The flagellin of candidate live biotherapeutic Enterococcus gallinarum MRx0518 is a potent immunostimulant

Many links between gut microbiota and disease development have been established in recent years, with particular bacterial strains emerging as potential therapeutics rather than causative agents. In this study we describe the immunostimulatory properties of Enterococcus gallinarum MRx0518, a candidate live biotherapeutic with proven anti-tumorigenic efficacy. Here we demonstrate that strain MRx0518 elicits a strong pro-inflammatory response in key components of the innate immune system but also in intestinal epithelial cells. Using a flagellin knock-out derivative and purified recombinant protein, MRx0518 flagellin was shown to be a TLR5 and NF-κB activator in reporter cells and an inducer of IL-8 production by HT29-MTX cells. E. gallinarum flagellin proteins display a high level of sequence diversity and the flagellin produced by MRx0518 was shown to be more potent than flagellin from E. gallinarum DSM100110. Collectively, these data infer that flagellin may play a role in the therapeutic properties of E. gallinarum MRx0518.

expressed on the surface of a range of host cells including epithelial cells, endothelial cells, macrophages, dendritic cells (DCs) and T cells [17][18][19] . As a member of the TLR family, TLR5 forms an important link between the innate and adaptive immune systems and plays a role in the maintenance of gut homeostasis. TLR5 interacts with the extracellular monomeric form of bacterial flagellin of both Gram-negative and Gram-positive bacteria, leading to activation of the NF-κB signalling pathway via the adaptor protein MyD88 and the serine kinase IRAK [20][21][22] . This can lead to systemic immune responses, stimulating the production of pro-inflammatory mediators including TNF-α, IL-1β, IL-6, IL-8, IL-12 and IL-23. A study by Cai et al. has shown that expression and activation of TLR5-associated pathways were elevated in breast carcinomas 23 . Furthermore, they demonstrated that flagellin activation of TLR5 in that context resulted in the local release of pro-inflammatory cytokines and anti-tumorigenic effects 23 .
Historically, flagellin has been studied as a virulence-associated trait but is also recognized as a host colonisation factor. Bacterial flagellin is characterised by highly conserved N-and C-terminal domains (D0 and D1 domains) which have been shown to interact directly with TLR5 21,24 . The hypervariable central region of flagellin (D2 and D3 domains) varies in size and structural organisation between species, and constitutes the main antigenic region of the protein 24,25 . Antigenic variation is thought to be one mechanism by which strains evolve to evade the host immune system 26 . Serologically distinct flagellins have been identified within bacterial species and have been used to track and type isolates 27,28 .
Herein, we characterised the immunostimulatory potential of E. gallinarum MRx0518, a human commensal bacterium with demonstrated anti-tumorigenic properties 12 . This is the first study to examine the role of E. gallinarum flagellin as potential immunogens in the human gut. This work provides insights into the molecular effectors through which strain MRx0518 elicits an immunostimulatory response in human intestinal epithelial cells (IECs), macrophages and DCs and potentially exerts its anti-tumorigenic activity in vivo.

Enterococcus gallinarum MRx0518 induces a strong immunostimulatory response in vitro.
To perform an initial assessment of the immunostimulatory potential of E. gallinarum MRx0518, we measured the cytokine responses of two key innate immune cell types, THP-1-derived macrophages and monocyte-derived DCs, after stimulation with live MRx0518 cells (Fig. 1). We assessed a panel of pro-and anti-inflammatory cytokines involved in innate immunity and recruitment and activation of adaptive immune cells (IL-8, TNFα, IL-6, IL-10, IL-12p70, IL-23 and IL-1β). Both macrophages (Fig. 1A) and DCs (Fig. 1B) when unstimulated showed little to no production of the cytokines tested. As expected, a broadly consistent inflammatory profile was observed in both cell types in response to lipopolysaccharide (LPS), which was used as a pro-inflammatory response control (Fig. 1A,B). However, while LPS induced production of pro-inflammatory cytokines IL-6, IL-8 and TNFα in both macrophages and DCs, LPS-mediated expression of IL-12p70 and IL-1β was lower in DCs in comparison to macrophages (Fig. 1A,B). Compared to LPS, the cytokine production profiles were more consistent across the two cell types following E. gallinarum MRx0518 treatment (Fig. 1A,B). Both LPS and MRx0518 significantly induced IL-8 production in macrophages and DCs (p < 0.0001). MRx0518 also induced production A multiplicity of infection (MOI) of 10:1 was employed for both stimulation assays. A MOI of 1:1 was employed for IL-8 detection in macrophages, due to saturation of the assays at the MOI of 10:1. The effect of LPS on cytokines production is also shown as a positive control. The graphs represent an average of at least three biological replicates. Statistical comparisons were performed with GraphPad Prism (La Jolla, CA, USA) using ordinary one-way ANOVA analysis followed by Tukey's multiple comparison tests. Statistically significant differences with the untreated control are shown on the graphs as *(p < 0.05), **(p < 0.01), ***(p < 0.001) and ****(p < 0.0001). of all other pro-inflammatory cytokines tested in both cell types. Significantly higher levels of TNFα (p < 0.001 and p < 0.01), IL-6 (p < 0.001 and p < 0.001) IL-12p70 (p < 0.05 and p < 0.001) and IL-23 (p < 0.001 and p < 0.05) in comparison to untreated cells were observed in macrophages and DCs. IL-1β production was also induced by MRx0518 stimulation in both cells types but only reached statistical significance in macrophages (p < 0.05).
In addition to up-regulating the production of pro-inflammatory cytokines, MRx0518 treatment also induced a significant increase in IL-10 levels in comparison to LPS-treated and untreated cells. Of note, the variation observed in cytokine production by DCs (Fig. 1B) is most likely attributable to inherent donor heterogeneity. Overall, the data indicate that strain MRx0518 has a clear and potent immunostimulatory effect on host immune cells by inducing the production of a range of pro-and anti-inflammatory cytokines associated with both innate and adaptive immunity.

E. gallinarum MRx0518 treatment affects gene expression in human intestinal epithelial cells.
IECs represent one of the primary points of contact between commensal bacteria and the host in the gut. A Transwell ® co-culture system was employed to assess the transcriptional response of the mucin-secreting cell line HT29-MTX to treatment with MRx0518, using a Human Transcriptome Microarray (Fig. 2). Treatments with metabolically active cells (MRx0518 LV ), inactivated cells (MRx0518 HK ) or culture supernatants (MRx0518 SN ) were found to induce distinct host responses. Treatment with MRx0518 SN elicited the largest number of differentially expressed genes, inducing upregulation of 275 genes, 228 of which were not upregulated by other MRx0518 treatments (Supplementary Table S1). MRx0518 LV and MRx0518 HK induced the upregulation of 106 and 63 genes respectively that were not upregulated in MRx0518 SN -treated cells. Similarly, the MRx0518 SN also induced the downregulation of the largest number of genes in IECs (Fig. 2). Only 14 upregulated genes and one downregulated gene were common to all treatment groups (Supplementary Table S1). MRx0518 HK cells had the least impact on IEC transcription levels (Fig. 2). Despite MRx0518 SN treatment inducing the largest number of differentially expressed genes in IECs, pathway enrichment analysis of the transcriptomic data indicated that MRx0518 LV had the largest impact on physiological pathways, with over-representation of pathways involved in innate inflammatory responses, interferon signalling and apoptosis ( Supplementary Fig. S1). Of particular interest was the upregulation of CCL20 (~20-fold) and CXCL8 (~5-fold) in MRx0518 LV -and MRx0518 SN -treated cells, both of which play a role in immune cell recruitment (Table 1). NFKBIA and TNFAIP3, genes involved in regulating NF-κB signalling, were also significantly upregulated in MRx0518 LV -treated cells. ICAM1 was significantly upregulated in MRx0518 LV -treated cells, but not in MRx0518 SN -treated cells (Table 1). HT29-MTX cells demonstrated modest upregulation of CXCL1 expression (2.41-fold) in response to MRx0518 LV , which was not observed with other treatments.
A protein in MRx0518 culture supernatant activates NF-κB and TLR5 reporter cells. The immunostimulatory response observed in macrophages, DCs and IECs following stimulation with MRx0518 appears to be exerted, in part, through NF-κB signalling. In order to confirm activation of this signalling pathway, we examined the effect of MRx0518 treatments (MRx0518 LV , MRx0518 HK and MRx0518 SN ) on NF-κB and TLR5 reporter cell lines (Fig. 3). NF-κB activation was assessed by measuring the expression of the secreted embryonic alkaline phosphatase (SEAP) reporter gene. MRx0518 LV did not activate NF-κB reporter cells but activated TLR5 reporter cells (p < 0.0001) (Fig. 3A,B). The lack of SEAP detection in NF-κB reporter cells is likely due to the growth of MRx0518 in the cell culture media during the 22 h incubation which possibly impacted the viability of  the reporter cells rather than genuine absence of signalling activation. MRx0518 HK induced a strong response in the TLR5 reporter cells, which was slightly higher than that observed in the NF-κB reporter cells (p < 0.0001 for both cell lines in comparison to untreated cells) (Fig. 3A,B). Both reporter cell lines were activated by MRx0518 SN , to the same extent as their respective positive controls (Fig. 3A,B). Overall, MRx0518 SN was the most potent stimulant of both NF-κB and TLR5. These results, combined with the transcriptional response of HT29-MTX cells to  MRx0518 SN , prompted us to investigate the active component in this fraction. In an effort to identify the nature of molecules responsible for the observed host response, we treated MRx0518 SN with a range of enzymes (i.e. DNase, proteases and apyrase). Trypsin treatment had the greatest effect on activation of NF-κB and TLR5 reporter cells, while other enzymatic treatments had smaller effects (data not shown). TLR5 activation was completely abolished by trypsin treatment, whereas low but detectable activation remained in the NF-κB reporter cells (p < 0.0001 when compared to MRx0518 SN ) (Fig. 3C,D). These assays established that a molecule of proteinaceous nature was present in MRx0518 SN which was most likely responsible for TLR5-mediated NF-κB activation. However, residual NF-κB activation (Fig. 3C) also suggested that other molecules in MRx0518 SN that are not affected by trypsin digestion may contribute to NF-κB signalling. Flagellin is the only known TLR5 ligand, and both expression profiling and phenotypic observations indicated that MRx0518 expresses flagellin in its late log growth phase and is motile under in vitro conditions ( Supplementary Fig. S2). NanoLC-MS/MS analysis confirmed the presence of flagellin in high abundance in MRx0518 SN (Supplementary Table S2), strongly suggesting that flagellin was the molecule responsible for the observed NF-κB activation in the reporter assays.
Flagellin is responsible for the activation of NF-κB and TLR5 reporter cells. An insertion mutation in the flagellin gene (fliC) in E. gallinarum MRx0518 (strain MRx0518 fliC::pORI19) was generated. Genetic manipulations in E. gallinarum had not been described in the literature prior to this study, it was therefore necessary to develop a transformation protocol for strain MRx0518. Electrocompetent cells were successfully generated by growing bacterial cultures in sub-inhibitory concentrations of glycine 29 followed by mutanolysin and lysozyme treatments to further weaken the cell wall peptidoglycan layer (see Material and Methods for details). The fliC gene was disrupted by homology-driven insertion of the suicide plasmid pORI19 30 . Insertion of pORI19 within the fliC gene was confirmed by DNA sequencing and the non-motile phenotype of the resulting mutant strain was confirmed in vitro ( Supplementary Fig. S2). Both MRx0518 SN and MRx0518 fliC::pORI19 culture supernatant (fliC SN ) were tested in the NF-κB and TLR5 reporter assays along with culture supernatant from an additional E. gallinarum strain, DSM100110 (DSM100110 SN ) ( Fig. 4A,B). Strain DSM100110 is a murine isolate which was chosen for its highly-motile phenotype in vitro ( Supplementary Fig. S2). A significant reduction of NF-κB activation (approximately 75% compared to MRx0518 SN ) was observed for fliC SN -treated NF-κB reporter cells (p < 0.0001) (Fig. 4A). The presence of additional stimulatory molecules in fliC SN may have contributed to the observed residual activation of NF-κB signalling, as previously noted for trypsinized-MRx0518 SN (Fig. 3C). Inactivation of the flagellin gene completely abolished TLR5 activation (no observable difference with the YCFA culture medium control) and was significantly reduced in comparison to MRx0518 SN (p < 0.0001) (Fig. 4B). Interestingly DSM100110 SN induced very little activation of the TLR5 reporter cells (not statistically significant when compared to YCFA). Analysis by nanoLC-MS/MS of the protein content of DSM100110 SN revealed the absence of flagellin (data not shown), which explains the observed lack of TLR5 activation elicited by this strain (Fig. 4B). The lack of flagellin in DSM100110 SN limited our ability to determine the immunogenic potential of its flagellin in supernatant-stimulation assays. Flagellins from strains MRx0518 and DSM100110 were therefore overexpressed and purified as N-terminally his-tagged recombinant proteins ( Supplementary Fig. S3) and used to perform a dose-response assay. Both recombinant flagellins were capable of activating NF-κB and TLR5 reporter cells (Fig. 4C,D). Lower protein concentrations stimulated a strong response in TLR5 reporter cells in comparison to the NF-κB reporter cells. This was perhaps unsurprising as flagellin is known to interact directly with TLR5 20 and this TLR5 reporter cell line is expected to be 20-100 fold more responsive than cell lines that express basal levels of TLR5, as indicated by the manufacturer. Purified FliC MRx0518 showed comparable levels of activation to MRx0518 SN in both reporter cell lines, with saturation levels dropping at concentrations less than 1 ng/ml. FliC DSM100110 was weakly active in the NF-κB reporter assay at the concentrations tested ( Fig. 4C). At concentrations greater than 5 ng/ml, both recombinant proteins induced comparable responses in TLR5 reporter cells, which were not significantly different (Fig. 4D). However, at lower concentrations FliC MRx0518 was significantly more stimulatory than FliC DSM100110 at equivalent concentrations (p < 0.0001 for tests at 0.2 and 1 ng/ml and p < 0.05 for 0.04 ng/ml). The same trend was observed in the NF-κB reporter cells (p < 0.0001 for 0.8, 4 and 20 ng/ml). FliC MRx0518 and FliC DSM100110 display sequence divergence and reside in distinct clusters of a FliC phylogenetic tree. FliC MRx0518 displayed a higher capacity to stimulate both TLR5 and NF-κB than FliC DSM100110 at low concentrations. This prompted the further examination of the flagellar loci and particularly the FliC protein sequences of both strains. Organisationally-conserved 40-kb motility loci were identified in the genome sequences of strains MRx0518 and DSM100110, both of which encode 47 contiguous genes (Fig. 5). Gene organisation was similar to that of other motile enterococci 4 and both strains share 69.3% nucleotide (nt) identity (ID) over the length of the operon. Indels are present between the two loci, which result in changes to the start and stop sites of a number of homologous genes but are not predicted to result in the formation of pseudogenes. Each strain encodes a single FliC protein which are 360 amino acid (aa) and 361 aa long in MRx0518 and DSM100110 respectively and share 77.2% aa identity ( Fig. S4 and Supplementary Dataset S1). Modelling of the structure of FliC MRx0518 revealed the presence of three domains (Fig. S4), as predicted using the Phyre 2 server 31 (data not shown).
Phylogenetic analyses indicated that the FliC proteins of E. gallinarum and E. casseliflavus branch closely to the FliC proteins of motile lactobacilli ( Supplementary Fig. S5, Supplementary Dataset S1) 32,33 . In order to assess the level of diversity within the flagellin of these closely related species, the FliC sequences of 15 E. gallinarum and 3 E. casseliflavus strains derived from the 4D Pharma plc and DSMZ culture collections (Supplementary Table S3   the examined proteins, with several E. gallinarum FliC proteins displaying higher levels of sequence homology to E. casseliflavus FliC, than to each other ( Supplementary Fig. S4, Supplementary Dataset S1). The highest level of sequence divergence both within and between the E. gallinarum and E. casseliflavus FliC proteins was observed in the D2 region whereas the D0 and D1 regions were more highly conserved (Fig. S4). The regions known to be critical for TLR5 interaction in other bacterial species 21,34 were found to be conserved (residues 87-96) in all strains examined. Three distinct clusters were present within the FliC-based Maximum Likelihood phylogenetic tree shown in Fig. 6, with two well-supported E. gallinarum clusters evident (E. gallinarum_1 and E. gallinarum_2) and the majority of E. casseliflavus strains grouping together. Interestingly, strains MRx0518 and DSM100110 were resident in distinct clusters, each of which broadly represent their sources of isolation (Fig. 6).

Inactivation of the flagellin gene abolishes MRx0518 immunogenic effects in IECs.
In order to confirm the involvement of flagellin in the observed immunostimulatory effects of strain MRx0518, the impact of MRx0518 SN and fliC SN together with FliC MRx0518 recombinant flagellin on gene expression and cytokine production levels in IECs were tested. The changes in IEC gene expression following stimulation with MRx0518 SN , fliC SN and FliC MRx0518 were investigated using a targeted panel of qPCR primers, designed based on the transcriptional profiles of IECs in response to the MRx0518 treatments described above (Table 1, Fig. 7A). The expression of NFKBIA was unchanged, despite a slight upregulation being observed in the microarray-derived data (1.61-fold). MRx0518 SN significantly induced the expression of the CCL20 and CXCL8 genes (p < 0.0001 and p < 0.05 respectively compared to the YCFA treatment) (Fig. 7A), which was consistent with the upregulation previously observed. fliC SN had no effect on expression levels of the five genes tested. Co-culture of HT29-MTX cells with fliC SN did not induce the stimulatory response observed with MRx0518 SN which strongly supports the role of flagellin as a major effector of MRx0518 immunogenicity. This was further confirmed by co-culturing HT29-MTX cells with recombinant MRx0518 flagellin. The addition of FliC MRx0518 led to a significant (p < 0.05) upregulation in the expression of all five genes in the panel, with fold changes higher than those observed with MRx0518 SN (Fig. 7A). The levels of IL-8 secreted by HT29-MTX cells stimulated with MRx0518 SN , fliC SN , DSM100110 SN and FliC MRx0518 was measured in cell-free supernatants following 24 h co-culture (Fig. 7B). MRx0518 SN induced a significant release of IL-8 in IECs in comparison to cells treated with YCFA (p < 0.0001). The inactivation of the flagellin gene in the MRx0518 strain reduced IL-8 secretion to levels comparable to those observed with YCFA. Treatment of cells with recombinant flagellin strongly stimulated IL-8 secretion in comparison to the untreated and YCFA groups (p < 0.0001) but also in comparison to cells treated with MRx0518 SN (p < 0.001). In contrast, DSM100110 SN had no observable impact on IL-8 stimulation (Fig. 7B).

Discussion
Enterococcus gallinarum MRx0518 is a candidate live biotherapeutic, isolated from a healthy human faecal sample. Oral delivery of this strain has demonstrated anti-tumour efficacy in murine models of breast, lung and renal carcinomas 12 . Only a limited number of studies have characterised strains of this species in any detail, and fewer still that have examined the interactions of E. gallinarum strains with the immune system 35,36 . To begin to understand how strain MRx0518 interacts with the host, we examined its effect upon IECs, macrophages and DCs, host cells which have distinct roles in the innate immune response. DCs are capable of priming T cells at distal sites and stimulating a homing response to drive T cell accumulation to sites of inflammation. Macrophages tend to act locally to maintain homeostasis and can induce secondary activation of T cells. Both cell types come into direct contact with luminal bacteria in the gut but can also play a role in anti-tumour immunity at distal tumour sites. MRx0518 LV elicited a strong and consistent pro-inflammatory signature in both macrophages and DCs, at levels similar to or higher than those elicited by the control inflammatory stimulant LPS. MRx0518 LV significantly elevated levels of TNFα, a known regulator of IL-6 and IL-8 production, in both DCs and macrophages. Similarly, a study by van den Bogert and colleagues found that E. gallinarum HSIEG1 is also capable of inducing cytokine secretion in vitro 35 . IL-10, a cytokine well-described for its anti-inflammatory and tolerogenic effects, was also induced by MRx0518 LV treatment and can actively suppress the expression of IL-6, IL-12, IL-1β and TNFα. Given the elevation of both pro-and anti-inflammatory cytokines in response to MRx0518 treatment, it is noteworthy that IL-10 and IL-6 are reciprocal cytokines, both utilising the activity of transcription factor STAT3 to alter cellular responses with broadly opposing effects 37 . IL-10 is not a pan-inhibitory cytokine of inflammatory responses; it is known to activate and increase CD8α cytotoxic capacity which may be significant for anti-tumour responses 38,39 . High levels of IL-8 production were observed in macrophages and DCs in response to MRx0518 LV exposure. Additionally, MRx0518 LV , MRx0518 SN and purified flagellin all induced expression of CCL20 and CXCL8 in HT29-MTX cells; the products of which are implicated in the recruitment of immune cells and the subsequent activation of the adaptive branch of the immune system. Similarly, Salmonella-derived flagellin has been shown to stimulate expression of the CCL20 gene in Caco-2 cells 40 and E. coli flagellin has been shown to induce secretion of IL-8 and CCL20 in HT29-19A and Caco-2 cells 41 .
Flagellin is a potent immunostimulant which acts through TLR5 and has been exploited in recent years for its capacity as a vaccine adjuvant and its anti-tumorigenic efficacy 13 . Purified MRx0518 flagellin showed significant activation of TLR5 in reporter cells and proved more potent than DSM100110 flagellin at equivalent nanomolar concentrations. Given the body of work that is emerging regarding the role of TLRs and their associated ligands in anti-cancer therapies, the potential contribution of flagellin to MRx0518 anti-tumour activity warrants further investigation. The administration of S. Typhimurium flagellin has been shown to reduce tumour growth and cell proliferation in colon and breast cancer cells 23,42 . An elegant study by Cai et al. demonstrated that 80% of breast carcinoma tissues tested were found to be positive for TLR5 expression and that TLR5-signalling was also upregulated in breast carcinomas 23 . They concluded that flagellin-mediated TLR5 activation is involved in modulation of the tumour microenvironment and mediates its anti-tumorigenic effect through pro-inflammatory Activation of TLRs on the surface of tumours may require transport or delivery of flagellin to distal tumour sites which could be achieved through translocation of the bacteria or their components from the gut. Manfredo-Viera et al. recently demonstrated that E. gallinarum was able to translocate from the murine gut to induce an autoimmune response in immunocompromised mice 36 . Translocation of another enterococcal species E. hirae to secondary lymphoid organs has been shown to enhance efficacy of a chemotherapeutic agent 9 . Additionally, E. gallinarum was found to be overly abundant in patients who responded to treatment with anti-PD1 11 , suggesting a potential role for this species in patient responsiveness to ICI treatments. Studies  Table S3) were aligned with MUSCLE 59 . The evolutionary history was inferred by using the maximum likelihood method based on the Le_Gascuel_2008 model 61 , using MEGA7 software 60 . Statistical support (above 60%) was estimated with bootstraps and is indicated at branch nodes. Well-defined clades are indicated. The type strain Lactobacillus mali DSM20444 was used as an outgroup for analyses. The origins of strains are indicated where the information was available on the National Center for Biotechnology Information (NCBI) website (https://www.ncbi.nlm.nih.gov). are currently ongoing to investigate the ability of MRx0518 to translocate from the gut to extra-intestinal sites. Testing the translocation potential of MRx0518-derivatives, including flagellin, is also underway using FliC MRx0518 -directed antibodies.
Inter-and intra-species comparative analysis of the FliC proteins of E. gallinarum and E. casseliflavus indicate that the D0 and D1 domains are highly conserved while the majority of variability lies within the D2 domains, as observed for flagellin of other species 25 . The sequence divergence displayed in the FliC sequence in E. gallinarum is comparable with that of C. difficile 43 and E. coli 44 and is less than that reported for P. aeruginosa 45 and B. thuringiensis 46 . Antigenic variation in the FliC sequence may contribute to differences in immunogenic potential of E. gallinarum strains. However, it is yet to be determined to what extent the variance observed in the D2 domains of FliC MRx0518 and FliC DSM100110 contributes to the immunogenic profiles of these strains.
Inactivation of flagellin in MRx0518 resulted in complete abrogation of TLR5-mediated activation of NF-κB. However, some residual activity remained in the NF-κB reporter cells when treated with fliC SN . This suggests the involvement of additional or complementary bacterial effectors present in culture supernatants. Of particular interest was the identification of enolase as the most abundant protein in the MRx0518 SN (Supplementary  Table S2) as enolase has been shown to play a role in host-interactions in lactic acid bacteria, through plasminogen binding 47 . Small molecules such as ATP and CpG DNA, can act synergistically with flagellin to trigger host immune responses [48][49][50] . The potential contribution of these molecules to the observed immunogenic effects of MRx0518 warrants further investigation. Taken together, these data demonstrate that E. gallinarum MRx0518, and more specifically its flagellin, is a strong immunostimulant of both immune and intestinal epithelial cells. Importantly FliC MRx0518 displays higher potency than FliC DSM100110 . The extent of the activity of MRx0518 flagellin in vivo remains to be determined. In this context, MRx0518 derivatives are currently being investigated in murine cancer models in order to shed light on their influence on the previously established therapeutic effect of E. gallinarum MRx0518.

Material and Methods
Bacterial strains, plasmids and culture conditions. E. gallinarum strains were routinely cultured in  Preparation of bacterial fractions for co-culture assays. Late log phase bacterial cultures were centrifuged at 5000 x g for 5 min at room temperature to generate bacterial fractions. Pelleted cells were washed once in phosphate-buffered saline (PBS) (Sigma-Aldrich) and resuspended in antibiotic-free cell culture media to the appropriate dilution (live fraction, MRx0518 LV ). Culture supernatants were harvested and filtered through a 0.22 μm pore size filter and diluted in water to provide equivalents for the live fraction described above (supernatant fraction, MRx0518 SN ). Bacterial cultures were heat-inactivated for 40 min at 80 °C and prepared as described above for the live fraction (heat-killed fraction, MRx0518 HK ). Viable cell counts were determined by spread plating. When required, culture supernatants were digested with 500 μg/ml trypsin or an equivalent volume of Hank's balanced salt solution (HBSS) (Thermo Fisher Scientific, Waltham, MA, USA) as a mock digestion control for 1 h at 37 °C, followed by inactivation with 10% (v/v) foetal bovine serum (FBS) (Sigma-Aldrich).

Immortalised and primary cells stimulation.
THP-1 cells were differentiated into macrophages by the addition of 5 ng/ml phorbol 12-myristate 13-acetate (PMA) (Sigma-Aldrich) to the culture media for 48 h. Cells were plated in 96-well plates (200,000 cells/well) in cRPMI without PMA, antibiotics and FBS and incubated for 3 h. Treatments (live bacteria at a MOI of 10:1 or a MOI of 1:1 for IL-8 detection as saturation was obtained with the 10:1 MOI) and controls (50 ng/ml LPS or PBS) were then added and incubated for 1 h at 37 °C under anaerobic conditions. Culture medium was then replaced with cRPMI and incubated for 24 h under standard growth conditions. Cell-free supernatants were then harvested, centrifuged for 3 min at 10,000 x g at 4 °C and stored at −80 °C for cytokine detection. Human PBMCs, obtained from STEMCELL Technologies (Vancouver, Canada) from healthy donors, were used to isolate primary monocyte populations by negative selection using a Human Monocyte Isolation kit. Monocytes were then differentiated into immature dendritic cells by incubation with 20 ng/ml recombinant human IL-4 and 50 ng/ml recombinant human GM-CSF for 8 days at 37 °C in a 5% CO 2 atmosphere in cRPMI supplemented with 55 μM 2-mercaptoethanol. Immature dendritic cells were recovered, washed, resuspended in cRPMI medium without antibiotics and plated in 96-well plates (200,000 cells/ well). Treatments (live bacteria at a MOI of 10:1) and controls (100 ng/ml LPS or cRPMI) were added to the cells and incubated for 1 h at 37 °C under anaerobic conditions. Culture medium was then replaced with cRPMI and incubated for 17 h under standard culture conditions. Cell-free supernatants were then harvested, centrifuged for 3 min at 10,000 x g at 4 °C and stored at −80 °C prior to cytokine detection.
Cytokine quantification. Cytokine quantification was conducted using a ProcartaPlex multiplex immunoassay following the manufacturers recommendations (Thermo Fischer Scientific, Waltham, MA, USA).
Briefly, 50 µl of cell-free co-culture supernatants (CFS) were used for cytokine quantification using a MAGPIX ® MILLIPLEX ® system (Merck, Darmstadt, Germany) with the xPONENT software (Luminex, Austin, TX, USA). Data was analysed using the MILLIPLEX ® analyst software (Merck) using a 5-parameter logistic curve and background subtraction to convert mean fluorescence intensity to pg/ml values. Motility assays. Motility in vitro was assessed using BBL ™ Motility Test Medium supplemented with 0.005% (w/v) 2,3,5-triphenyltetrazolium chloride (BD, Sparks, MD, USA). In brief, a fresh colony was stab-inoculated in 20 ml equilibrated media and incubated for 48 h at 37 °C in anaerobic conditions. All assays were performed in triplicate.

Protein identification by nanoLC-MS/MS. Sample preparation and protein identification by LC-MS/
MS were performed by Aberdeen Proteomics (University of Aberdeen, UK). In brief, 40 ml culture supernatants were concentrated down to 0.5 ml and washed with ultrapure water. Proteins were precipitated using a ReadyPrep 2-D Cleanup Kit (Bio-Rad) and resuspended in 100 µl 50 mM ammonium bicarbonate. Proteins were incubated with porcine trypsin (Promega, Madison, WI, USA) for 16 h at 37 °C and the resulting supernatants were dried by vacuum centrifugation and dissolved in 0.1% trifluoroacetic acid. Peptides were further desalted using µ-C18 ZipTips (Merck). Peptides were then eluted into a 96-well microtiter plate, dried by vacuum centrifugation and dissolved in 10 µl LC-MS loading solvent (2% acetonitrile, 0.1% formic acid). Peptides were separated and identified by nanoLC-MS/MS (Q Exactive hybrid quadrupole-Orbitrap MS system) (Thermo Fischer Scientific) using a 15-cm PepMap column, 60-minute LC-MS acquisition method and an injection volume of 5 µl. Data analysis was performed with Proteome Discoverer (Thermo Fischer Scientific) and the workflow included the Mascot Server as the search engine with the following parameters: enzyme = trypsin, maximum mixed cleavage sites = 2, precursor mass tolerance = 10 ppm, dynamic modifications = oxidation (M), static modifications = carbamidomethyl (C). Identified peptides were matched against a strain-specific protein sequence database, which was constructed based on the sequenced genome of E. gallinarum MRx0518 (3068 sequences).

Recombinant flagellin expression and purification. E. gallinarum MRx0518 and DSM100110
full-length fliC genes were amplified by PCR using primer pairs DC022/DC023 and DC024/DC025, respectively (Supplementary Table S3). Gene products were then cloned into the pQE-30 vector (Supplementary Table S3

Comparative analysis of the flagellar loci of E. gallinarum strains MRx0518 and DSM100110.
Nucleotide alignments were generated using a local BLAST v 2.7.1+ installation which were then visualised and analysed for gene conservation and sequence synteny using EasyFig. 2.2.2 58 .
Phylogenetic analyses. FliC protein sequences were downloaded from the NCBI protein database or were derived from sequence data available for the strains outlined in Table S5, using BLASTP-based homology searches against the FliC MRx0518 sequence. Protein sequences were aligned using MUSCLE 59 and evolutionary analyses were conducted in MEGA7 60 . Phylogenies were inferred using the Maximum Likelihood method based on the Le_Gascuel_2008 model 61 . A discrete Gamma distribution was used for the multispecies FliC tree (Fig. S5), to model evolutionary rate differences among sites (5 categories (+G)). The rate-variation model allowed for some sites to be evolutionarily invariable ([+I]). The trees with the highest log likelihood are displayed and the reliability of the groups were evaluated by bootstrap testing with 1,000 re-samplings. The FliC of Lactobacillus mali DSM20444 (accession number KRN11091.1) was used as an outgroup during E. gallinarum and E. casseliflavus interspecies phylogenetic analyses.
Generation of an E. gallinarum MRx0518 flagellin gene insertion mutant. The flagellin insertion mutant was created using the non-replicative plasmid pORI19 30 (Supplementary Table S3). An internal fragment of E. gallinarum MRx0518 fliC gene was amplified using primers DC020 and DC021 (Supplementary Table S3) and cloned into pORI19. Restriction enzymes and Quick Ligase (New England Biolabs, Ipswich, MA, USA) were used according to the manufacturer's instructions. This construct was propagated in E. coli EC101 by chemical transformation (Supplementary Table S3) and isolated using the Genopure Plasmid Maxi Kit (Roche Diagnostics, Basel, Switzerland) from a 500-ml culture. Isolated plasmid DNA was concentrated using 0.3 M sodium acetate pH 5.2 and ethanol down to 20 μl. A protocol was developed to prepare E. gallinarum MRx0518 electrocompetent cells, which was adapted from a previously published method 29 . In brief, E. gallinarum MRx0518 was grown for 18 h in GM17 broth, supplemented with 0.5 M sucrose and 3% (w/v) glycine (Sigma-Aldrich). Cells were then washed twice with 0.5 M sucrose and 10% (v/v) glycerol and treated with 10 μg/ml lysozyme and 10 U/ml mutanolysin (Sigma-Aldrich) for 30 min at 37 °C. E. gallinarum MRx0518 cells were then transformed by electroporation with 10 μg of plasmid DNA and recovered in BHI broth before plating on selective BHI agar. Plasmid insertion was confirmed for successful transformants by PCR amplification and sequencing (GATC Biotech, Konstanz, Germany) using primers listed in Supplementary Table S3. In vitro motility of the flagellin insertion mutant was assessed as described for strain MRx0518.
Gene expression profiling by qPCR. HT29-MTX cells were cultured in Transwell ® and incubated with bacterial culture supernatant (MOI 100:1 equivalent) or recombinant flagellin (1 µg/ml) for 24 h. Mammalian RNA was isolated as described above. cDNA was synthesized using a High-Capacity cDNA Reverse Transcription Kit (Thermo Fischer Scientific). qPCR analysis was carried out using the primers detailed in Supplementary