Complex Bacterial Consortia Reprogram the Colitogenic Activity of Enterococcus faecalis in a Gnotobiotic Mouse Model of Chronic, Immune-Mediated Colitis

Inflammatory bowel diseases (IBD) are associated with compositional and functional changes of the intestinal microbiota, but specific contributions of individual bacteria to chronic intestinal inflammation remain unclear. Enterococcus faecalis is a resident member of the human intestinal core microbiota that has been linked to the pathogenesis of IBD and induces chronic colitis in susceptible monoassociated IL-10-deficient (IL-10−/−) mice. In this study, we characterized the colitogenic activity of E. faecalis as part of a simplified human microbial consortium based on seven enteric bacterial strains (SIHUMI). RNA sequencing analysis of E. faecalis isolated from monoassociated wild type and IL-10−/− mice identified 408 genes including 14 genes of the ethanolamine utilization (eut) locus that were significantly up-regulated in response to inflammation. Despite considerable up-regulation of eut genes, deletion of ethanolamine utilization (ΔeutVW) had no impact on E. faecalis colitogenic activity in monoassociated IL-10−/− mice. However, replacement of the E. faecalis wild type bacteria by a ΔeutVW mutant in SIHUMI-colonized IL-10−/− mice resulted in exacerbated colitis, suggesting protective functions of E. faecalis ethanolamine utilization in complex bacterial communities. To better understand E. faecalis gene response in the presence of other microbes, we purified wild type E. faecalis cells from the colon content of SIHUMI-colonized wild type and IL-10−/− mice using immuno-magnetic separation and performed RNA sequencing. Transcriptional profiling revealed that the bacterial environment reprograms E. faecalis gene expression in response to inflammation, with the majority of differentially expressed genes not being shared between monocolonized and SIHUMI conditions. While in E. faecalis monoassociation a general bacterial stress response could be observed, expression of E. faecalis genes in SIHUMI-colonized mice was characterized by up-regulation of genes involved in growth and replication. Interestingly, in mice colonized with SIHUMI lacking E. faecalis enhanced inflammation was observed in comparison to SIHUMI-colonized mice, supporting the hypothesis that E. faecalis ethanolamine metabolism protects against colitis in complex consortia. In conclusion, this study demonstrates that complex bacterial consortia interactions reprogram the gene expression profile and colitogenic activity of the opportunistic pathogen E. faecalis toward a protective function.

Inflammatory bowel diseases (IBD) are associated with compositional and functional changes of the intestinal microbiota, but specific contributions of individual bacteria to chronic intestinal inflammation remain unclear. Enterococcus faecalis is a resident member of the human intestinal core microbiota that has been linked to the pathogenesis of IBD and induces chronic colitis in susceptible monoassociated IL-10-deficient (IL-10 −/− ) mice. In this study, we characterized the colitogenic activity of E. faecalis as part of a simplified human microbial consortium based on seven enteric bacterial strains (SIHUMI). RNA sequencing analysis of E. faecalis isolated from monoassociated wild type and IL-10 −/− mice identified 408 genes including 14 genes of the ethanolamine utilization (eut) locus that were significantly up-regulated in response to inflammation. Despite considerable up-regulation of eut genes, deletion of ethanolamine utilization ( eutVW) had no impact on E. faecalis colitogenic activity in monoassociated IL-10 −/− mice. However, replacement of the E. faecalis wild type bacteria by a eutVW mutant in SIHUMI-colonized IL-10 −/− mice resulted in exacerbated colitis, suggesting protective functions of E. faecalis ethanolamine utilization in complex bacterial communities. To better understand E. faecalis gene response in the presence of other microbes, we purified wild type E. faecalis cells from the colon content of SIHUMI-colonized wild type and IL-10 −/− mice using immuno-magnetic separation and performed RNA sequencing. Transcriptional profiling revealed that the bacterial environment reprograms E. faecalis gene expression in response to inflammation, with the majority of differentially expressed genes not being shared between monocolonized and SIHUMI conditions. While in E. faecalis monoassociation a general bacterial stress response could be observed, expression of E. faecalis genes in SIHUMI-colonized mice was characterized by up-regulation of genes involved in growth and replication. Interestingly, in mice colonized with SIHUMI lacking E. faecalis enhanced inflammation was observed in comparison to SIHUMI-colonized mice, supporting the hypothesis that E. faecalis ethanolamine metabolism protects against colitis in complex consortia. In conclusion, this study demonstrates that complex bacterial consortia interactions reprogram the gene expression profile and colitogenic activity of the opportunistic pathogen E. faecalis toward a protective function.

INTRODUCTION
Inflammatory bowel diseases (IBD) with the two dominant types Crohn's disease (CD) and ulcerative colitis (UC) are chronic relapsing inflammatory disorders affecting the distal intestine. Several factors appear to be involved in the pathogenesis of IBD, including genetic susceptibility (1,2) together with diverse environmental triggers resulting in an inappropriate T-cell mediated activation of immunity toward components of the intestinal microbiota (3)(4)(5). IBD is associated with compositional (6-10) and functional (11,12) changes of the intestinal microbiota referred to as dysbiosis. Dysbiotic changes associated with IBD are characterized by an overrepresentation of opportunistic pathogens and a loss of beneficial commensal organisms, indicating that the pathogenic potential of a dysbiotic community can be linked to certain organisms. Consequently, the specific contribution of individual bacteria to disease pathogenesis needs to be investigated in detail.
A relevant bacterial species in the context of IBD is Enterococcus faecalis. E. faecalis is a Gram-positive resident member of the human intestinal core microbiota harboring a number of pathogenic traits, which explains the association of this bacterium with inflammatory diseases and fatal nosocomial infections (13)(14)(15). Relative to healthy volunteers, the abundance of enterococci is increased in fecal samples of CD patients (16)(17)(18). This is in line with a high frequency of E. faecalis housekeeping and virulence genes in CD cohorts (19) and a high likelihood of the presence of virulence factors in E. faecalis isolates originating from inflamed IBD mucosa (20). UC patients have increased numbers of mucosa-associated E. faecalis correlating with disease activity (21) and a high E. faecalis-specific IgG sero-reactivity (22).
The dual characteristics of E. faecalis as core member of the human intestinal microbiota and as an opportunistic pathogen make this bacterium an ideal model to study microbe-host interactions relevant to the development of chronic colitis in a genetically susceptible host. In the IL10 tm1Cgn mouse (IL-10 −/− ), which is a well-characterized model of human chronic colitis, monoassociation with E. faecalis induces severe intestinal inflammation, whereas wild type mice remain disease-free (23)(24)(25)(26). In previous studies, we could show that certain bacterial structures contribute to the colitogenic activity of E. faecalis in monoassociated IL-10 −/− mice (27,28).
Using germ-free wild type and IL-10 −/− mice monoassociated with E. faecalis or colonized with a colitogenic human enteric bacterial consortium (SIHUMI) (29), we aim to unravel the functional relevance of E. faecalis in colitis development with regard to gene expression and interactions with co-colonizing bacteria. Based on RNA sequencing (RNA seq) analysis, we characterize the role of ethanolamine (EA) utilization for E. faecalis survival and colitogenic activity in a susceptible host.
The intestine is a rich source of the metabolite EA, which is derived from phospholipid phosphatidylethanolamine of bacterial and eukaryotic cell membranes (30,31). Bacteria capable of EA catabolism can utilize this compound as a source of carbon and/or nitrogen to promote growth as well as a signal influencing virulence during host colonization (32). EA utilization is a well-recognized property of diverse pathogens, for example Salmonella and enterohemorrhagic Escherichia coli (EHEC) (33)(34)(35)(36)(37). The only commensals known to carry EA utilization genes are E. faecalis and some commensal strains of E. coli (32,(38)(39)(40)(41). E. faecalis catabolizes EA using enzymes encoded by the eut genes that are contained in a locus consisting of 19 genes. The eut gene expression is controlled by the EutVW two-component system, comprising the sensor histidine kinase EutW and the response regulator EutV (42,43). Using an E. faecalis eutVW double deletion mutant (43), we are able to establish a novel correlation between E. faecalis EA utilization and an attenuation of the pro-inflammatory host response in the presence of a complex bacterial consortium.
Using immuno-magnetic separation, we are able to purify wild type E. faecalis cells from the colon content of SIHUMI-colonized wild type and IL-10 −/− mice and analyze the gene expression in the presence of other bacteria. We show that E. faecalis adapts transcriptionally to the other co-colonizing bacteria. To unravel whether these transcriptional alterations have functional consequences, we colonized wild type and IL-10 −/− mice with the SIHUMI consortium in the presence and absence of E. faecalis. We determine a protective activity of E. faecalis in SIHUMI colonized susceptible mice, demonstrating that complex bacterial consortia interactions can reprogram the colitogenic activity of the opportunistic pathogen E. faecalis toward a protective function.

Bacterial Strains and Cultivation
For this study we used the characterized human oral E. faecalis strain OG1RF and an isogenic eutVW double deletion mutant kindly provided by Danielle Garsin (Department of Microbiology and Molecular Genetics, The University of Texas Health Science Center at Houston, Houston, Texas, USA) (43). E. faecalis strains ( Table 1) were cultivated in Brain Heart Infusion (BHI) broth

Animal Experiments-Experimental Design
For monoassociation, germ-free 129S6/SvEv wild type and IL-10 −/− mice were colonized at the age of 10-12 weeks with either E. faecalis wild type OG1RF or the eutVW mutant strain by oral and rectal swab. The mice were sacrificed after different periods of colonization (1,4,8,12,16,20, and 24 weeks). For colonization with the SIHUMI consortium, the bacterial mixture was prepared as described above. Germfree 129S6/SvEv wild type and IL-10 −/− mice were colonized at the age 10 weeks with 1 × 10 8 bacteria/strain by repetitive oral gavage (at week 0 and week 1). The composition of the bacterial consortia differed between the experimental groups as shown in Table 2. The mice were sacrificed after colonization periods of 4 or 16 weeks. A schematic overview of all experimental setups is included in the Supplementary Material.

Preparation of Bacterial Lysates
Bacterial lysates for stimulation of mesenteric lymph node (MLN) cell cultures were prepared as described previously (29). Details are provided in the Supplementary Material.

Bacterial DNA Extraction
Bacterial cells were lysed by mechanical lysis using a FastPrep-24 (MP Biomedicals). DNA was isolated using phenol/chloroform/iso-amyl alcohol (Carl Roth) and contaminating RNA was digested with 0.1 µg/µl RNase A (VWR) for 30 min at 37 • C. Subsequently, the DNA was purified using the gDNA clean-up kit (Macherey-Nagel) according to the manufacturer's instructions.

Bacterial Community Analysis via Quantitative Real-Time Polymerase Chain Reaction (qPCR) Assay
For bacterial community analysis, oligonucleotide primers were used to target species-specific regions of the 16S ribosomal RNA (rRNA) gene. Quantification was performed using the LightCycler 480 Universal Probe Library System (UPL, Roche). Primers (Sigma Aldrich) and corresponding probes (Roche) are listed in Table 3. Positive and negative controls were included in every run. Standard curves were generated for each bacterial strain and used to enumerate the 16S rRNA gene copy numbers per species in intestinal content. 16S rRNA gene copy number per ng DNA was calculated via the molecular weight of the genome and 16S rRNA gene copy numbers/genome. The abundance of each bacterium was calculated as previously described (29) using the following formula: [(copies of 16S rRNA gene for a specific

Isolation of E. faecalis Cells From Intestinal Content by Immuno-Magnetic Separation
The isolation method was modified according to protocols of Nolle (48). Frozen colonic content from SIHUMI-colonized mice was disrupted by bead beating in 1 ml RNAlater (Sigma Aldrich) using glass beads (1 mm, Carl Roth). Subsequently, samples were diluted with wash buffer (PBS, 0.5% biotinfree BSA, 10% RNAlater) and filtered through three cell strainers with decreasing pore size (VWR; 100, 70, and 30 µm, respectively). The samples were kept on ice throughout and only pre-chilled buffer solutions were used. The bacterial cells were pelleted by centrifugation and resuspended in 1 ml wash buffer. The samples were mixed with 10 µl Enterococcusspecific antibody (biotinylated, Thermo Fisher Scientific PA1-73123) and incubated for 10 min at 4 • C under gentle agitation. The samples were centrifuged for 2 min at 9600 × g. After a washing step, the cells were resuspended in 100 µl wash buffer.

Bacterial RNA Isolation
Bacterial total RNA was isolated as described previously (49). Bacterial RNA prepared from the colonic content of E. faecalis OG1RF monoassociated wild type and IL-10 −/− mice (n = 8 mice/group; colonization period: 16 weeks) was sequenced as described previously (28). RNA isolation and rRNA depletion was done by Jonathan Hansen and Sandrine Tchaptchet from the Division of Gastroenterology and Hepatology of the University of North Carolina in Chapel Hill. Library preparation and sequencing was performed at the Washington University St. Louis Genome Technology Access Center generating in total 190 million unidirectional 50-bp reads using an Illumina HiSeq2000 instrument (Illumina).

RNA Seq of E. faecalis Isolated From SIHUMI Colonized Mice
E. faecalis cells in the colonic content of SIHUMI colonized wild type (n = 6 mice/group; colonization period: 16 weeks) and IL-10 −/− mice (n = 13 mice/group; colonization period: 16 weeks) were isolated by immuno-magnetic separation. Total bacterial RNA was isolated as described above and purified using the RNeasy MinElute Cleanup kit (Qiagen). Contaminating genomic DNA was eliminated using TURBO DNase (Thermo Fisher Scientific) and RNA integrity was validated by agarose gel electrophoresis. Ribosomal RNA was depleted by using the RiboMinus TM Transcriptome Isolation Kit (Bacteria, Thermo Fisher Scientific) according to the manufacturer's instructions. Subsequently, the RNA was fragmented for 180 s using a Covaris E220 sonicator set at 10% duty cycle, intensity of 5 and 200 cycles/burst in 120 µl 1 mM Tris-EDTA pH 8. The fragmented RNA was precipitated together with 1 µl glycogen (Thermo Fisher Scientific), 1:10 sample volume 3 M sodium acetate (Ambion) and 2.5 sample volume of 100% EtOH. Next, the RNA was dephosphorylated using 10 U Antarctic phosphatase per 300 ng RNA at 37 • C for 30 min (New England Biolabs). The RNA fragments were recovered using the miRNeasy Mini kit (Qiagen) followed by 5'-phosphorylation using 20 U T4 polynucleotide kinase at 37 • C for 60 min (Thermo Fisher Scientific). Again, the miRNeasy Mini kit was used for recovery. RNA fragment size was examined using the BioAnalyzer on an RNA Pico Chip (Agilent Technologies). The cDNA library was prepared as described in the TruSeq Small RNA Library Prep Reference Guide (Illumina) with the following exceptions: After adapter ligation and reverse transcription, fragments were size-selected between 145 and 300 bp corresponding to an insert length between 18 and 173 bp. The sequencing was conducted at the Core Facility Microbiome, generating in total 300 million unidirectional 50 bp reads using an Illumina HiSeq2000 instrument.

Bioinformatics
Data analysis was performed using the Galaxy web platform (50)

Histopathological Analysis
The intestinal tract of gnotobiotic wild type and IL-10 −/− mice was removed immediately after sacrificing the animals and cleaned from stool. Cross-sections or "swiss-roles" (52) of parts from the intestinal tract were fixed in 4% formalin, dehydrated, and embedded in paraffin. For histopathological analysis, tissue samples were cut into 5 µm sections, deparaffinized and stained with hematoxylin and eosin. The histopathology was evaluated by an independent examiner in a blinded manner as described previously (53), resulting in scores from 0 (not inflamed) to 12 (highly inflamed).

RNA Isolation (Mouse Tissue)
Tissue sections of distal colon were stored in RNAlater (Sigma Aldrich) at −80 • C. Tissue samples were transferred into 500 µl RA1 buffer (Macherey Nagel) supplemented with 10 mM DTT (Sigma Aldrich) and the tissue was homogenized using a Miccra D-1 homogenizer. RNA was isolated using the NucleoSpin RNA Kit (Macherey Nagel) according to the manufacturer's instructions.

Reverse Transcription and Gene Expression Analysis via qPCR (RNA From Mouse Tissue)
Complementary DNA (cDNA) was synthesized from 1 µg RNA using the Moloney murine leukemia virus (MMLV) Point Mutant Synthesis System (Promega) and random hexamers. Quantification was performed using the LightCycler 480 Universal Probe Library System (UPL, Roche). Primers (Sigma Aldrich) and corresponding UPL probes are listed in Table 4. Relative quantification of gene expression was calculated by means of 2 − Ct values: Relative gene expression values were normalized to GAPDH and HPRT expression.

MLN Cell Cultures
MLNs were aseptically harvested from gnotobiotic wild type and IL-10 −/− mice associated with the SIHUMI consortium. Subsequently, tissue was homogenized and filtrated through 70 µm cell strainers (VWR) resulting in unfractionated single cell suspensions. MLN cells, isolated from individual mice (for IL-10 −/− mice) or a pool of mice (wild type mice), were cultured (5 × 10 5 cells per well) in RPMI-1640 medium (Sigma Aldrich) containing 10% fetal calf serum (Biochrom) and 1% antibiotics/antimycotics (Sigma Aldrich) and stimulated with 15 µg/ml of bacterial lysate from the respective colonizing species. After incubation for 72 h, the supernatants were collected for cytokine analysis and stored at −80 • C.

Cytokine Quantification
IL-12p40 and IFNγ concentration in cell culture supernatant was quantified using commercially available ELISA kits (eBioscience).

Ethanolamine Quantitation in Colonic Content by LC-MS/MS Analysis
Quantitation of ethanolamine in colonic content from wild type and IL-10 −/− mice monoassociated with E. faecalis OG1RF or eutVW was performed at the Bavarian Center for Biomolecular Mass Spectrometry (BayBioMS, Freising, Germany). Sample preparation and mass spectrometry quantitation were modified after (54). Details are provided in the Supplementary Material.

Statistical Analysis
Statistical analyses were performed using Prism version 6.05 (GraphPad). Unless otherwise stated, data are presented as mean ± standard deviation. Shapiro-Wilk or D'Agostino-Pearson omnibus test was used to assess the normality of distribution of investigated parameters. Statistical significant differences between two groups were determined using two-tailed unpaired parametric t-test. Statistical significant differences between three or more unpaired groups were determined using oneway ANOVA followed by pairwise testing (Tukey). Effects of genotype and colonization groups were compared using two-way ANOVA followed by pairwise testing (Tukey). Differences were considered significant for * p ≤ 0.05, * * p ≤ 0.01, * * * p ≤ 0.001, * * * * p ≤ 0.0001.
Analysis of RNA seq data was performed using Galaxy platform web platform (50) or R as described in detail above.

E. faecalis Adapts to a Chronic Inflammatory Environment by Altering Its Gene Expression Profile
To analyse the kinetics of E. faecalis-driven inflammation, we monoassociated germ-free wild type and IL-10 −/− mice at the age of 10 weeks with E. faecalis wild type OG1RF for different periods of time. After 4, 8, and 12 weeks of colonization, IL-10 −/− mice showed mild signs of inflammation in the distal colon as indicated by the histological score.
Between the 12th and the 20th week of colonization, IL-10 −/− mice started to develop severe distal colonic inflammation reaching a plateau after week 20. A similar disease pattern was also observed in mid colon with an overall lower level of inflammation (Supplementary Figure 1A). In proximal colon, the inflammation persisted at a very low level throughout the analyzed colonization periods (Supplementary Figure 1B). All wild type mice remained disease-free ( Figure 1A). The colonization density as detected by countable colony forming units (CFU) in luminal colon contents was not affected by the level of inflammation ( Figure 1B). Based on the kinetic of distal colonic inflammation and on previous publications, 16 weeks of colonization were used in subsequent experiments. At this time, inflammation was still in the ascending phase and therefore advantageous for the analysis of factors modulating disease severity. We hypothesized that E. faecalis quickly adapts to a chronic inflammatory environment by altering its gene expression profile. These adaptation mechanisms are likely responsible for the persistence and virulence of these bacteria in a susceptible host. To assess E. faecalis disease-associated transcriptome, we monoassociated mice with E. faecalis OG1RF for 16 weeks and performed RNA sequencing of bacteria isolated from luminal colon content of healthy wild type and inflamed IL-10 −/− mice. Transcriptional profiling identified 408 significantly up-regulated and 468 significantly down-regulated genes under conditions of chronic inflammation (Figure 1C and Supplementary Figure 1C, Supplementary Table 1).
The samples clearly clustered according to mouse genotype as detected by principal component analysis (PCA) indicating that inflammation modulates E. faecalis gene expression (Supplementary Figure 1D). E. faecalis response to inflammation was characterized by an up regulation of genes associated with a bacterial stress response toward unfavorable growth conditions. KEGG pathway enrichment analysis showed an enrichment of the biosynthesis of amino acids, quorum sensing, mismatch repair, and oxidative phosphorylation among up-regulated genes ( Figure 1D). Induction of these processes has also been observed in previous studies analyzing E. faecalis and E. coli adaptation during periods of stress (55-57). Nucleotide synthesis genes were in general down regulated, as indicated by an enrichment of purine-and pyrimidine metabolism transcripts in wild type mice. Furthermore, a large number of genes involved in metabolism and nutrient acquisition were differentially expressed. Sugar metabolism and transport were repressed with an enrichment of starch and sucrose metabolism, ABC transport, PTS transport and glycolysis among downregulated genes. On the other hand, genes necessary for the utilization of alternative carbon sources were up regulated with an enrichment of inositol phosphate metabolism and glycerolipid metabolism ( Figure 1D). Interestingly, genes of the ethanolamine utilization (eut) locus were also identified among the top 20 up-regulated genes in response to an inflammatory environment ( Figure 1E). In total, 14 E. faecalis eut genes were significantly up regulated in inflamed IL-10 −/− mice indicating an important function of this metabolic pathway for the survival and virulence of E. faecalis in the inflamed intestine ( Figure 1F).

E. faecalis Ethanolamine Utilization Has Protective Functions in Gnotobiotic Il-10 −/− Mice Colonized With a Simplified Microbial Consortium
To determine whether the utilization of ethanolamine (EA) is relevant for E. faecalis colonization and colitogenic activity, we monoassociated germ-free wild type and IL-10 −/− mice with E. faecalis wild type OG1RF or an eutVW mutant strain ( eut) for 1 and 16 weeks (Figure 2A). The eut strain does not express the eut genes and is not able to utilize EA (43,58,59). After 1 week of colonization, wild type and IL-10 −/− mice showed no signs of inflammation. After 16 weeks of colonization, IL-10 −/− mice developed severe inflammation in the distal colon, independent of the E. faecalis strain used for monoassociation (OG1RF histopathological score: 5.0 ± 0.9, eut histopathological score: 4.6 ± 0.5). All wild type mice remained disease-free (Figures 2B,C). E. faecalis OG1RF and eut mutant strain showed a similar colonization density as detected by CFU in luminal contents from colon ( Figure 2D). The colon inflammation in IL-10 −/− mice was accompanied by increased spleen weights (Supplementary Figure 2A) and increased expression of pro-inflammatory markers in the distal colon tissue. The expression levels of TNF, IL-12p40, and IL-17 were similar for both colonization groups, only IFNγ expression levels were increased in E. faecalis wild type OG1RF compared to eut mutant monoassociated IL-10 −/− mice ( Figure 2E). To determine whether E. faecalis catabolizes EA in vivo, we measured EA concentrations in intestinal fluid. Luminal EA concentrations were significantly higher in wild type mice monoassociated with eut mutant, indicating catabolism of EA by E. faecalis OG1RF. This trend was also observed in IL-10 −/− mice after 4 weeks of colonization, but after 16 weeks of colonization intestinal EA concentrations were similar in both E. faecalis OG1RF and eut-mutant monoassociated mice ( Figure 2F).
Despite considerable up regulation of eut genes, the deletion of ethanolamine utilization revealed no impact on E. faecalis colitogenic activity in monoassociated mice. To test the hypothesis that E. faecalis EA utilization is only important for bacterial fitness in competition with other microbes, we characterized the colitogenic activity of E. faecalis eut mutant in the context of a simplified human microbial consortium (SIHUMI). We colonized germ-free wild type and IL-10 −/− mice with SIHUMI consortium either containing E. faecalis wild type OG1RF (SIHUMI) or E. faecalis eut (SIHUMI eut) mutant strain for a period of 16 weeks (Figure 2G). When we introduced the bacterial consortium to germ-free mice, 6 out of 7 species successfully colonized the mouse intestine, as determined by 16S rRNA gene-based quantitative qPCR ( Figure 2J). Faecalibacterium prausnitzii could not be detected by qPCR and droplet digital PCR at any time ( Figure 2J, Supplementary Figures 2B,C) in intestinal content (taken at week 4 and 16 of colonization) and feces (taken at week 2, 4, 8, and 16 of colonization). While IL-10 −/− mice colonized with the SIHUMI consortium gradually developed inflammation in the distal colon (SIHUMI histopathological score: 3.1 ± 2.3), colitis severity was significantly increased when E. faecalis OG1RF was replaced by eut mutant strain (SIHUMI eut histopathological score: 6.1 ± 1.8) (Figures 2H, I). These results surprisingly suggest protective functions of E. faecalis EA utilization in the context of chronic intestinal inflammation. Tissue pathology was accompanied by increased spleen weights in IL-10 −/− mice colonized with the SIHUMI eut consortium (Supplementary Figure 2D). As expected, all colonized wild type mice remained disease-free. The replacement of E. faecalis wild type OG1RF strain by eut mutant strain resulted in a shift of the bacterial community with a significant increase of Bacteroides vulgatus relative abundance accompanied by a decrease in Ruminococcus gnavus abundance ( Figure 2J, Supplementary Figure 5). Interestingly, the E. faecalis eut mutant strain showed an increased colonization density compared to the wild type OG1RF strain, as detected by CFU in luminal contents from colon (Supplementary Figure 2E). However, E. faecalis abundance did not correlate with the histopathology score of the distal colon (Supplementary Figure 2F). The only significant correlation between bacterial abundance and the level of colonic inflammation was detected for R. gnavus with a decreased abundance being associated with higher inflammatory scores (Supplementary Figures 2G, 5). Tissue sections from the distal colon of IL-10 −/− mice showed high expression of proinflammatory cytokines in comparison to wild type mice. Tissue expression of TNF and IL-12p40 reflected the degree of histopathology with higher expression levels in SIHUMI eut colonized compared to SIHUMI colonized IL-10 −/− mice. Expression levels of IFNγ were similar between SIHUMI and SIHUMI eut colonized IL-10 −/− mice, while IL-17 expression was decreased in SIHUMI eut colonized IL-10 −/− mice ( Figure 2K). To address the bacterial antigen-specific immune responses in colonized mice, we isolated MLN cells from IL-10 −/− mice colonized with SIHUMI or SIHUMI eut consortium and re-stimulated them with respective lysates of each of the colonizing species. Unfractionated MLN cells produced high levels of IL-12p40 when stimulated with E. coli lysate, while IL-12p40 secretion in response to all other lysates was similar to the negative control. The level of MLN activation was similar for all E. faecalis strains ( Figure 2L). Since replacement of E. faecalis wild type by the eut mutant resulted in a more severe inflammatory phenotype, we investigated whether increased inflammation was associated with an enhanced humoral immune response against specific SIHUMI species. It has been demonstrated by others that certain commensal bacteria induce serum IgA responses that confer immunity against polymicrobial sepsis (60). When we probed lysates of SIHUMI species with sera from SIHUMI or SIHUMI eut colonized IL-10 −/− mice, we observed the highest number of IgA immunoreactive bands for lysates of B. vulgatus and R. gnavus, the two most abundant species in colonized mice (Supplementary Figure 2H and Figure 2J). Similar results have been obtained for serum IgG antibodies (data not shown). The serum IgA response against SIHUMI species was not influenced by the level of inflammation, as the number and intensity of immunoreactive bands was similar for lysates probed with SIHUMI or SIHUMI eut sera (Supplementary Figure 2H).

Complex Bacterial Consortia Reprogram E. faecalis Gene Expression
Since we observed a protective effect of E. faecalis EA utilization only in a complex bacterial community and not under conditions of E. faecalis monoassociation, we aimed to gain a better understanding of E. faecalis gene expression in the presence of other microbes. We purified E. faecalis cells from the colon content of SIHUMI colonized wild type and IL-10 −/− mice using immuno-magnetic separation and performed RNA sequencing ( Figure 3A). Transcriptional profiling identified 180 significantly up-and 148 significantly down-regulated genes under inflammatory conditions (Supplementary Figure 3A). The samples clustered according to mouse genotype, as shown by PCA analysis (Supplementary Figure 3B). Among the top 20 up-regulated genes, several genes associated with arginine biosynthesis and glycerol metabolism were identified ( Figure 3B). Arginine biosynthesis genes have been shown to be up-regulated in lactic acid bacteria in response to acid stress (55) and genes involved in glycerol metabolism have been shown to be induced during E. faecalis-mediated mouse peritonitis as well as intestinal colonization (61, 62). Additionally, an E. faecalis mutant affected in glycerol metabolism showed attenuated virulence in a mouse peritonitis model (62).
Surprisingly, we observed no overlap with the top upregulated genes in E. faecalis isolated from monoassociated mice (Figures 3B, 1E). Additionally, the functional categories enriched among all differentially regulated genes were mainly different between E. faecalis monoassociation or in the context of the SIHUMI community (Figures 3C, 1D). In the SIHUMI consortium, E. faecalis response to inflammation was characterized by an induction of cellular pathways required for growth and replication, including ribosome, aminoacyl-tRNA biosynthesis, nucleotide synthesis, pyruvate and propanoate metabolism, ABC transport and fatty acid biosynthesis and metabolism. In contrast to E. faecalis monoassociation, inositol phosphate metabolism and the biosynthesis of amino acids were repressed. Quorum sensing and glycerolipid metabolism were induced, whereas phosphotransferase systems and starch and sucrose metabolism were repressed in both mono-and complex association of IL-10 −/− mice (Figures 3C, 1D). Interestingly, E. faecalis isolated from SIHUMI colonized mice showed higher expression of eut genes in healthy ( Figure 3D) and mildly inflamed ( Figure 3E) compared to inflamed animals.
In IL-10 −/− mice colonized with the SIHUMI consortium, we observed a gradient of colitis severity. Six out of 12 mice had a colitis score <3 and were defined as mildly inflamed animals, the remaining 6 mice with a colitis score >3 were defined as severely inflamed animals (Figure 3F). When comparing E. faecalis gene expression between severely and mildly inflamed IL-10 −/− mice, only 11 genes showed statistical significant regulation (Supplementary Figures 3C,D). In particular, the lipoprotein SpoIIIJ (OG1RF_12576), which has been shown to be important for bile resistance of E. faecalis (63) and a cell wall surface anchor family protein (OG1RF_10871), which is associated with E. faecalis biofilm formation (64), appeared to be interesting candidates for E. faecalis virulence (Figure 3G).
When comparing E. faecalis transcriptional response to inflammation between bacteria isolated from mono-and complex-associated mice on the gene level, the majority of differentially regulated genes were not shared. This suggests that the bacterial environment reprograms E. faecalis gene expression. Only 125 genes were shared between both colonization environments, whereas 751 genes were exclusively regulated in monoassociation and 203 genes were exclusively regulated in the SIHUMI community ( Figure 3H). In addition, the majority of shared genes showed opposite patterns of regulation. Several genes associated with fundamental cellular pathways were down regulated in E. faecalis monoassociated mice, whereas these genes were strongly up regulated in E. faecalis of the SIHUMI community ( Figure 3I). This includes genes associated with the translation apparatus (lysS, trpS2, rpmB2, rpmH, rpsT, engA), RNA synthesis (rpoZ), DNA replication (parE), citrate cycle (fdrA), peptidoglycan biosynthesis (amiD, cwlO), nucleotide metabolism (gmk2, guaD, upp2), and transport (ecfA1, tagH, ctpE, secA) ( Figure 3I). Vice versa, we observed an up regulation of several stress-response genes in E. faecalis of monoassociated mice, including genes that encode for proteases and chaperones (clpE, clpP, dnaK, dnaJ, groEL, groES), a ribosomal stress protein (rplY), a detoxifying enzyme (gloA2) and proteins associated with DNA repair (ung, dinB2) and oxidative stress resistance (gor) (55, 65). In addition, genes associated with the biosynthesis of amino acids (luxS, aroA, aroB, tyrA, pheA), transport (yidK, nhaC), transcriptional regulation (glcR, lacI) and a protein homologous to RNA polymerase subunit RpoE, were up-regulated in E. faecalis of monoassociated mice, but down-regulated in E. faecalis SIHUMI colonized mice ( Figure 3J).

Complex Microbial Consortia Interactions Reprogram E. faecalis Colitogenic Activity
We next addressed the question of whether the transcriptional differences of E. faecalis in mono-or complex-associated mice have functional consequences by applying the SIHUMI consortium with or without E. faecalis (SIHUMI-E. fae). Wild type and IL-10 −/− mice were colonized at the age of 10 weeks for a period of 4 or 16 weeks (Figure 4A). Surprisingly, the colonization of IL-10 −/− mice with SIHUMI in the absence of E. faecalis induced an aggravated inflammatory host response compared to the control. After 16 weeks of colonization, IL-10 −/− mice colonized with SIHUMI-E. fae consortium developed severe inflammation in the distal colon (score: 4.8 ± 2.4), while the degree of colitis was significantly reduced in IL-10 −/− mice colonized with SIHUMI in the presence of E. faecalis (score: 2.8 ± 2.5) (Figures 4B,C). After 4 weeks of colonization, IL-10 −/− mice showed mild pancolitis independent of the colonizing bacteria. All colonized wild type mice remained disease-free (Figures 4B,C). The histological inflammation was accompanied by increased spleen weights (Supplementary Figure 4A).
The absence of E. faecalis resulted in changes of the relative abundances of specific SIHUMI species in the colonic  content. Compared to SIHUMI colonized mice, the relative concentrations of E. coli and B. longum were slightly higher, while B. vulgatus concentrations were lower in SIHUMI-E. fae colonized mice ( Figure 4D, Supplementary Figure 6). In addition, the relative abundance of E. coli and B. longum correlated positively with the development of histopathology. E. faecalis abundance was not correlated with colitis activity (Figures 4E,F, Supplementary Figure 4D). The expression of the pro-inflammatory cytokine IFNγ was significantly increased in IL-10 −/− mice colonized with SIHUMI-E. fae, supporting the observation at the level of tissue pathology. TNF and IL-17 expression levels followed the same pattern, whereas IL-12p40 expression was not influenced by the composition of the colonizing consortium ( Figure 4G). To address the bacterial antigen-specific immune responses of colonized mice, we isolated MLN cells from IL-10 −/− mice colonized with SIHUMI in the presence and absence of E. faecalis and re-stimulated them with respective lysates of each of the colonizing SIHUMI species. After 4 weeks of colonization, the secretion of IFNγ and IL-12p40 in response to E. coli lysate stimulation was slightly increased in MLN cells isolated from mice colonized with SIHUMI. All other lysates failed to activate the MLN cell cultures (Supplementary Figures 4B,C). After 16 weeks of colonization, MLN cells produced maximal levels of IFNγ and IL-12p40 upon stimulation with E. coli lysate, although the relative luminal concentrations of this bacterium were quite low (Figures 4D,H,I). The MLN response toward E. coli lysate was even more pronounced in MLNs isolated from mice colonized with SIHUMI-E. fae consortium (Figures 4H,I), suggesting compensatory mechanisms in the absence of E. faecalis. The level of serum IgA antibodies against SIHUMI species was only slightly influenced by the composition of the bacterial consortium. When we probed lysates of SIHUMI species with sera from SIHUMI or SIHUMI-E. fae colonized IL-10 −/− mice, we observed the highest number of immunoreactive bands for lysates of B. vulgatus, R. gnavus, and E. coli. The number and intensity of immunoreactive bands against E. coli and B. vulgatus lysates were slightly increased when probed with SIHUMI-E. fae sera, indicating increased levels of IgA antibodies against these species in mice colonized with SIHUMI in the absence of E. faecalis (Supplementary Figure 4E).

DISCUSSION
In this study, we demonstrate that E. faecalis functionally adapts to chronic inflammation, supporting the hypothesis that host-related mechanisms contribute to the colitogenic activity of opportunistic pathogens. E. faecalis is a disease-relevant pathobiont and colonization of IBD-related mouse models, such as IL-10 −/− mice, supports the concept that commensal bacteria harboring pathogenic traits drive chronic inflammation in genetically susceptible but not wild type hosts (66). In addition to previous identified genes modulating the colitogenic effect of E. faecalis (28), we performed transcriptional profiling of E. faecalis under inflammatory conditions and identified 876 genes to be differentially regulated compared to normal conditions. Despite considerable up regulation of 14 genes of the eut locus in inflamed IL-10 −/− mice, the deletion of E. faecalis ethanolamine utilization ( eutVW) showed no effect on the colonization density and colitogenic activity of E. faecalis in monoassociated mice. Interestingly, E. coli responds to inflammation in monoassociated IL-10 −/− mice by the up regulation of stress response genes, but targeted deletion of these genes aggravated the inflammatory phenotype (11,12). This supports our observation that bacterial adaptation processes correlating with intestinal inflammation are not necessarily harmful for the host. In line with this, we have shown in a previous study that genes that produce two colitis-relevant structures in E. faecalis are not transcriptionally regulated in inflamed monoassociated IL-10 −/− mice (28). The measurement of luminal intestinal EA allowed us to demonstrate that E. faecalis utilizes this substrate in-vivo in wild type mice and not-inflamed IL-10 −/− mice. However, EA concentrations in inflamed IL-10 −/− mice were similar for both E. faecalis wild type OG1RF and eut mutant colonization. This is surprising as E. faecalis eut genes were massively up regulated in inflamed IL-10 −/− mice relative to wild type controls. This paradox could be explained by limitations of our model as we only analyzed gene expression and not eut protein levels. Additional studies to understand E. faecalis' eut-gene expression and in-vivo EA metabolism are needed. We provide for the first time evidence for a protective effect of bacterial EA utilization in the context of chronic intestinal inflammation. Using a simplified microbial consortium based on human strains (SIHUMI), we demonstrate that E. faecalis EA utilization attenuates colitis in mice colonized with a complex bacterial community. The replacement of E. faecalis wild type strain by a eut mutant resulted in exacerbated colitis in SIHUMI colonized IL-10 −/− mice, as indicated by increased histological inflammation, increased tissue expression of TNF and IL-12p40 and increased spleen weights. The colonization density of eut mutant was slightly increased compared to wild type E. faecalis. However, the relative abundance of E. faecalis did not correlate with the histological inflammation, suggesting that increased E. faecalis eut numbers do not explain the aggravated inflammatory phenotype. MLN cells secreted high levels of IL-12p40 in response to E. coli lysate, but the highest IgA immunoreactivity was observed for B. vulgatus and R. gnavus, the two most abundant species. The bacterial community changed when E. faecalis wild type OG1RF was replaced by eut mutant strain, with a significant increase in the relative abundance of B. vulgatus accompanied by a decrease in the abundance of R. gnavus. B. vulgatus induces colitis in HLA/B27-β 2 m rats, but does not induce inflammation in other colitis models (e.g., IL-10 −/− or IL-2 −/− mice) (67), which makes it difficult to assess whether this bacterium has colitogenic activity in a complex community in IL-10 −/− mice. Taken together, our data suggest that E. faecalis is not the driver of inflammation in the SIHUMI community, but modulates disease by interaction with co-colonizing bacteria. The catabolism of EA by E. faecalis may prevent the utilization of this metabolite by the more colitogenic species E. coli. It has been shown that EA can promote E. coli growth and virulence (35,54). However, the expression of selected E. coli virulence (fliC, fimH, ompA, ompC) and ethanolamine utilization genes (eutR, eutC, eutQ) was not affected in our model (data not shown) and also the relative abundance of E. coli was not increased in the presence of E. faecalis eut mutant.
Our data provide new insights into the role of EA utilization in microbiota-host interactions. EA plays a well-recognized role in host adaption and virulence of enteric pathogens including EHEC, Salmonella and Listeria monocytogenes (35,36,54,68,69). In contrast, we show that E. faecalis EA utilization has a protective function in IL-10 −/− mice in the presence of other resident bacterial species. These discrepancies might arise from differences in bacterial physiology and pathogenicity. The pathogen Salmonella requires alternative electron acceptors released from inflamed host tissue for EA utilization (34,70), whereas E. faecalis can grow anaerobically on EA in the absence of alternative electron acceptors (71). In addition, others have demonstrated a protective effect of EA utilization by opportunistic pathogens as well. The metabolism of EA resulted in reduced virulence together with a delayed onset of disease in a hamster model of Clostridium difficile infection (72). Furthermore, EA utilization by commensal E. coli isolates allowed them to outcompete the pathogen EHEC (41). Consistent with our data, the colonization efficiency of E. faecalis eutVW in the murine gastrointestinal tract was increased relative to the wild type strain (73). This stands in contrast to enteric pathogens, where the loss of ethanolamine utilization resulted in a competitive disadvantage (34,36,54). In Caenorhabditis elegans, an E. faecalis eut mutant was attenuated in virulence, but in this host some E. faecalis strains are regarded as a pathogenic organism resulting in the death of infected nematodes (38). In contrast, probiotic E. faecalis protect C. elegans against EHEC, underlining the importance of strain-level based experiments (74). In any case, the interesting correlation between bacterial pathogenicity and the role of EA utilization for bacterial survival and virulence awaits further investigation.
To the best of our knowledge, this is the first published study to analyze the influence of co-colonization with a bacterial consortium on gene expression and function of an opportunistic pathogen. Most importantly, we demonstrate that complex bacterial consortia interactions reprogram the gene expression profile and the colitogenic activity of E. faecalis toward a protective function. Transcriptional interactions between bacteria have already been illustrated for two members of the dominant gut bacterial phyla. Eubacterium rectale and Bacteroides thetaiotaomicron adapt their profile of utilized substrates in response to the co-colonizing species in gnotobiotic mice (75). In addition, Plichta et al. (76) demonstrated transcriptional species-specific interactions in a complex bacterial community with central metabolisms being strongly affected by coexistence with other microbes. Using immunomagnetic separation, we were able to isolate E. faecalis cells from SIHUMI colonized mice. This allowed us to study E. faecalis transcriptional response to host inflammation depending on the microbial environment. Interestingly, the majority of regulated genes were not shared between E. faecalis colonizing in monoassociation or in the SIHUMI community. This shows that E. faecalis adapts transcriptionally to the co-colonizing microbes.
In monoassociated IL-10 −/− mice, E. faecalis gene expression was characterized by an enrichment of functions important for bacterial stress adaptation. General stress-response genes, amino acid biosynthesis genes and genes involved in the utilization of alternative carbon sources were up regulated. A similar regulation pattern has been observed by others analyzing E. faecalis transcriptome during mammalian infection. The intraperitoneal transcriptome of E. faecalis in gnotobiotic mice is characterized by the up regulation of stress-response genes and a metabolic adaptation to an inflamed environment (62). Frank et al. (77) demonstrated that E. faecalis cells undergo transcriptional adaptation and exist in a stringent response state in a rabbit subdermal abscess model. The stringent response, a bacterial stress adaptation mechanism, is characterized by the repression of fundamental cellular pathways, whereas amino acid biosynthesis genes and stress-survival genes are strongly induced. In addition, metabolic rearrangements including the utilization of alternative carbon sources are a hallmark of the stringent response (65,78). Since we observe similar pathways regulated in response to inflammation and a link between E. faecalis stringent response and virulence has been demonstrated in a study by Gaca et al. (65), one could speculate that the gene expression of E. faecalis explains the colitogenic character of this bacterium in monoassociated IL-10 −/− mice. The detailed investigation of E. faecalis genes and their biological role in the pathophysiology of chronic inflammation awaits further investigation.
In co-colonization with other microbes, however, E. faecalis gene expression in inflamed mice is not primarily characterized by the induction of stress-response pathways, but rather by an enrichment of pathways important for bacterial growth and replication. Although some genes important for bacterial adaptation to environmental stress were up regulated (e.g., arginine deaminase pathway genes, opuC, phrB) (38, 55), the transcriptional profile does not indicate a stringent response state as observed in monoassociated E. faecalis. Fundamental cellular pathways like replication and translation were induced and the expression of general stress-response genes like clpE, clpP, dnaK, dnaJ, groEL, and groES were repressed (62,79).
Previous studies analyzing E. faecalis' function in chronic inflammation were conducted in monoassociated IL-10 −/− mice and revealed a colitogenic behavior (23,24,27,28). By characterizing E. faecalis as part of a microbial consortium, we are adding new knowledge to the complex interdependence of opportunistic pathogens, the genetically predisposed host and the luminal microbial environment. Colonization of IL-10 −/− mice with SIHUMI under omission of E. faecalis or in the presence of E. faecalis with defective EA utilization resulted in a more severe phenotype as indicated by the histological inflammation, spleen weights and the pro-inflammatory cytokine expression in colon tissue. This supports the hypothesis that E. faecalis provides protective mechanisms in complex consortia by the utilization of EA. A massive pro-inflammatory response of reactivated MLN cells to E. coli lysate stimulation and a positive correlation between histological score and E. coli abundance point to this bacterium and not E. faecalis as a main driver of SIHUMI mediated colitis. One could speculate that E. faecalis outcompetes E. coli or other SIHUMI strains for EA, but when E. faecalis is missing, or E. faecalis EA utilization is defective, the virulence of E. coli or other EA-responsive strains is enhanced leading to an increased immune response and inflammation.
Our data emphasize the dualistic character of E. faecalis and show that the colitogenic function of a bacterial strain is not only defined its gene repertoire, but also by co-colonizing microbes. Simultaneous colonization of IL-10 −/− mice with E. coli and E. faecalis resulted in more aggressive colitis as compared to disease induced by each species individually (80). Furthermore, Helicobacter hepaticus and Lactobacillus reuteri do not induce disease in monoassociated IL-10 −/− mice, but when both strains are used in a dual-association setup they induce severe colitis (81). The infection of Rag2-deficient and wild type mice with Klebsiella pneumoniae, Proteus mirabilis, or a combination of both strains triggered inflammation in the presence of a complex specific-pathogen free (SPF) microbiota, but K. pneumoniae and P. mirabilis dual-association of germ-free mice did not result in the development of colitis (82). Similarly, in the severe combined immunodeficiency mouse model, segmented filamentous bacteria were only effective in triggering intestinal inflammation in combination with a complex SPF microbiota (83). Vice versa, E. coli induces colitis in gnotobiotic IL-2-deficient mice, but co-association with B. vulgatus prevents colitis development (84).
In summary, we demonstrate that E. faecalis gene expression and colitogenic activity is influenced by co-colonizing microbes. A protective effect of E. faecalis EA utilization was only apparent in SIHUMI colonized mice, but not under monocolonized conditions. Additionally, we report a shift of phenotype from a colitogenic to a protective activity of E. faecalis in the IL-10 −/− mouse model depending on the bacterial environment. This model study with manageable bacterial complexity shows that monoassociation studies do not necessarily allow conclusions to be drawn about bacterial colitogenicity in IBD. This underlines that IBD is a complex disease in which both different host genes and bacterial genes play a role. The gradual development of a human-oriented complexity in carefully selected microbial consortia could be a viable way to unravel the complex microbehost and microbe-microbe interactions in disease etiology. This may contribute to a basic understanding of the pathogenesis of human IBD and have therapeutic options.

DATA AVAILABILITY
The RNA seq datasets generated for this study can be found in the SRA archive of NCBI under the accession No. PRJNA527872 and PRJNA528098.

ACKNOWLEDGMENTS
We thank Danielle Garsin for providing the E. faecalis eut VW strain. We thank Josef Sperl for the support with droplet digital PCR. We thank Maureen Bower for her support with gnotobiotic mice. We thank Lisa Holt and Nico Gebhard for their help with the samples. We thank Kurt Gedr for his support with biostatistics.