Intestinal Bile Acids Induce a Morphotype Switch in Vancomycin Resistant Enterococcus that Facilitates Intestinal Colonization

Summary

Vancomycin resistant Enterococcus (VRE) is a highly antibiotic-resistant and readily transmissible pathogen that causes severe infections in hospitalized patients. We discovered that lithocholic acid (LCA), a secondary bile acid prevalent in the cecum and colon of mice and humans, impairs separation of growing VRE diplococci causing the formation of long chains and increased biofilm formation. Divalent cations reversed this LCA-induced switch to chaining and biofilm formation. Experimental evolution in the presence of LCA yielded mutations in the essential two-component kinase yycG/walK and three-component response regulator liaR that locked VRE in diplococcal mode, impaired biofilm formation and increased susceptibility to the antibiotic, daptomycin. These mutant VRE strains were deficient in host colonization due to their inability to compete with intestinal microbiota. This morphotype switch presents a potential non-bactericidal therapeutic target that may help clear VRE from the intestines of dominated patients, as occurs frequently during hematopoietic stem cell transplantation.

Graphical Abstract

eTOC Blurb

Enterococci are common, increasingly antibiotic resistant gut microbes that grow as diplococci in liquid media. McKenney et al. describe a morphotype switch to chained growth driven by bile acids and reversed by cations that was necessary for persistence in the intestine by vancomycin-resistant Enterococcus faecium.

Introduction

Single bacterial cells exist in a range of sizes from near the 200nm limit of resolution of light microscopy (Luef et al., 2015), up to the hundreds of microns resolvable by the human eye (Schulz et al., 1999). Bacterial cell size, however, is not static and it varies depending on conditions of the local environment (Schaechter et al., 1958; Weart et al., 2007). Some bacteria switch between distinct growth modes called morphotypes that alter cell size and shape in a predictable and regulated manner. Transitions between morphotypes are important characters in bacterial evolution, pathogenesis and resistance to antibiotics (Deforet et al., 2015; Justice et al., 2008; Kysela et al., 2016; Weiser, 2013).

An example of morphotype switches in Gram-positive bacteria occurs in liquid cultures of Bacillus subtilis laboratory strains, which grow in long chains of fully septated cells during exponential growth. When the cultures approach stationary phase, they express high levels of autolysin which breaks the chains apart and separates the individual cells to prepare for sporulation (Margot et al., 1999). Streptococcus pneumonia performs the reverse switch–diplococci to chains–when it encounters host-associated media, such as nasal lavage fluid (Dalia and Weiser, 2011). The long chains adhere better to surfaces, which aids S. pneumoniae in colonization of the host (Rodriguez et al., 2012). Interestingly, chaining also increased the likelihood of cells being bound by complement and engulfed by macrophages, suggesting that morphotypes may confer context-specific advantages during host interactions(Dalia and Weiser, 2011).

Any bacterium that colonizes the intestine encounters bile acids, which are rarely found elsewhere in nature (Hofmann et al., 2010). Bile acids are present throughout the intestine in concentrations ranging from low millimolar to low micromolar and they affect gut microbes since whole bile is anti-microbial at high concentrations (Begley et al., 2005). Bile is produced in the liver and is chemically altered by the intestinal microbiota into so-called secondary bile acids (Ridlon et al., 2016). Dozens of chemically distinct bile acids have been identified and each one is likely to have a unique spectrum of chemical and biological activities. The universal distribution of enterococci among coelomates (Lebreton et al., 2014) suggests that their association with vertebrate hosts is ancient. Thus, VRE may possess mechanisms to respond to bile acids and facilitate to host colonization. Enterococci mount a stress response to high levels of bile (Bøhle et al., 2010; Choudhury et al., 2011; Michaux et al., 2011; Saito et al., 2014; Solheim et al., 2007; Zhang et al., 2013), but little else is known about the effects of individual bile acids.

Here we identified a reversible morphotype switch in VRE that is triggered by bile acids, reversed by cations and plays a key role in intestinal colonization. We used experimental evolution to select for loss of the morphotype switch and obtained mutants with reduced biofilm formation and deficient intestinal colonization in the presence of a competing gut microbiota. Unexpectedly, these strains were highly susceptible to the antibiotic daptomycin due to the accumulation of mutations in genes associated with resistance. This morphotype switch represents a potential target for decolonization therapies for VRE and illustrates the linkage between phenotypes selected for during the evolution of host-associated microbes and antibiotic resistance.

Results

Secondary bile acids cause chaining in VRE

To determine how individual bile acids may affect the growth of enterococci in the intestine, we grew a Vancomycin-resistant strain of Enterococcus faecium ATCC 700221 (VRE) in liquid media in the presence of physiologically-relevant concentrations of single bile acids. Primary bile acids such as cholic acid (CA) had little effect on growth. The secondary bile acids deoxycholic acid (DCA) and lithocholic acid (LCA) caused a dose-dependent reduction of the growth rate of VRE (Figure S1A–K). The final optical density of the culture was significantly, but slightly, decreased at physiologically relevant concentrations of 125 μM of LCA (Devlin and Fischbach, 2015; Hamilton et al., 2007), but not (Figure S1C–D), but was unaffected by a diverse panel of primary and secondary bile acids tested mostly at 250 μM. After overnight culture, we noticed a visually striking precipitation of VRE treated with LCA. Light microscopy revealed that this was due to a switch from diplococci to long chains (Figure 1A). Growth in the presence of 250 μM LCA resulted in a 2-log reduction in VRE CFUs (Figure S1E). This reduction in CFUs was eliminated when cultures were grown in the presence of DNAse, which breaks cell chains of E. faecium (Paganelli et al., 2013), suggesting that LCA is not toxic at this concentration.

Figure 1. Lithocholic acid induced chained growth and biofilm in VRE.

A) Light microsocopy and electron microscopy of VRE grown overnight without and with 125 μM LCA. B) Quantification of chain length of VRE cultures grown overnight in the presence of 125 μM of the indicated primary bile acids: CA = Cholic acid, CDCA = Chenodeoxycholic acid, and secondary bile acids: DCA = Deoxycholic acid, LCA = Lithocholic acid. (n ~200–350 chains per condition, Kruskal-Wallace one-way ANOVA with Dunn’s correction vs. DMSO) C) Fluorescence microscopy of VRE grown for 4 hours in the presence or absence of LCA and stained for 20 minutes with HADA before fixing and membrane staining with Mitotracker Red. D) VRE biofilm formation on polystyrene plates in 125 μM or 250 μM of the indicated bile acids, following growth for 36 hours, performed 3 times in triplicate (Kruskal-Wallace one-way ANOVA with Dunn’s correction vs. DMSO). E) E. faecium E1162 biofilm formation in the presence and absence of 125 μM of LCA, following growth for 36 hours, performed 3 times in triplicate (Unpaired t-test with Holm-Sidak correction vs. DMSO). * p < 0.05, ** p < 0.005, **** p < 0.0005

Transmission electron microscopy of LCA-treated cells showed complete division septa; however, the daughter cells failed to separate into the diplococci which were the most common cell morphology in normal growth media (Figure 1A). LCA-treated cells also generated horn-like structures of cell-envelope material (Figure S2), which often appear to flank nascent mid-cell division sites. These structures resembled those reported in daptomycin resistant isolates of Enterococcus faecalis (Arias et al., 2011). We quantified the chain length of VRE after 24 hours in the presence of the primary bile acids cholic acid and chenodeoxycholic acid and secondary bile acids DCA and LCA. LCA was the strongest inducer of cell chaining at physiologically-relevant concentrations with a mean chain length of 5 and a wide range that exceeded 20 cells (Figure 1B). This is likely an underestimate of the true mean chain length of the population, as longer chains often fail to separate from each other and often do not adhere to the coverslip in a single plane of focus.

These data suggested that LCA treatment impacts cell separation without affecting cell division, and that LCA may interfere with peptidoglycan synthesis. To determine if peptidoglycan synthesis was properly localized we performed short pulse labeling of peptidoglycan synthesis by adding fluorescent-labeled d-Alanine probe HADA (Kuru et al., 2012) in the presence and absence of LCA. In untreated cells, the HADA incorporated at mid-coccus (Figure 1C), suggesting this was the location of the wall peptidoglycan synthesis machinery, which is consistent with other ovococci such as S. pneumoniae (Boersma et al., 2015; Kuru et al., 2012). The dominant pattern at mid-log phase was of connected diplococci with triplet bands of HADA perpendicular to the long axis of the cells. Peptidoglycan synthesis co-localized with the site of membrane fission during cell division, suggested by the presence of overlapping HADA and Mitotracker Red at mid-cell (Figure 1C). In untreated cells the septal HADA band, which marked the sites of autolysin activity that leads to eventual cell separation, was often shorter than the growth bands of HADA incorporation at mid-coccus (Figure 1C). In LCA-treated cells, the HADA also incorporated into a triplet pattern; however, the middle septal band was longer (Figure 1C). These longer septa could be due to an imbalance in peptidoglycan synthesis and autolysin activity, which is required for successful cell separation. The presence of triplets of HADA bands, however, suggests that overall cell polarity and geometry were unaffected by LCA and that the synthesis machinery of peptidoglycan assembled normally at, or near to, mid-coccus. Our laboratory strain of E. faecium undergoes spontaneous autolysis in static cultures left over the course of a few days. The presence of LCA, DCA and CDCA resulted in significant reductions in spontaneous autolysis (Figure S1F), suggesting that bile acids may at least partially inhibit autolysin activity.

LCA induces biofilm formation

Biofilm formation is an important factor in multiple colonization models with enterococci (Paganelli et al., 2012) and it is possible that gut microbes may respond to gut-associated molecules by increasing biofilm formation. We tested whether a panel of bile acids also increased VRE biofilm formation. All bile acids tested increased VRE biofilm formation compared to DMSO control, but LCA was the most potent, with a significant increases at 125 μM and 250 μM (Figure 1D). Biofilm formation in E. faecalis and E. faecium requires autolysin (Guiton et al., 2009; Mohamed et al., 2004; Paganelli et al., 2013; Thomas et al., 2009). To determine if LCA-induced biofilm formation depended on the major autolysin AtlA in E. faecium E1162, we assayed an atlA deletion mutant and a plasmid-complemented strain. LCA-induced biofilm formation was lost in atlA mutant cells and was rescued in the plasmid-complemented strain (Figure 1E). These data suggest that LCA mediates biofilm formation through the canonical AtlA-dependent pathway in VRE.

Cations reverse the effect of LCA

Bile acids are amphipathic molecules with the potential to chelate cations. Cation concentration is critical to peptidoglycan biosynthesis and mediates cell separation of S. pneumoniae (Rochford and Mandle, 1953). To test if cations played a role in VRE chaining we treated cultures with MgCl₂ and asked whether this ameliorated the effects of LCA on VRE morphotype. Growth in media supplemented with MgCl₂ inhibited LCA-induced sedimentation of VRE chains after overnight growth (Figure S3A). VRE chain length and width was smaller in the presence of LCA plus MgCl₂ after 4 hours of growth, suggesting that normal cell separation had been partially restored (Figure 2A–B). Treatment with MgCl₂ also reversed the effect of LCA on biofilm formation (Figure S3B, 2C). These data support a link between chaining, biofilm formation and VRE cell surface charge. To test if cation chelators can affect cell separation, we grew VRE in the presence of the cation chelator EDTA and observed significant chaining which was also reversed by adding MgCl2 to the media (Figure S3D). We did not, however, detect significant differences in biofilm formation upon treatment with EDTA (Figure S3E).

Figure 2. LCA-induced morphotype switch, biofilm and aggregation in intestine was inhibited by magnesium chloride.

A&B) Quantification of cell length and width of VRE processed for HADA staining in Figure 1C. Cells were grown for 4 hours in the presence of DMSO, 125 μM LCA and/or 100 mM MgCl2, as indicated, n = 160–200 chains per condition, Kruskal-Wallace one-way ANOVA with Dunn’s correction vs. DMSO. Boxes represent median and quartiles, whiskers represent range. C) Biofilm formation in the presence or absence of DMSO, LCA and/or MgCl2, as indicated, following 36 hours of growth. Assays were performed three times in triplicate, Kruskal-Wallace one-way ANOVA with Dunn’s correction vs. DMSO for each strain. Bars represent mean +/− SD. D) Chain length of VRE following overnight culture in naïve mouse cecal content extract. Data are from 4 pooled experiments n=800–1000 chains per condition, Unpaired Mann-Whitney U-test. E) Quantification of VRE aggregation in ceca of LCA-treated mice: Mice were treated with antibiotics then fed 5mg LCA on days −2, −1, and 0 of colonization in the presence or absence of MgCl2. Mice were sacrificed 24 hours after colonization and processed for fluorescence in-situ hybridization. Each data point represents percentage of aggregates containing 4 or more cells in ROIs of equal size. 5 ROIs were quantified per mouse in a single section, containing 400–3000 cells each, Kruskal-Wallace one-way ANOVA with Dunn’s correction vs. Corn oil + Water. F) One representative field from FISH imaging (quantified in Figure 2E) per condition. White triangles point to VRE aggregates. Sub-field in each dashed box is magnified in 4× in the adjacent image.

Scale bars = 10μm. * p < 0.05, ** p < 0.005, **** p < 0.0001.

To determine if the effects of LCA and MgCl₂ on biofilm formation were conserved more widely among enterococci we investigated Vancomycin-sensitive E. faecium E1162, Vancomycin-resistant E. faecalis V583 and Vancomycin-sensitive E. faecalis JH2–2 and OG1RF. LCA increased biofilm formation in all strains and this was inhibited by 100 mM MgCl₂ (Figure 2C). These results confirmed that the effect of bile acids on biofilms is conserved across diverse enterococci.

Next, we examined the effects of cations and bile acids on VRE during intestinal colonization. We treated mice with Ampicillin and then gavaged the mice with either MgCl₂ or a control for three successive days before and concomitant with VRE gavage. (Figure S3C) We did not find a statistically significant decrease in VRE burden. This could be due to many complexities of the environment in vivo, such as the presence of other cation sequestering molecules or the absorbance of MgCl₂ by the host.

Enterococci colonize antibiotics naïve mice when added to drinking water over a period of weeks (Kommineni et al., 2015), however, it likely does not colonize at a titre high enough to detect by microscopy against the background of a full gut microbiota. Antibiotic treatment breaks anti-VRE colonization resistance, but also alters the bile acid pool and greatly reduces secondary bile acid levels (Buffie et al., 2015). To determine if the morphotype switch occurs in a context more directly relevant to gut colonization we cultured VRE ex vivo in liquid Brain Heart Infusion media mixed 1:1 with an extract of cecal content from naïve mice. In these conditions we observed a modest, but significant increase in mean chain length of VRE (Figure 2D).

To determine if LCA affects VRE morphology in the mouse intestine, we treated mice with antibiotics then fed mice LCA in the presence or absence of MgCl2. We observed VRE in the cecum using fluorescence in situ hybridization (Caballero et al., 2015). We were not able to clearly distinguish chains of VRE in LCA-treated mice, however, we did observe a substantial difference in aggregation of VRE in mice treated with LCA (Figure 2E–F). These VRE aggregates were not present in mice treated with both LCA and MgCl₂. Taken together, these data suggest that LCA and MgCl₂ affect VRE morphotype and aggregation in vitro and during intestinal colonization.

Experimental evolution selection for mutants locked in diplococcal mode

To determine the molecular mechanism underlying the Bile acid-induced chaining, we adapted experimental evolution (van Ditmarsch et al., 2013; Kawecki et al., 2012) to take advantage of the relative density of chains versus diplococci. Chains fall to the bottom of standing cultures while diplococci are enriched at the top, resulting in low optical density of the planktonic portion of an overnight culture treated with LCA (Figure 3A). We started serial passages from the top of cultures to fresh culture media in two lineages in parallel: DMSO and LCA treated (50 μM), daily for five weeks, comprising approximately 350 generations. Cultures were saved as freezer stocks periodically during the course of selection, then the stocks were assayed for sedimentation, chain length and mouse intestinal colonization. After 14 days the planktonic optical density of the LCA-treated lineage increased to a level similar to the ancestral untreated strain (Figure 3B). The mean chain length of the LCA-treated population, however, did not decrease to that of the ancestral untreated strain until day 35 of passage (Figure 3C). These data suggest that selection for buoyancy preceded selection for smaller chains.

Figure 3. Experimental evolution selection against LCA-induced chaining.

A) Schematic of selection protocol. Two lines were passaged in parallel. Cells were grown overnight in 50 μM LCA or DMSO and diluted 1:1000 in fresh media for 5 weeks. B) Chain precipitation in selection stocks. Planktonic = optical density of top of culture, Total = optical density of mixed culture, following overnight growth in LCA. Each bar represents the mean and SD of 2 experiments performed in triplicate for the planktonic and total OD for the culture of a single stock after 18 hours of growth in LCA. C) Chain length of selection stocks following overnight growth in the presence or absence of LCA as indicated, n ~ 200–400 chains per condition, Kruskal-Wallace one-way ANOVA with Dunn’s correction vs. VRE - LCA, **** p < 0.0001 D & E) Ampicillin-treated mice were colonized with VRE or the d35-LCA population stock. D) Fecal CFU burden 12 hours post-colonization. Data are combined from 3 experiments, each data point represents 1 mouse, Mann-Whitney test, **** p < 0.0001. E) Long-term kinetics of fecal CFU burden of mice from D. Each point represents the mean and SD of 12–14 mice.

We then tested whether the day 35 LCA-treated population (D35LCA) could colonize ampicillin treated mice. Mice gavaged with D35LCA harbored less CFUs in fecal pellets at 12 hours post-infection compared with mice gavaged with the same amount of the ancestral VRE strain (Figure 3D). After 12 hours, mice remained stably colonized with no difference between ancestral VRE and D35LCA, indicating that the colonization defect was transient (Figure 3E).

This selection could have been confounded by the simultaneous selection for buoyancy and other effects of prolonged serial passage. In order to identify the genes underlying LCA-induced chain formation and to minimize the number of accumulated mutations, we isolated clones from day 7 of the serial passage, a time-point when the planktonic optical density had begun to rise, but had not yet reached the level of the ancestral strain. We screened 500 colonies from the LCA-treated passage for optical density after overnight growth in 50 μM LCA. Cultures with a planktonic OD greater than 0.4 were observed by light microscopy. From this, we isolated 4 clones that appeared to form predominantly diplococci in the presence of LCA. We also isolated 4 clones from the LCA passage with a planktonic OD of >0.4 that appeared to form chains when grown overnight in LCA. We also isolated 4 clones with a planktonic OD of < 0.1 to uncover SNPs associated with continuous LCA treatment, without an effect on chain formation. From the DMSO-treated passage we isolated 4 clones with a planktonic OD of > 0.2 after growth in LCA and 4 clones with a planktonic OD < 0.1. These clones comprised a collection of 5 distinct phenotypic classes of isolates, which we refer to as phenotypic classes I–V (Table 1). We sequenced the whole genomes of 19 clones using Illumina MiSeq and we identified the nucleotide variants using breseq (Deatherage and Barrick, 2014). The isolates accumulated 12 variants (non-synonymous SNP, substitution, or deletion) shared among all members of a phenotypic class, suggesting that they were linked to the underlying phenotypes (Figure 4, Table S1). Surprisingly two of the SNPs occurred in genes frequently mutated in daptomycin-resistant enterococci, the essential two-component kinase walK/yycG and the three-component response regulator liaR (Diaz et al., 2014; Miller et al., 2013).

Table 1:

Selection strains, Phenotypic classes.

Class	Passage	Chains?	OD
I	LCA	No	High
II	LCA	Yes	High
III	LCA	Yes	Low
IV	DMSO	Yes	Medium
V	DMSO	Yes	Low

Figure 4. Identification of mutations accumulated during selection.

Plot of variants mapped to the reference genome of VRE (E. faecium ATCC 700221). Outer ring, reference genome comprising a 2.85 Mb chromosome and three plasmids. Inner blue rings, 3 classes of isolates from the LCA-treated passage: 4 isolates of Class I - Diplococcus High OD, 4 isolates of Class II - Chains High OD, 3 isolates of Class III - Chains Low OD. Inner grey rings, 2 classes of isolates from the untreated passage: 4 isolates of Class IV - Chains Medium OD, 4 isolates of Class V - Chains Low OD. Here, OD refers to the planktonic OD of the isolate after culture overnight in LCA. Variants are mapped to the inner rings as vertical hatch marks and are colored according to the legend. Most variants were identical in all affected isolates. Exceptions include arsC, present as a 1bp deletion in 13/16 affected isolates and D58Y variant in 3/16 affected isolates. walK contained A88R in 1/4 affected isolates and R262P in 3/4 affected isolates.

VRE locked in diplococcus mode are daptomycin sensitive and defective in mouse colonization

To link phenotype to genotype, we screened the clone collection in a series of assays. We measured changes in LCA-induced chain length by flow cytometry for the 19 sequenced strains. Only the strains from Class I had mean forward scatter values that were significantly lower than VRE treated with LCA (Figure 5A). As walK is the only gene to contain a SNP in all 4 clones sequenced from class I, it suggests that signaling through WalK modulates chain length in response to LCA. However, we cannot rule out the effects of pleiotropy from other SNPs in the strain background. Recent developments in the field suggest that attempts to genetically alter the function of walK, which is an essential gene, frequently results in confounding off-site compensatory mutations (42–44).

Figure 5. Phenotypic linkage.

A) Strains were grown for 12 hours in the presence of LCA with the exception of VRE UT which was grown in the presence of DMSO. Forward scatter was measured by flow cytometry as a proxy for chain length, twice in triplicate, Kruskal-Wallis One-way ANOVA with Dunn’s correction relative to VRE+LCA. B) Daptoymycin MIC for each strain as measured by Etest three times, Kruskal-Wallis One-way ANOVA with Dunn’s correction relative to VRE. C) Biofilm on polystrene in the presence and absence of LCA. Assays were performed twice in triplicate, unpaired t-test with Holm-Sidak correction for multiple comparisons, relative to DMSO for each strain. * p < 0.05, ** p < 0.005.

In strains from class II, forward scatter was not significantly reduced. Chain length of class II strains was significantly reduced and between that of class I and ancestral VRE treated with LCA when quantified by light microscopy (Figure S4). The response regulator liaR contains the only SNP exclusive to Class II, suggesting that it regulates VRE chain length along with WalK. Both walK and liaR are frequently mutated in daptomycin-resistant enterococci (Diaz et al., 2014). Therefore, we measured daptomycin MIC values for all strains in the isolate collection and found that strains from class I and class II were significantly more sensitive than the ancestral strain (Figure 5B). Deletion of liaR was sufficient to restore sensitivity to daptomycin resistant E. faecium (Panesso et al., 2015), suggesting that our allele is likely to be a loss-of-function. We assayed LCA-induced biofilm formation and found that only a single strain from class V retained the ability to form robust biofilms (Figure 5C), suggesting that this phenotype was highly sensitive to the effects of serial passage alone. SNPs in guaC, epsD and pstS are present in all phenotypic classes except class V, suggesting that they may be linked to loss of LCA-induced biofilm formation.

To determine which phenotypic classes of strains were impaired in mouse intestinal colonization, we colonized mice by oral gavage with the VRE stock strain, or one representative of each class. We found that the isolate from class I (L14), the only class that is locked in diplococcus mode (Table 1), was significantly impaired in colonization of ampicillin-treated mice (Figure 6A). We then colonized mice with L14 and the other diplococcus-locked isolates of Class I (L67, L127 and L131) and found that 2 of 4 isolates were significantly impaired in intestinal colonization (Figure 6B). While secondary bile acids are undetectable in antibiotic-treated mice, total bile acid levels increase due to disruption of a negative feedback loop on bile acid production mediated by the gut microbiota (Kuribayashi et al., 2012; Sayin et al., 2013). It is likely that the colonization defect seen in the Class I strains is representative of the inability to respond to high levels of primary bile acids, which also induce chaining albeit with less affinity than LCA (Figure 1B). The variability in colonization levels among the Class I strains (L14 and L127 vs. L67 and L131) may also reflect the effects of genetic changes unrelated to chain formation. To determine if the colonization defect was simply caused by growth inhibition in intestinal content, we cultured VRE for 12 hours ex vivo in cecal content from antibiotic naive and ampicillin-treated mice (Figure 6C). We observed no difference in CFUs of VRE or L14 in cecal content ex vivo, suggesting that the colonization defect is not simply due to growth inhibition, but is specific to the host intestine.

Figure 6. Diplococcus-locked isolates have a colonization defect and reduced persistence in the intestine.

A) Ampicillin treated mice were colonized with VRE or one isolate of each phenotypic class. CFU burden of cecal content 12 hours post-colonization. Strains from the LCA passage are colored blue. Strains from the untreated passage are colored grey. Class I = L14, Class II = L61, LCA Class III = L111, Class IV = N5, Class V = N1. Data are combined from two experiments, each data point represents 1 mouse, Kruskal-Wallis one-way ANOVA relative to VRE, *** p = 0.0002. B) CFU burden in cecal content 12 hours post-colonization with each strain from Class I. Data are combined from two experiments, each data point represents 1 mouse, Kruskal-Wallis one-way ANOVA relative to VRE, ** p = 0.0069, *** p = 0.0002.C) VRE and L14 cultured ex vivo in cecal content from antibiotic naive and ampicillin-treated mice after 12 hours of culture. Data are combined from 2 experiments, each data point represents ex vivo culture in content from 1 mouse. D) Fecal CFU burden of germ-free mice colonized with VRE and L14. Arrow indicates the time point of fecal microbiota transplantation, n = 3 mice per group.

To determine if the colonization defect of L14 was predominantly due to the microbiota or the host, we colonized germ free mice with VRE or L14. Both strains colonized germ-free mice to similar levels as ampicillin-treated mice (Figure 6D). Observing the same level of germ-free colonization in the two strains indicated that the L14 colonization defect was microbiota-dependent, and not due to factors such as intestinal peristalsis, or an inability to survive in the anoxic gut. Then, because VRE can be readily cleared from the mouse intestine by fecal microbiota transplantation (FMT) (Ubeda et al., 2013), we performed FMT on the mono-colonized ex-germ free mice with feces from age-matched conventional mice. L14 was cleared more efficiently by the FMT than the ancestral VRE (Figure 6D). Taken together, these data suggest that the ability to form chains in response to LCA is necessary for VRE colonization and persistence in the intestine in the presence of the gut microbiota.

Discussion

We found that bile acids–especially the secondary bile acid lithocholic acid (LCA)–induce a morphotype switch from diplococcus to chains in the antibiotic-resistant enteric pathogen VRE. Genetic studies in enterococci have identified multiple genes that regulate chain length. Deletion of autolysins resulted in increased chain length in both E. faecalis (Qin et al., 1998; Salamaga et al., 2017) and E. faecium (Paganelli et al., 2013). Deletion of gelE, encoding a zinc-metalloprotease gelatinase, also resulted in longer chains of strain OG1RF and decreased autolytic activity, suggesting that gelatinase was necessary for autolysin activation (Waters et al., 2003). Proteolysis of AtlA by GelE was later confirmed biochemically (Thomas et al., 2009). However, in a study with a large collection of clinical isolates, chain length did not correlate with GelE production (Arias et al., 2007), suggesting that other genes may be involved in the regulation of chain length.

Bile acids are only one class of molecules present in the intestine. Previous work found that another common intestinal molecule, lysozyme, also increased chain length of E. faecalis. That switch depended on the Extra-Cytoplasmic Function (ECF) sigma factor SigV and the membrane-bound zinc metalloprotease Eep (Varahan et al., 2013). Altogether, these data suggest that enterococci may have evolved to switch to chained growth in the presence of multiple intestinal molecular cues. LCA treatment resulted in long chains and inhibition of autolysin activity in vitro, as well as increased biofilm formation. Linkage here between chaining and biofilm could reflect either residual autolysin activity due to incomplete inhibition, or other alterations of the cell surface proteome by bile acids that remain to be discovered. An experimental evolution study selecting for increased resistance to daptomycin in E. faecalis produced isolates that have increased chaining and increased biofilm formation (Miller et al., 2013). These data suggest that cell separation, biofilm formation and antibiotic resistance are mechanistically linked.

We also observed that the switch to chained growth could be reversed by MgCl₂. Bile acids are amphipathic and have the potential to chelate cations away from the cell surface, which may affect the ability of individual VRE daughter cells to fully separate after cell division. The strongest inhibitors of separation were the secondary bile acids LCA and DCA, both found predominantly in the cecum and colon. Upon entering the cecum, enterococci may switch from diplococcus to chains, increasing adherence, biofilm formation, resistance to antibiotics, and potentially also to inhibitory molecules present in the lower intestine such as RegIIIγ (Brandl et al., 2008) and IgA (Hendrickx et al., 2015). Our inability to detect obvious long chains in the cecum of LCA-treated mice suggests that the ability of bile acids to affect chain length in the intestine may be affected by other gut metabolites in the intestinal milieu. The final length of a VRE chain and its commitment to biofilm versus vegetative growth is likely to be determined by the integration of the effects many gut metabolites.

In S. pneumoniae, a mechanism of morphotype switching was proposed that balances distinct advantages of short and long chains. Longer chains have the disadvantage of an increased probability of being bound by complement and of coming into contact with phagocytic host cells, however, the increased surface area of longer chains may also advantageously increase adherence to host surfaces (Weiser 2013). In a zebrafish model of systemic infection, long chain autolysin mutants of E. faecalis were less virulent and more susceptible to phagocytosis (Salamaga et al. 2017). During colonization of the intestine, formation of chains by VRE in response to bile acids could be advantageous for adherence to host surfaces, but may also increase the likelihood of binding by RegIIIγ and IgA and adherence of VRE to feces destined for excretion.

The chelation of cations away from the enterococcal cell surface by bile acids may also explain–at least partially–why daptomycin fails to treat intestinal colonization by enterococci, as daptomycin requires cations for its activity (Miller et al., 2016). In vitro, E. faecalis is capable of increasing resistance to daptomycin by incorporating lipids into the cell membrane from serum, whole bile and purified fatty acids such as oleic acid (Saito et al., 2014). Although treatment of mice with MgCl₂ alone was not sufficient to reduce colonization of mice by VRE, it is possible that infusion of MgCl₂ or CaCl₂ may increase the separation of VRE and increase the effectiveness of daptomycin in the intestine. The intestinal microbiota of hematopoietic stem cell transplant patients is often chronically dominated by VRE, which is linked to less favorable treatment outcomes (Taur et al., 2014). An ability to selectively flip the morphotype switch so that it favors clearance of VRE, may enhance both the elimination of VRE and the efficacy of fecal microbiota transplantation (McKenney and Pamer, 2015).

Our data suggests a conserved link between chain morphotype and antibiotic resistance, a finding relevant for the treatment of VRE. Isolates of Gram-positive bacteria resistant to vancomycin and daptomycin often exhibit defects in cell separation and autolysis. A series of Staphylococcus aureus isolates with intermediate vancomycin sensitivity (VISA), taken over time from the same patient, showed rates of autolysis that were inversely correlated with vancomycin MIC, and the strains with the highest MIC failed to separate and grew as aggregates (Sieradzki and Tomasz, 2003). Cell separation was also affected in daptomycin resistant (technically non-sensitive) Enterococcus isolates from a patient (Arias et al., 2011; Tran et al., 2013) and in strains selected for increased daptomycin resistance by serial passage in the presence of the antibiotic (Miller et al., 2013). Imaging with fluorescently labeled daptomycin suggests that access to the septal membranes is required for killing enterococci (Miller et al., 2016; Tran et al., 2013). Further supporting the association between antibiotic resistance and chain length, we saw that mutants selected for defective LCA-induced chaining had SNPs in genes commonly associated with VISA (Howden et al., 2011) and daptomycin-resistant enterococci. Importantly, the three-component system liaFSR and the two-component system yycGH (also: walKR) are among the genes most commonly mutated in DAP-resistant patient isolates (Diaz et al., 2014). Decreased cell separation may be an adaptation that helps enterococci survive antibiotics that target the cell envelope.

Although both vancomycin and daptomycin are derived from microbes, it is unclear to what extent antibiotics function as killers in the natural environment (Romero et al., 2011) and it is unlikely that enterococci would have encountered clinical concentrations of daptomycin before the modern era. Therefore, adaptations to life in the mammalian gut, such as sensing and responding to bile acids may have resulted in an exaptation (or pre-adaptation) that potentiated the rise of antibiotic-resistant enterococci (Didelot et al., 2016; Gould and Vrba, 1982). Recent data suggests that the emergence of stress-resistant enterococci occurred around the terrestrialization of vertebrates (Lebreton et al., 2017). It is possible that millions of years of evolution within the bile acid-saturated intestinal tracts of vertebrates inadvertently primed enterococci to become successful antibiotic-resistant hospital pathogens.

Taken together, these data suggest that the host environment alters the antibiotic sensitivity of a colonizing pathogen. In the future it may be useful to study antibiotic resistance ex vivo in host-derived fluids such as serum or intestinal content extracts. Through increased knowledge of cell division, adherence and resistance in assays that more accurately reflect the intestinal niche of VRE, we may be able to design therapeutics that more effectively target this transmissible and problematic microbiont (Miles et al., 2015).

STAR Methods

CONTACT FOR REAGENT AND RESOURCE SHARING

Further information and requests for resources and reagents should be directed to and will be fulfilled by the Lead Contact, Eric G. Pamer (pamere@mskcc.org).

EXPERIMENTAL MODEL AND SUBJECT DETAILS

Bacteria and growth conditions.

We used the Vancomycin-resistant Enterococcus faecium strain 700221 from ATCC and its derivative mutants except when noted. Brain Heart Infusion liquid media was used for routine culture. Enterococcosel agar containing 100 μg/mL streptomycin and 8 μg/mL Vancomycin was used for selective plating in assays where CFUs were quantified. Bile acid stocks were prepared as 50 mM stock solutions in DMSO and equivalent amounts of DMSO were used as untreated controls.

Mouse colonization.

Animal experiments were approved by the Institutional Animal Care and Use Committee at MSKCC. All animals were maintained in a specific pathogen-free facility at Memorial Sloan Kettering Cancer Center. All mouse handling was performed in a biosafety level 2 facility wearing sterile gowns, masks and gloves. C57BL/6 female mice aged 6–8 weeks were purchased from one of two specified breeding rooms at Jackson Laboratories to ensure relative consistency of microbiota, Ampicillin sensitivity and VRE colonization. C57BL/6 female germ free mice aged 6 weeks (Figure 6D) were obtained from an isolator maintained by the gnotobiotics core facility at Weill Cornell Medical College. Mice were transferred to the SPF facility at MSKCC and were immediately colonized under sterile conditions.

METHOD DETAILS

Biofilm assays.

Polystyrene plate biofilm assays were performed using standard methods (Iyer and Hancock, 2012). Assays were performed in 24-well plates in the outer ring of wells to minimize the effect of plate position and gas exchange. An overnight culture was diluted 1:50 and was allowed to grow to mid-log phase. It was then diluted 1: 1000 in Todd-Hewitt Broth containing 0.25% Glucose and was grown 36 hours in the plate at 37C. The biofilms were not strongly adherent, so care was taken to avoid disrupting them. Media removal, all washes, and adding of stain was performed by tipping the plate to an angle of 45° and pipetting gently up or down the side of the well. Culture was removed and washed twice with water before staining for 15 minutes with 0.1% Crystal Violet. Stain was removed and washed twice with water. Plates were inverted and dried overnight. Stain was solubilized in 30% Acetic acid and optical density was read in cuvettes in a Beckman Coulter DU720 Spectrophotometer.

Microscopy and image quantification.

Routine light microscopy (Figure 1A, Figure 3C) was performed on an Axioplan 2 upright light microscope equipped with an Axiocam HRm digital camera (Carl Zeiss). Cells were pelleted, resuspended in PBS, applied to Poly-L Lysine coated cover slips and imaged. Chain length was quantified manually with ImageJ in most experiments due to difficulty in computationally segmenting long chains of VRE with existing software tools.

High-resolution microscopy.

Peptidoglycan was labeled by adding 1mM HADA (7-hydroxycoumarin-3-carboxylic acid (HCC-OH) coupled to 3-amino-D-alanine) to the culture (Botella et al., 2017; Kuru et al., 2012). Bacteria were washed 3 times with PBS and fixed with 3.2% paraformaldehyde for 30 minutes. Bacterial suspensions were deposited on soft agar pads and visualized using a DeltaVision image restoration microscope (GE Healthcare) equipped with an Olympus IX-71 microscope with a 100×/1.4 numeric aperture (NA) UPlanSApo objective and appropriate filter sets (for HADA, excitation at 390/10 and emission at 435/48), a pco.edge scientific complementary metal oxide-semiconductor (sCMOS) camera (PCO-Tech), and an Insight SSI 7-color solid-state illumination system. Images were analyzed using MATLAB and MicrobeTracker (Sliusarenko et al., 2011).

Transmission electron microscopy.

The pellet was fixed for one hour in 2.5% Glutaraldehyde / 2% Paraformaldehyde in Cacodylate buffer for one hour. It was then rinsed in Cacodylate buffer and post fixed in 2% Osmium Tetroxide for one hour. The sample was then rinsed in double distilled water followed by a graded series of alcohol 50%, 75%, 95% and absolute alcohol twice. This was followed by Propylene Oxide and finally a 50/50 mixture of Epon 812 resin and Propylene Oxide overnight. The sample was embedded the following day in Polyscience Epon 812 resin in Beem capsules and cured in a 60C oven for forty eight hours. Ultra-thin sections were obtained with a Leica Reichert Ultracut S Ultra microtome. The sections were stained in 5% Uranyl Acetate followed by Lead Citrate. Images were obtained using a Jeol 1200 EX Transmission Electron microscope.

Mouse colonization.

All animals were maintained in a specific pathogen-free facility at Memorial Sloan Kettering Cancer Center. All mouse handling was performed in a biosafety level 2 facility wearing sterile gowns, masks and gloves. C57BL/6 female mice aged 6–8 weeks were purchased from one of two specified breeding rooms at Jackson Laboratories to ensure relative consistency of microbiota, Ampicillin sensitivity and VRE colonization. All mice were single-housed in sterile cages with irradiated food and acidified water for all experiments. In experiments involving Ampicillin treatment (Figure 3D–E, Figure 6A–C), mice were administered 0.5g/L of Ampicillin (Fisher) in drinking water for 7 days, with water and cage changes every 3 days. Antibiotic treatment for FISH (Figure 2C–D) was Metronidazole, Vancomycin and Neomycin (0.25 mg/L each) in drinking water for 3 days, followed by 2 days rest on facility water and subcutaneous injection with 200 μg Clindamycin 24 hours before gavage with VRE. Mice in this experiment were gavaged with 5mg LCA in corn oil, or corn oil, followed by 100 μL 100mM MgCl2 on days −2, −1 and 0 (2 hours prior to gavage). Strains to be administered were grown overnight in BHI, diluted 1:4 in fresh media, were allowed to grow for 2 hours at 37C and then diluted 1:10,000 in sterile PBS to yield a consistent inoculum of 5×10³ CFU. Inoculum was confirmed by plating. All mice were inoculated at night, during the active period (Zeitgeiber Time 14–18 on a 12-hour light dark cycle) when bile acid levels are naturally high.

C57BL/6 female germ free mice aged 6 weeks (Figure 6D) were obtained from an isolator maintained by the gnotobiotics core facility at Weill Cornell Medical College. Mice were transferred to the SPF facility at MSKCC and were immediately colonized under sterile conditions. Fecal microbiota transplantation was performed using 2 age-matched female C57BL/6 mice obtained from Jackson Laboratories as donors. One fecal pellet from each mouse was resuspended in 1mL sterile and anaerobically reduced PBS. A single 200 μL dose of the fecal suspension was administered to each mouse by oral gavage.

Ex vivo Culture

One antibiotics naïve mouse was sacrificed at night. Cecal content was harvested, weighed and resuspended in sterile PBS at 0.1 g/mL. Content slurry was centrifuged briefly at 8000rpm, then filtered through a 0.22μm filter, split in half and extracted with 50 mg/ml of cholestyramine, or not, for 1 hour with continuous mixing at room temperature. The extract was then added to Brain Heart Infusion liquid media at 1:1.

FISH.

Cecal contents were fixed in methacarn solution (60% methanol, 30% chloroform, and 10% acetic acid). Tissues were washed in 70% ethanol, processed with a Leica ASP6025 processor and embedded in paraffin. Paraffin blocks were sectioned into 5μm sections and baked at 56°C for 1 hour prior to staining. Tissue sections were deparaffinized by xylene washes (2× 10 minutes) and were rehydrated through an ethanol gradient (95% for 10min, 90% for 10min, water). Probes were diluted in 0.9M NaCl, 20mM Tris-HCl at pH7.2 and 0.1%SDS. Sections were incubated with the VRE-specific probe at 50°C for 3 hours. Sections were washed in 0.9M NaCl, 20mM Tris-HCl at pH7.2 twice for 10 minutes and were then counterstained with Hoechst (1:3000 in wash buffer) for nuclear staining. The slides were digitally scanned using Pannoramic Flash 250 (3DHistech, Budapest, Hungary) using 40×/0.95NA objective with DAPI and FITC filters. Z-stacks spanning the section thickness were imaged with 0.2um step size. The projection images from the regions of interest within the lumen were exported to tiled .tiff images and analyzed using a custom macro in ImageJ/FIJI. Clusters of bacteria were segmented and filtered by size. The shape parameters for each cluster was measured and analyzed.

Experimental evolution.

The serial culture selection was started from a fresh single colony of E. faecium ATCC 700221 that was used to inoculate BHI containing 50 μM LCA or an equivalent volume of DMSO. Cultures were grown statically at 37C in 5mL polystyrene round bottom tubes (Falcon) overnight. Each culture was then sub-cultured by diluting 1:1000 in fresh media. The inoculum was taken from ~ 1 cm below the air-media interface, in an attempt to select for diplococci, which do not settle to the bottom of the tube, as opposed to large chains. Every 2–3 days a stock of the population was saved and frozen in 15% glycerol at −80C. Blank tubes of the two media were used to monitor contamination and the previous day’s cultures were kept at 4C in case of contamination. No contamination occurred over the 5 weeks of passage.

Day 7 clone isolation.

Individual clones were isolated from the day 7 freezer stocks by streaking on enterococcosel in the absence of antibiotics. Individual colonies were then streaked a second time to ensure clonal isolation. 500 individual colonies from the LCA-treated population were grown overnight in BHI containing 50 μM LCA without shaking and the planktonic optical density was measured from 1mL of the top of the static culture. The 40 colonies with the highest planktonic OD were observed by light microscopy and four clones (L14, L67, L127, L131) that appeared to lack LCA-induced chain formation, comprise phenotypic Class I: LCA-treated Diplococcus High OD. Clones with the highest planktonic OD that appeared to form chains by light microscopy (L61, L101, L259, L281) were saved and comprise phenotypic Class II: LCA-treated Chain High OD. Clones with the lowest planktonic OD (L111, L115, L120), comprise phenotypic Class III LCA-treated Chain Low OD. 100 individual colonies from the DMSO-treated population were similarly streaked twice on enterococcosel and 4 clones with the highest planktonic OD (N1, N32, N62, N63) and the lowest planktonic OD (N5, N13, N18, N23) comprise the phenotypic Class IV: Not treated Chain Medium OD, and Class V: Not treated Chain Low OD respectively.

Flow cytometry.

Bacterial cells were pelleted and fixed in 3.2% paraformaldehyde for 30 minutes. Cells were washed 3 times in PBS and resuspended in PBS containing 5 μg/mL DAPI. Flow cytometry was performed on an LSR II (Becton Dickinson) equipped with a 405nm laser and 450/50 bandpass filter for detecting DAPI. Unstained and unfixed controls were used to distinguish DAPI-positive VRE from debris.

SNP identification.

DNA was extracted from the 19 isolated strains using bead beating lysis and phenol-chloroform as described previously (Ubeda et al., 2010). Libraries were prepared for Illumina MiSeq by using the Nextera XT DNA library kit and MiSeq Reagent Kit v3 (Illumina) for 2×300bp paired-end sequencing. Reads were processed using a custom bash shell script. Read quality was examined using FastQC version 0.11.4. Reads were trimmed using Trimmomatic version 0.36 (Bolger et al., 2014). Reads were aligned to the reference and variants were called by breseq version 0.26 (Deatherage and Barrick, 2014). Variant call .gd files from breseq were manipulated using custom bash scripts. The variant plot in Figure 4A was generated using Circos version 0.69–5 (Krzywinski et al., 2009).

QUANTIFICATION AND STATISTICAL ANALYSIS

Statistical analysis.

Statistical analysis for comparison of more than 2 means was performed using ANOVA, Kruskal-Wallace one way ANOVA with Dunn’s correction. Comparison of 2 means was performed using the unpaired nonparametric Mann-Whitney U-test, or unpaired t-test with Holm-Sidak correction, as noted in the figure legends and text. Analyses were performed in GraphPad Prism version 7. Sample sizes can be found within figure legends and text for each experiment.

DATA AND SOFTWARE AVAILABILITY

Illumina Sequencing.

Raw sequencing reads were submitted to NCBI under BioProject accession number PRJNA377979. The E. faecium ATCC 700221 genome (McKenney et al., 2016) was re-annotated using the PATRIC web portal and is available on PATRIC (Wattam et al., 2014), with genome IDs 1352.804, 1352.2809, 1352.2810, 1352.2811.

KEY RESOURCES TABLE

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Bacterial and Virus Strains
Enterococcus faecium ATCC 700221	ATCC	ATCC: 700221
Enterococcus faecium E1162	Helen Leavis	NCBI: txid535206
Enterococcus faecium E1162 atlAEfm::genta	(Paganelli et al., 2013)	N/A
Enterococcus faecium E1162 atlAEfm::genta+ atlAEfm	(Paganelli et al., 2013)	N/A
Enterococcus faecalis V583	Lynn Hancock	NCBI: txid535206
Enterosoccus faecalis JH2–2	Stephane Mesnage	NCBI: txid1320322
Enterosoccus faecalis OG1RF	Stephane Mesnage	NCBI: txid474186
VRE L14	This Study	NCBI: SAMN06479922
VRE L67	This Study	NCBI: SAMN06479923
VRE L127	This Study	NCBI: SAMN06479924
VRE L131	This Study	NCBI: SAMN06479925
VRE L61	This Study	NCBI: SAMN06479926
VRE L101	This Study	NCBI: SAMN06479927
VRE L259	This Study	NCBI: SAMN06479928
VRE L281	This Study	NCBI: SAMN06479929
VRE L111	This Study	NCBI: SAMN06479930
VRE L115	This Study	NCBI: SAMN06479931
VRE L120	This Study	NCBI: SAMN06479932
VRE N1	This Study	NCBI: SAMN06479933
VRE N32	This Study	NCBI: SAMN06479934
VRE N62	This Study	NCBI: SAMN06479935
VRE N63	This Study	NCBI: SAMN06479936
VRE N5	This Study	NCBI: SAMN06479937
VRE N13	This Study	NCBI: SAMN06479938
VRE N18	This Study	NCBI: SAMN06479939
VRE N23	This Study	NCBI: SAMN06479940
Chemicals, Peptides, and Recombinant Proteins
Ampicillin	Fisher Scientific	Cat# 69-52-3
Paraformaldehyde	Electron Microscopy Sciences	Cat# 15714S
Glutaraldehyde	Electron Microscopy Sciences	Cat# 16320
Sodium Cacodylate Buffer	Electron Microscopy Sciences	Cat# 11652
Osmium Tetroxide	Electron Microscopy Sciences	Cat# 19152
Propylene Oxide	Electron Microscopy Sciences	Cat# 20414
Epon 812 Resin	Electron Microscopy Sciences	Cat# 14901
Uranyl Acetate	Electron Microscopy Sciences	Cat# 224002
Lead Citrate	Electron Microscopy Sciences	Cat# 17810
HADA	Sabine Ehrt	N/A
Brain Heart Infusion	BD	Cat# 211059
Todd Hewitt Broth	Neogen	Cat# 7161A
Crystal Violet	Fisher	Cat# C581
Acetic Acid	Fisher	Cat# BP1185
Enterococcosel Agar	BD	Cat# 212205
Vancomycin	Novaplus	Cat# 6340026
Streptomycin	Fisher	Cat# BP910
Metronidazole	Millipore Sigma	Cat# M1547
Neomycin	Fisher	Cat# BP26695
DAPI	Millipore Sigma	Cat# 268298
Magnesium Chloride	Fisher	Cat# BP214
Cholate	Sigma	Cat# C6445
Chenodeoxycholate	Sigma	Cat# C8261
Deoxycholate	Sigma	Cat# 30970
Lithocholate	Sigma	Cat# L6250
Critical Commercial Assays
Nextera XT DNA Library kit	Illumina	Cat# FC-131
MiSeq Reagent Kit v3	Illumina	Cat# MS-102
Deposited Data
VRE ATCC 700221 annotation	PATRIC	GenomeID: 1352.804
Experimental evolution isolate sequencing	NCBI BioProject	NCBI: PRJNA377979
Experimental Models: Organisms/Strains
Mouse: C57BL/6	The Jackson Laboratory	JAX: 000664, RRID: IMSR_JAX000664
Oligonucleotides
FISH Probe EUB388: [Cy3]-GCTGCCTCCCGTAGGAGT-[AmC7~Q+Cy3es	(Caballero et al., 2015)	N/A
FISH Probe Enfm93: [AminoC6+Alexa488]-GCCACTCCTCTTTTTCCGG-[AmC7~Q+Alexa488]	(Caballero et al., 2015)	N/A
Software and Algorithms
ImageJ	(Schneider et al., 2012)	RRID:SCR_003070
MATLAB	The Mathworks	RRID:SCR_001622
Microbe Tracker v. 0.937	(Sliusarenko et al., 2011)	RRID:SCR_015939
FlowJo v. 10.0	FlowJo, LLC	RRID:SCR_008520
FastQC v. 0.11.4		RRID:SCR_014583
Trimmomatic v. 0.36	(Bolger et al., 2014)	RRID:SCR_011848
PATRIC	(Wattam et al., 2014)	RRID:SCR_004154
breseq v. 0.28.1	(Deatherage and Barrick, 2014)	RRID:SCR_010810
Circos v. 0.69–5	(Krzywinski et al., 2009)	RRID:SCR_011798
Prism 7	Graphpad Software, Inc.	RRID:SCR_002798

Highlights.

• VRE forms long chains and biofilms in physiological concentrations of bile acids

• This morphotype switch is reversed by cations

• Selection against chaining is linked to sensitivity to the antibiotic, daptomycin

• Chaining-deficient VRE mutants exhibit reduced persistence in the gut