. Lozupone, P. J. Turnbaugh, R. R. Ramey, J. S. Bircher, M. L. Schlegel, T. A. Tucker, M. D. Schrenzel, R. Knight, and J. I. Gordon. 2008. Evolution of mammals and their gut microbes. Science 320: 1647–1651. (Erratum, 322:1188.) 13. MacLean, R. C. 2005. Adaptive radiation in microbial microcosms. J. Evol. Biol. 18:1376–1386. 14. Maharjan, R., S. Seeto, L. Notley-McRobb, and T. Ferenci. 2006. Clonal adaptive radiation in a constant environment. Science 313:514–517. 15. Manche´, K., L. Notley-McRobb, and T. Ferenci. 1999. Mutational adaptation of Escherichia coli to glucose limitation involves distinct evolutionary pathways in aerobic and oxygen-limited environments. Genetics 153:5–12. 16. Minocha, A. 2009. Probiotics for preventive health nutrition in clinical practice. Nutr. Clin. Pract. 24:227–241. 17. Myhal, M. L., D. C. Laux, and P. S. Cohen. 1982. Relative colonizing abilities of human fecal and K 12 strains of Escherichia coli in the large intestines of streptomycin-treated mice. Eur. J. Clin. Microbiol. 1:186–192. 18. Orndorff, P. E., A. Devapali, S. Palestrant, A. Wyse, M. L. Everett, R. B. Bollinger, and W. Parker. 2004. Immunoglobulin-mediated agglutination and biofilm formation by Escherichia coli K-12 requires the type 1 pilus fiber. Infect. Immun. 72:1929–1938. 19. Rainey, P. B., and M. Travisano. 1998. Adaptive radiation in a heterogeneous environment. Nature 394:69–72. 20. Rang, C. U., T. R. Licht, T. Midtvedt, P. L. Conway, L. Chao, K. A. Krogfelt, P. S. Cohen, and S. Molin. 1999. Estimation of growth rates of Escherichia coli BJ4 in streptomycin-treated and previously germfree mice by in situ rRNA hybridization. Clin. Diagn. Lab. Immunol. 6:434–436. 21. Ribot, E. M., M. A. Fair, R. Gautom, D. N. Cameron, S. B. Hunter, B. Swaminathan, and T. J. Barrett. 2006. Standardization of pulsed-field gel electrophoresis protocols for the subtyping of Escherichia coli O157:H7, Salmonella, and Shigella for PulseNet. Foodborne Pathog. Dis. 3:59–67.

VOL. 76, 2010


22. Rozen, D. E., and R. E. Lenski. 2000. Long-term experimental evolution in Escherichia coli. VIII. Dynamics of a balanced polymorphism. Am. Nat. 155:24–35. 23. Schluter, D. 2000. The ecology of adaptive radiation. Oxford University Press, Oxford, United Kingdom. 24. Umesaki, Y., and H. Setoyama. 2000. Structure of the intestinal flora responsible for development of the gut immune system in a rodent model. Microbes Infect. 2:1343–1351.

4663

25. Wilson, K. H., J. N. Sheagren, R. Freter, L. Weatherbee, and D. Lyerly. 1986. Gnotobiotic models for study of the microbial ecology of Clostridium difficile and Escherichia coli. J. Infect. Dis. 153:547–551. 26. Wynn, S. G. 2009. Probiotics in veterinary practice. J. Am. Vet. Med. Assoc. 234:606–613. 27. Zaneveld, J., P. J. Turnbaugh, C. Lozupone, R. E. Ley, M. Hamady, J. I. Gordon, and R. Knight. 2008. Host-bacterial coevolution and the search for new drug targets. Curr. Opin. Chem. Biol. 12:109–114.