Identification of a muropeptide precursor transporter from gut microbiota and its role in preventing intestinal inflammation

Significance

Muropeptides are bacterial-derived cell wall components known to trigger immune response in the host. To date, the production of these compounds was exclusively linked to the activity of peptidoglycan hydrolases through the mechanism known as “peptidoglycan recycling”. The present study highlights an alternative and performant way, implemented by some bacteria, to release important amount of muropeptide precursor that can directly interact with the host. The mechanism described here contributes to the understanding of the complex relationship host–microbiota and suggests that possible mechanisms of compound delivery or accumulation are the first steps for their development as therapeutics.

Abstract

The gut microbiota is a considerable source of biologically active compounds that can promote intestinal homeostasis and improve immune responses. Here, we used large expression libraries of cloned metagenomic DNA to identify compounds able to sustain an anti-inflammatory reaction on host cells. Starting with a screen for NF-κB activation, we have identified overlapping clones harbouring a heterodimeric ATP-binding cassette (ABC)-transporter from a Firmicutes. Extensive purification of the clone’s supernatant demonstrates that the ABC-transporter allows for the efficient extracellular accumulation of three muropeptide precursor, with anti-inflammatory properties. They induce IL-10 secretion from human monocyte-derived dendritic cells and proved effective in reducing AIEC LF82 epithelial damage and IL-8 secretion in human intestinal resections. In addition, treatment with supernatants containing the muropeptide precursor reduces body weight loss and improves histological parameters in Dextran Sulfate Sodium (DSS)-treated mice. Until now, the source of peptidoglycan fragments was shown to come from the natural turnover of the peptidoglycan layer by endogenous peptidoglycan hydrolases. This is a report showing an ABC-transporter as a natural source of secreted muropeptide precursor and as an indirect player in epithelial barrier strengthening. The mechanism described here might represent an important component of the host immune homeostasis.

The human gut represents the perfect habitat for a complex and dynamic community of bacteria that constantly interact with the host. The microbiota contributes to many physiological processes of the host, and perturbation of the host’s microbiota, called dysbiosis, has been associated with inflammatory bowel diseases (IBDs), metabolic diseases, neuroimmune disorders, and cancer (1–3). A large number of reports from the literature describe the immunomodulatory properties of bacteria-derived metabolites (4, 5), with butyrate and other short-chain fatty acids (SCFAs) being among the most investigated compounds. Derived from the degradation of dietary fibers by microbial carbohydrate-active enzymes, combined with an intense fermentation process (6), SCFAs signal through cell surface G-protein coupled receptors and act on a plethora of cellular events linked to the maintenance of epithelial integrity, glucose homeostasis, and adipogenesis (7–9). Other relevant bacterial metabolites as tryptophan-derived compounds were highlighted for their role on gastrointestinal functions (10). Finally, polysaccharide A from Bacteroides fragilis was also extensively studied for its ability to reduce experimental colitis in mice through induction of IL10-secreting CD4 and CD8 T cells (11). Besides that, a huge number of bioactive compounds still wait to be discovered, and while metaproteomic and metabolomic analyses provide qualitative and quantitative measurements of all proteins and metabolites from a given ecosystem, library screening helps to identify new compounds that could be useful for treating diseases (12). Functional metagenomics is such kind of approach. Emerged 20 y ago, functional metagenomic refers to the peculiar characteristic and growing conditions of bacteria in the gastrointestinal tract (GIT) (13, 14). In the GIT, the overall conditions for microbiota to survive and proliferate greatly change, mainly due to the lowering of oxygenation and the pH variation. The colon with a concentration of 10¹² to 10¹³ cfu/mL is the largest source of bacteria, predominantly anaerobes (Bacteroides, anaerobic Streptococci, and clostridiales), difficult to cultivate in vitro. Investigations on these bacteria were made possible via metagenomic sequencing allowing to generate catalogues of genes and libraries of metagenomic clones from either healthy subject or specific cohorts (15–17). For functional screening, metagenomic libraries were preferably generated in Escherichia coli, each E. coli clone harboring a large DNA fragment of 30 to 40 kb (18, 19).

Here, we show results on one metagenomic clone (clone F4) identified from the screening of a library generated from human gut microbiota using a NF-κB reporter system. NF-κB-dependent promoter regulation was selected as readout due to its key role in host response to microbes and the peculiar characteristic to induce both inflammatory response and tissue protection (20). When deregulated, NF-κB activation is associated to aberrant T cell activation, hence to autoimmune and inflammatory responses (21, 22). NF-κB complete inhibition severely compromises epithelial integrity and triggers spontaneous chronic intestinal inflammation in mice (23, 24). Therefore, proper regulation of NF-κB activation at epithelial interfaces is crucial for the maintenance of tissue homeostasis and for efficient host defense against environmental insults.

We provide data showing that the F4 clone, initially selected as an NF-κB inducer, encodes a heterodimeric transporter leading to the accumulation and secretion of muropeptides precursor.

Muropeptides are peptidoglycan-derived fragments sensed through the innate intracellular receptors NOD1 and NOD2 (25, 26). Mutations in these receptors are associated to increased predisposition for the development of IBD such as ulcerative colitis or Crohn’s disease, respectively (27, 28). Since the 70s, muropeptides have been extensively studied and used as adjuvants for boosting the potency of drugs and vaccines (29). Besides, muropeptides were also proven to contribute to the reduction of inflammation in both DSS and TNBS mouse models of colitis (30, 31). Interestingly, the anti-inflammatory effect observed by Fernandez and coworkers was due to a Lactobacillus salivarius Ls33 muropeptide inducing local IL-10 secretion via NOD2 recognition and generation of CD103⁺ dendritic cells (DCs) and CD4⁺Foxp3⁺ regulatory T cells. More recently, a novel peptidoglycan hydrolase (SagA) from the commensal bacterium Enterococcus faecium has been described for its ability to generate small muropeptides that are effective against Salmonella typhimurium and Clostridioides difficile infections (32). Altogether, this supports the concept that some muropeptides may act on immune and epithelial homeostasis in the gut.

In the present study, we describe the identification of a muropeptide precursor transporter whose expression contributes to reduce epithelial damage upon inflammatory stress in ex vivo and in vivo models. We also suggest that possible mechanisms of compounds delivery or accumulation are the first steps for their development as therapeutics.

Results

Identification of F4 Clone through High-Throughput Screening of a Metagenomic Library.

The F4 clone has been identified after the screening of 5,000 clones from a metagenomic library generated from fecal samples of healthy subjects. Total lysates of metagenomic clones were screened on HT-29 cells stably transfected with a NF-κB reporter system described previously (19). Comparative analysis was performed plotting reporter gene activity (normalized to that of control E. coli Epi300) versus growth level for each metagenomic clone (SI Appendix, Fig. S1A).

The whole fosmid insert was 41 kb long. It was aligned on the complete set of more than 200,000 Metagenomes Assembled Genomes (MAGs) from Almeida et al. (17). The fosmid insert alignment was split on two MAGs (GUT_GENOME259254 and GUT_GENOME139568) with more than 97% identity. Both MAGs are annotated as being part of the Acutalibacteraceae family. Putative transcription units and operons were detected using SoftBerry’s software FGENESB. Putative open reading frames (ORFs) were determined using GeneMark for prokaryotes. As highlighted by BlastP analysis on predicted ORFs, F4 sequence harbors multiple transporter and permease systems grouped in several operons (SI Appendix, Fig. S1 B and C).

Gene(s) responsible for biologic activity were identified through systematic mutagenesis by random transposon insertion on the F4 fosmid. Mutated DNA was retransformed into E. coli, and a small library of 184 potential revertants was screened again on the NF-κB reporter system. Twenty-three transposed clones out of the 184 were revertants for the activity. Interestingly, 22/23 mutants were found in two genes corresponding to ORFs 3 and 4 and encoding a heterodimeric ABC-transporter (SI Appendix, Fig. S1C). The last mutant was a conserved hypothetical lipoprotein. As the ORF3 was the most affected by transposon insertions, the revertant F4D5, carrying a mutated ORF3 at position +203 (first Methionine considered as +1, SI Appendix, Fig. S14), was selected for further characterization studies. Proteins forming the heterodimer contain both ATPase and permease components of an ABC multidrug transport system and are related to the MsbA superfamily whose members are involved in lipopolysaccharide synthesis through the export of lipid A (33). As determined by BlastP, the most similar sequences (proteins) were from Acutalibacter species and a Firmicutes bacterium CAG:94 with 99% and 96% identity, respectively.

As indicated before, the fosmid insert alignment is split on two different MAGs from the same taxonomical family. Given that the MAGs are not complete genome sequences from bacterial isolates but are in silico reconstruction of complete and partial bacterial genomes performed using the sequencing data from multiple metagenomic samples, there is indeed the possibility that this insert may be originating from an unidentified bacterial strain, which may still be not completely covered or characterized. To further explore this hypothesis, we analyzed the mapping of the metagenomic sequencing data from 169 human gut microbiota samples, with a focus on the portion of the fosmid insert which includes the split region between the two MAGs (Fig. 1A). Even if the number of mapped reads is lower in this region, the overall coverage remains homogeneous, supporting the hypothesis for a still unidentified bacterial strain.

Fig. 1.

(A) Reads sequences of 169 human gut metagenomic samples mapped to the F4 sequence with a focus on the portion of the fosmid insert which includes the split region between the two MAGs (GUT_GENOME259254 and GUT_GENOME139568). The red arrow indicates the separation point between the two MAGs. (B) Distance tree view of the top 100 match using blast pairwise alignment with the ABC heterodimer as query.

The portion of the insert including the two genes encoding for the ABC-transporter aligns on the GUT_GENOME139568 MAG, which is annotated on the genus Acutalibacter.

Moreover, sequences corresponding to ORF3 and ORF4 were used to derive the corresponding protein sequences and to perform a similarity search and phylogenetic analysis using the sequences available in the UniRef90 database. A distance tree was created using the results of this similarity search and loaded into the ITOL web tool (https://itol.embl.de/) to generate a dendrogram (SI Appendix, Fig. S2). As it is possible to observe, the two proteins forming the ABC-transporter are on two opposite branches of the tree, which is expected considering that the two genes have different sequences and functions. Both are anyway very close in the dendrogram to protein sequences originating from an unannotated Firmicutes bacterium. The same results were obtained when the two proteins forming the heterodimer were aligned together; the ABC heterodimer is ranged in a separate and distant cluster of the distance tree view (Fig. 1B).

The NF-κB-Dependent Activity of F4 Clone Is Mediated by a Heterodimeric Transporter.

F4 activity was found in the supernatant of an overnight culture. Under the same conditions, the supernatant from the F4D5 transposed clone was unable to induce any NF-κB-dependent activity in HEK293 Null NF-κB-SEAP reporter cells (Fig. 2A). This cell line was chosen to follow the multiple purification steps, necessary for compound identification, because of the low Pattern Recognition Receptors (PRRs) expression and acceptable basal activity induced by E. coli MAMPs. Preliminary investigations on F4 clone highlighted a marked instability of F4 DNA resulting in complete loss of biologic activity after few cycles of expansion in liquid or agar culture. We hypothesized that this loss of activity was a consequence of the natural transposition/inactivation of the heterodimer by the transposition elements present in the next operon.

Fig. 2.

The activity of F4 clone is mediated by the ABC heterodimeric transporter by a MyD88-independent pathway. NF-κB activity in (A–C) HEK Null- or (D and E) THP-1-NF-κB/SEAP reporter cells induced by (A) supernatant of F4 clone and of the transposed F4D5 clone; (B) supernatants from E. coli Epi300 transformed with single (p-ORF3 and p-ORF4) or both ORFs of the ABC heterodimer (pBAD-ABC in the Figure); (C) supernatants from F4D5:ABC (complemented clone), pBAD-ABC (ABC-transporter cloned in pBAD30 vector), and F4D5-pBAD (transposed clone transformed with empty pBAD30 vector). (D) NF-κB activity induced on THP-1 and (E) THP-1 MyD88^−/− NF-κB/SEAP reporter cells by LPS (10 ng/mL), TNF (10 ng/mL), Flagellin (100 ng/mL), F4, F4D5, and Epi300 supernatants. Supernatants were all used at 10% v/v. Data (A–C) are pooled from biological triplicates, and error bars represent SEM (*P < 0.05, **P < 0.01, and ***P < 0.001). Data (D and E) are shown as mean ± SD of triplicate measurements of a representative of two independent experiments.

Purification of the active compound from F4 supernatant was not recommended under these conditions, and subcloning of the ABC-transporter was performed to investigate whether the transporter was itself responsible for the biologic activity and whether the gene complementation was able to restore the activity of the F4D5 transposed clone.

To that end, the DNA coding for the two ORFs forming the ABC-transporter (ORF3 + ORF4) was amplified from the original fosmid library, cloned into the arabinose-induced pBAD30 vector (to get the plasmid pBAD-ABC) and verified by sequencing (SI Appendix, Fig. S3C). Also, sequences corresponding to each of the two ORFs were cloned individually in pBAD30 to get p-ORF3 and p-ORF4 (SI Appendix, Fig. S3 D and E). As shown in Fig. 2B, only supernatant from pBAD-ABC, but not that from p-ORF3 or p-ORF4, was able to activate the NF-κB reporter system on HEK293 cells, thus confirming the absolute necessity for the concomitant expression of the two genes for proper activity (SI Appendix, Fig. S3H).

Plasmid pBAD-ABC was used to complement the F4D5 revertant to get the F4D5:ABC clone (SI Appendix, Fig. S3G). Supernatants from both pBAD-ABC and F4D5:ABC clones were tested on the NF-κB reporter HEK293 cells (Fig. 2C). As expected, both supernatants were able to induce the NF-κB activity, but higher signals were measured with the F4D5:ABC-derived supernatant, suggesting that other genes in the metagenomic insert could be important for the total activity observed in vitro or that the presence of specific PRR was necessary to get full activity.

To clarify this point, F4 supernatant was tested on NF-κB reporter THP1 cells (a monocytic cell line expressing the majority of PRR) and on THP1 cells deficient in MyD88, the adapter protein involved in the TLR/IL1R signalling pathway. F4 supernatant activated NF-κB reporter system on both THP1 cell lines while LPS, Flagellin, Epi300, and F4D5 supernatants failed to activate it in MyD88^−/− -THP1, indicating that signalling did not involve MyD88 (thus most TLRs) (Fig. 2 D and E). The ORF encoding the putative lipoprotein, also identified through F4 fosmid transposition, was cloned into pBAD30 vector. Various L-Arabinose concentrations were tested to induce the Ara promoter and produce the protein. No biologic activity was associated with supernatants obtained from the recombinant E. coli strain carrying the lipoprotein DNA; thus, no further investigations on this protein were undertaken.

However, additional genes, other than the heterodimeric transporter, could be involved to get the full activity of F4 clone. For this reason, the F4D5:ABC (complemented F4D5 clone) was used for all the experiments described below and F4D5-pBAD (transposed clone transformed with the empty vector) was used as the negative control.

F4 Biological Activity Is Mediated by Several Small-Size Compounds.

Preliminary characterization tests were made on F4D5:ABC supernatant to understand whether the active compound was a protein (coded by ORFs 3+4) or a metabolite that could result from enzymatic or transporter activity of encoded proteins. The use of ultrafiltration, with a range of molecular weight cutoff filters between 1 and 50 kDa, allowed to identify a small compound below 3 kDa as the active one secreted by the F4 clone, thus excluding the direct effect of the dimeric transporter on NF-κB activity.

Preliminary High-Performance Liquid Chromatograpy (HPLC) fractionation of a prepurified supernatant using a Hypercarb column allowed the identification of three NF-κB positive fractions, F13, F21, and F24 (SI Appendix, Fig. S4A). We focused on F13 and F24 purification since F21 did not show activity on another readout (see below).

Active Compounds from F4 Supernatant Are Muropeptide precursor.

For the identification of the active compounds present in F13 and F24, a scale-up for both production and purification processes was necessary (SI Appendix, Figs. S4–S8 and SI Material and Method). Three compounds, belonging to the muropeptide’s family, were identified as the ones responsible for the biologic activity of F4 clone. Fraction 24 contains a mixture of two closely related structures of MW 1051.7 and 1122.8 g/mol, respectively, corresponding to Uridine Diphosphate-N-Acetylmuramic-L-Ala-gamma-D-Glu-mDAP (UDP-M-TriDAP) and Uridine Diphosphate–N-Acetylmuramic–L-Ala–D-isoGlu–meso-DAP–D-Ala (UDP-M-Tetrapeptide) (SI Appendix, Figs. S5H and S6H), while N-Acetylmuramic-L-Ala-gamma-D-Glu-mDAP -monophosphate (M-TriDAP-MP, MW of 745.2 g/mol) present in fraction 13 proved to be a fragment of compound UDP-M-TriDAP having lost the uracil moiety (SI Appendix, Fig. S8H).

As determined by LC-MS analysis, the three precursor were released in relatively high amount (μM range) in the supernatant of F4D5:ABC clone but not of F4D5-pBAD demonstrating the key role of the F4 transporter in the export of these compounds outside the cells instead of being incorporated into the de novo synthesized peptidoglycan. (SI Appendix, Fig. S9).

UDP-M-TriDAP and M-TriDAP-Monophosphate Are Dual NOD1 and NOD2 Agonists.

As the F4 active compounds correspond to muropeptide precursor, HPLC fractions enriched in these molecules were tested on NF-κB reporter HEK cells overexpressing either NOD1 or NOD2 receptors and compared to commercial muropeptides, TriDAP (NOD1 agonist) and MDP (NOD2 agonist) tested between 0.01, and 1 µg/mL. As shown in Fig. 3A, F13, enriched in M-TriDAP-MP, slightly activated NOD1. Similar activation was observed for TriDAP at 0.1 µg/mL. Higher activation was observed upon F24 incubation, enriched in both UDP-M-TriDAP and UDP-M-Tetrapeptide, with a fivefold increase in respect to the negative fraction and similar effect as TriDAP at 1 µg/mL.

Fig. 3.

F4D5:ABC-derived active fractions signal through NOD1 and NOD2 receptors and induce IL-10 secretion from moDC. NF-κB activity (A and B) induced by F13 and F24 fractions in (A) HEK-NOD1 NF-κB-SEAP and (B) HEK-NOD2 NF-κB-SEAP reporter cells. Cytokine secretion (C) IL-10 and (D) IL-8 by human moDC following stimulation with F13 and F24 fractions. TriDAP and MDP were used at 0.01 to 0.1 to 1 µg/mL. TNF (10 ng/mL) and PHA (3 µg/mL) were used as positive controls. F13 and F24 fractions were from purification of F4D5:ABC (blue bars) and F4D5-pBAD (blue-striped bars) supernatants. Results are shown as mean ± SD of triplicate measurements of a representative of two independent experiments.

As expected, no NF-κB activity was induced by Tri-DAP on NOD2 overexpressing HEK cell line (Fig. 3B), while both F13 and F24 induced a strong NF-κB-dependent signal that was, respectively 5 and 3 times higher than those induced by the respective negative fractions.

To properly evaluate the effectiveness of F4-derived precursor, chemical synthesis of the three molecules was considered. M-TriDAP-MP was selected as the most suitable compound for chemical synthesis as it lacks the UDP moiety. Pure (≥99%) M-TriDAP-MP was successfully synthetized as a diastereoisomer mixture (mixture of d-DAP and l-DAP moieties) although at a very low yield (2%).

The NF-κB-inducing activity of M-TriDAP-MP, used between 10 and 0.01 µM, was compared to that of commercial M-TriDAP (described as having dual NOD1/2 activity) used at the same concentrations. The two compounds behaved similarly, inducing a limited activity in the NOD1 expressing cell line (SI Appendix, Fig. S10A) and a strong dose-dependent stimulatory effect on the NOD2-HEK cells (SI Appendix, Fig. S10B). As expected, no NF-κB-dependent activity was measured using the HEK Null NF-κB/SEAP reporter cells (SI Appendix, Fig. S10C). UDP-M-TriDAP and UDP-M-TetraDAP (>90%) were obtained following F24 fractionation and tested on the NOD1 and NOD2-expressing cell lines at 45 µM and 1.5 µM, respectively. As shown in SI Appendix, Fig. S10 D and E, no activity was associated with the UDP-M-TetraDAP on both NOD1 and NOD2 cell models, denoting that the activity beared by F24 was due to the UDP-M-TriDAP only. The UDP-M-TriDAP seemed less effective than M-TriDAP in NOD1 stimulation, but the two compounds behaved similarly on NOD2 activation.

Muropeptide Precursor Present in F13 and F24 Are IL-10 Inducers.

As mentioned before, only F13 and F24 were considered for purification and compound identification because of interesting results on other readouts.

Indeed, the three fractions F13, F21, and F24 were assessed for the capacity to induce cytokines secretion from human monocyte-derived dendritic cells (moDCs). Interestingly, F13 and F24 induced IL-10 secretion, while F21 did not (Fig. 3C), which is why F21 investigation was not pursued. A similar pattern was observed for IL-8 (Fig. 3D). In this case, the basal level measured for the negative fractions was already non-negligible probably because of the presence of remaining contaminants.

We investigated whether the dual agonist nature of synthetic M-TriDAP-MP was associated to an improved IL-10 secretion with respect to commercial M-TriDAP, TriDAP, and MDP. Indeed, protective capacity of selected bacterial strains has been linked to the induction of local secretion of IL-10 by bacterial cell wall components and seems to be abrogated in NOD2-deficient mice (30). MDP, M-TriDAP-MP, M-TriDAP and TriDAP were tested with a dose response between 1 and 100 µM on human moDCs, for their capacity to induce IL-10 and TNF secretion (Fig. 4 A and B). M-TriDAP-MP, M-TriDAP, and TriDAP were able to induce IL-10 secretion when used at the highest concentration. Under these conditions, M-TriDAP-MP was identified as the best inducer for IL-10 (Fig. 4A) but not for TNF (Fig. 4B). We also investigated the effect of muropeptides and LPS costimulation on both IL-10 and TNF secretion (SI Appendix, Fig. S11 A and B). Only the IL-10 levels were increased by the presence of LPS but the overall profile was similar for all the compounds.

Fig. 4.

The muropeptide precursor M-TriDAP-MP induces IL-10 secretion from human moDC. (A) IL-10 and (B) TNF secretion by human moDC stimulated for 18 h with MDP (gray bars), M-TriDAP-MP (orange bars), M-TriDAP (green bars), and TriDAP (blue bars) at 1 to 10 to 100 mM. Dataare pooled from biological triplicates. Error bars represent SEM (*P < 0.05; ns, not significant) as calculated by the paired t test.

Anti-Inflammatory Potential of Muropeptide Precursor.

To further characterize the phenotype induced by the muropeptide precursor produced by F4 clone in human moDC and highlight possible difference beetween the M-TriDAP-MP from F13 and the mix UDP-M-TriDAP + UDP-M-TetraDAP from F24, microarray analysis was performed to compare transcriptome of moDC (from three different donors) stimulated with the corresponding fractions from both positive (F4D5:ABC) and negative (F4D5-pBAD) clones to those stimulated with the agonists TNF, LPS, PHA, Pam3CSK4, and MDP. Results were first normalized to the untreated condition, then, to exclude unspecific signals due to residual contaminants still present in the purified fractions, an additional normalization was done for F4D5:ABC fractions (versus the corresponding fractions from F4D5-pBAD sample). Were considered up-regulated genes having a fold increase in expression of ≥1.5 and down-regulated all genes having a fold decrease below 0.75. The overall expression profile was similar for both F13 and F24 fractions (SI Appendix, Fig. S12A); thus, only data derived from F24 are shown (SI Appendix, Table S1). Significant differences were observed between F24- and reference compounds-treated moDC (SI Appendix, Fig. S12B). Interestingly, several up-regulated genes by F4-derived fractions were members of the metalloproteinase family. Matrix metalloprotease are enzymes participating in the turnover and degradation of extracellular matrix proteins for which a role in innate immune defense and wound healing has been suggested (34). Other interesting transcripts up-regulated by F13 and F24 were Slc39A14 (Solute Carrier family 39 member 14), a zinc transporter described in epithelial cells for promoting epithelial barrier function (35) and Tjp2 (Tight Junction protein 2) (SI Appendix, Fig. S12C). Under the experimental conditions used, Il10 expression was below the threshold limit for all the conditions tested. However, we observed that most proinflammatory genes like Il6, ccl1, ccl8, and tnfα were down-regulated by the action of F4 fractions (SI Appendix, Table S1 and Fig. S12C), suggesting a general protective role of muropeptide precursor.

Muropeptide Precursor Transporter from F4 Clone Exerts Anti-Inflammatory Effects in Ex Vivo Model.

Human intestinal explants were used to simulate the complex human intestine’s microenvironment and evaluate the capacity of F4-derived muropeptide precursor to strengthen the epithelial barrier and improve tissue repair during inflammatory events as in IBD. Tissue infection with live E. coli LF82, an adherent invasive E. coli (AIEC) at 1 × 10⁹ CFU was used to trigger intestinal inflammation. Indeed, AIEC represent one of the main risk factors to develop Crohn’s disease through their adhesion to the intestinal epithelium via the interaction between the bacterial adhesin FimH and epithelial TLR4 and CEACAM6 (36, 37).

Explants were pretreated for 1 h with F4D5:ABC or F4D5-pBAD enriched samples before being exposed to live bacteria for 4 h. As shown in Fig. 5A, following 4 h of infection, LF82 bacteria caused a decrease in transepithelial electrical resistance (TEER) between 30 and 60% that was not prevented by treatment with F4D5-pBAD sample. As expected, the decrease in TEER was associated with an advanced degradation of the epithelial layer as shown by histologic analysis (Fig. 5C). Increased levels of IL-8 secretion confirmed an established inflammatory environment (Fig. 5B). On the contrary, following F4D5:ABC pretreatment, IL-8 levels remained unchanged as in control condition (Fig. 5B). Additionally, F4D5:ABC preserved the epithelial barrier from mechanical degradation induced by the pathogenic bacteria. This observation relies on both TEER values, that for resections cotreated with F4D5:ABC and LF82 were at the same levels than the uninfected ones, and on histologic data (Fig. 5C). The noninflammatory baseline status was confirmed by histological analysis at T = 0. The intestinal mucosa was perfectly homogeneous. No desquamation of the mucosa was visible, and the thickness of the mucosa and submucosa were normal.

Fig. 5.

Muropeptide precursor present in F4D5:ABC extract protect intestinal epithelium from E. coli LF82-induced inflammation. Human ileal resections (mounted on Ussing chambers) were treated with culture media (control, white bars), 1 × 10⁹ CFU E. coli LF82 (orange bars), F4D5-pBAD (blue-striped bars), and F4D5:ABC (blue bars) supernatants submitted to partial purification (ultrafiltration using 10-kDa cutoff filters followed by two cycles of acetone precipitation and a C18 solid phase extraction). (A) Transepithelial resistance (TEER), (B) IL-8 quantification by AlphaLisa, and (C) histology. One supernatant from F4D5-pBAD-treated resection lost during manipulation. Experiments are biological quadruplicates, and error bars represent SEM (*P < 0.05) as calculated by the unpaired t test.

As expected, for the untreated tissue at T = 4 h, a thinner mucosa especially in the crypts was observed, indicative of a preinflammatory stage of the tissue. This mild preinflammatory state is normally associated with the stress experienced by the tissue during chamber assembly and incubation at 37 °C for 4 h and is considered as a validated control in this experimental setting.

When the explants were incubated for 4 h in the presence of E. coli LF82, tissues were in an advanced inflammatory state with a mucosa that was totally desquamated at its apical pole and 75% desquamated at the level of 2/3 of the villus.

When F4D5:ABC fraction was incubated in the presence of E. coli LF82, a protective effect was observed. Indeed, the mucosa and submucosa remained equivalent to the control T = 4 h.

F4-derived muropeptide precursor were then effective in maintaining epithelial barrier integrity and prevented tissue inflammation.

Muropeptide Precursor Reduce the Inflammatory Process in DSS-Treated Mice.

The potential anti-inflammatory properties of F4D5:ABC supernatant were evaluated in the DSS-induced model of colitis in C57bl/6 mice (Fig. 6A).

Fig. 6.

Anti-inflammatory properties of F4D5:ABC supernatant in the DSS model of colitis in C57bl/6 mice. (A) Design of the study: Mice were gavaged daily with 200 µL of either vehicle, F4D5-pBAD supernatant, F4D5:ABC supernatant, or Pentasa granules, starting 2 d before the beginning of DSS treatment in the drinking water. All mice were killed at day 12 postbeginning of DSS treatment. (B) Loss of body weight was monitored all along the study. (C) Ratio weight/size (mg/cm) of the colon of mice at killing. (D) Histological score at killing. Data represent the mean of 10 mice/group. Errors bars represent the SEM (*P < 0.05 and **P < 0.01).

It should be highlighted that, unlike the human counterpart (hNOD1), murine NOD1 (mNOD1) recognizes M-TriDAP only at high concentrations (38); thus, the effect of F4 supernatant could be partially lowered. However, as shown in SI Appendix, Fig. S9 high levels of the three precursor were measured in the supernatant and could potentially be detected by mNOD1.

Vehicle-gavaged DSS-treated mice showed a marked decrease in body weight compared to naive animals (Fig. 6B). F4D5:ABC-gavaged mice treated with DSS loss significantly less weight than the F4D5-pBAD-gavaged mice treated with DSS, while the Pentasa-treated mice failed to recover from the weight loss induced by DSS (Fig. 6B). Despite no positive effect exerted on macroscopic parameters (Fig. 6C), F4D5:ABC treatment induced a significant improvement of the histological score (Fig. 6D and SI Appendix, Fig. S13). The analysis of the five parameters driving the Global Histological Score (SI Appendix, Fig. S13A) clearly supports the effect of muropeptide precursor on epithelial regeneration and tissue integrity. The severity of inflammation was significantly reduced (P < 0.05) in the group treated with F4D5:ABC supernatant (SI Appendix, Fig. S13B). This group showed a stratified response with few mice (n = 3) nonresponding to supernatant treatment and most of them showing slight to moderate inflammation (Score 1 to 2). This result is quite encouraging compared to the DSS group where most animals were affected by a severe inflammation associated to transmural infiltration (score 3), (SI Appendix, Fig. S13 B, C, and G). The data are also encouraging toward the Pentasa group with many mice strongly affected by DSS treatment.

In addition to the anti-inflammatory effect, the muropeptide precursor also played a protective role on the epithelial layer. As shown in SI Appendix, Fig. S13E, while most of the DSS mice presented a crypt score of 4 corresponding to complete epithelial destruction and to the absence of tissue repair, mice treated with F4D5:ABC supernatant were either preserved or slightly affected by DSS aggression (P < 0.05). Consistently, mice from the F4D5:ABC group were characterized by a marked regeneration (P = 0.0521) which corresponds to a normal tissue (SI Appendix, Fig. S13 F and G).

The overall results are consistent with the positive effect of muropeptide precursor, corresponding to improved epithelial barrier activities in vivo.

Homologs of F4-ABC-transporter.

The F4 clone and, in particular, its heterodimeric transporter allows delivery of muropeptide precursor that can contribute to intestinal homeostasis by strengthening the intestinal epithelium and controlling tissue permeability.

Components of ABC-transporters are highly represented in both Gram-negative and -positive bacteria where they play different functions (39). F4 heterodimeric transporter differs from classic ABC-transporters as it allows the secretion of multiple muropeptide precursor.

To corroborate the hypothesis that the accumulation of muropeptide precursor is specifically linked to the activity of the heterodimeric transporter present in the F4 clone, the two closest DNA sequences (76,63% identity on 99% coverage) to the heterodimer (from Acutalibacter muris and the Hungateiclostridiaceae bacterium, called hereafter ABC-Am and ABC-Hb, respectively) were cloned in the pBAD30 vector to generate pBAD-ABC-Am and pBAD-ABC-Hb. Protein sequences derived from ABC-Am and ABC-Hb genes are identical except for the first 21 amino acids that are missing in the former homolog. When compared to the F4 transporter the two homolog heterodimers present an insertion of five amino acids and several other mutations (SI Appendix, Fig. S14) mainly located in the first protein of the heterodimer.

Liquid cultures of recombinant E. coli pBAD-ABC-Am, pBAD-ABC-Hb, pBAD-ABC, and the respective negative controls were performed under the same inducing conditions. Supernatants were recovered at similar OD600 values and tested for both NF-κB activity (HEK cells) and IL-10 secretion (moDCs). As shown in Fig. 7 A and B, neither NF-κB-dependent activity nor IL-10 secretion were associated to E. coli clones harboring any of the two F4 homologous transporters.

Fig. 7.

Homolog genes to the F4-derived ABC-transporter fail to stimulate NF-κB activity in HEK-Null-NF-κB cells and IL-10 secretion from human moDC. (A) NF-κB activity on HEK-Null reporter cells and (B) IL-10 secretion from human moDC induced by supernatants of E. coli Epi300 expressing ABC-Am and ABC-Hb heterodimeric transporters. Supernatant of pBAD-ABC was used as positive control. Supernatants were derived from cultures performed under identical conditions and recovered at similar OD600. Supernatants were used at 10% v/v. Experiments are biological duplicates, and error bars represent SD as calculated by the unpaired t test.

Altogether, these results indicate that the biologic activity of F4 clone is specifically linked to the sequence of the heterodimeric ABC-transporter and that the full-length protein is necessary for optimal functioning.

AlphaFold and SWISS Model Structure Comparaison.

We investigated the F4D5:ABC, the ABC-Am, and ABC-Hb heterodimers predictions using AlphaFold2 (AF2) through the Jupyter Notebook inside Google Collaboratory program called ColabFold. For each of the 3 heterodimers, we displayed the sequence coverage, the pLDDT (predicted Local Distance Difference Test) and inter PAE (Predicted Aligned Error) scores for the Top5 predictions (rank 1 to 5) and the 3D representation for the best prediction (rank1) (Fig. S15 A–C).

The topological similarity was assessed for these protein structures using the rmsd and template modelling scores (TM score) (40, 41). The AF2 predicted F4D5 exhibited a 0.55 Å RMSD and 0.9955 TM-score values when compared with the ABC-Am structure and a 0.52 Å RMSD and 0.6096 TM-score values when compared with the ABC-Hb structure (SI Appendix, Fig. S15 D and E). Those data demonstrate that the ABC-F4D5 structure is closest to the ABC-Am structure.

As described above, a Blastp of ABC-F4D5 sequence produced several matches with more than 90% coverage and more than 80% sequence identity. However, a distance tree view of the top 100 match using blast pairwise alignment (Fig. 1C) highlighted that the ABC-F4D5 heterodimer is ranged in an isolate and distant cluster.

Usually, the structure alignment of the TM regions of distant ABC proteins exhibits low sequence conservation with respect to their dissimilar functions and substrates. However, they normally fold in highly conserved family; thus, homology modelling can provide high-quality models (42, 43).

Therefore, we examined the Blast of the ABC-F4D5 heterodimer based on the protein sequence and structure using the Modelling template software for SWISS-Model website. The top 20 templates, according to the GMQE values, were used to generate a model for the ABC-F4D5; however, those models were showing only 30% of sequence coverage (see report on analysis in SI Appendix, SI Text S1). In addition, and as recommended, we used a model template giving us the best QMEANDisCo Global value. The 6quz.1 template is at 0.78 of QMEANDisCo Global value, a nonsatisfactory model according to the comparison with Non-redundant Set of PDB Structures. To date, the interesting portion that could potentially be related to the function of that ABC-F4D5 heterodimer is the no-match sequence in the predicted closed D-lock of the protein (red arrow Fig. 8 A and B). We observed the same when the AF2 and SWISS Modelling were matching the Matchmaker sequence analysis tools from chimeraX1.3 software. The prediction of the closed D-lock region of the protein was clearly different (Fig. 8 D–F). However, we cannot exclude that this could be due to the choice of the model template used during our analysis.

Fig. 8.

F4D5:ABC transporter modelling results. (A) Structure homology comparison between the F4D5:ABC transporter prediction template and 6quz.1 template using the SWISS-Model system. (B) 3D prediction overlaps between F4D5:ABC (red) and 6quz.1 (blue). (C) Output data including coordinates, target-template alignment, modelling log, and quality estimation information. (D) SWISS-Model template Modelling: in pink the chain A and in blue the chain B. (E) AlphaFold2 Modelling (AF2) with in red the chain A and in blue the chain B. (F) Matching sequences using Matchmaker sequence analysis tools from ChimeraX 1.3. The zoomed D-lock regions of the 2 structures are shown. The arrows indicate the D-lock regions.

All together, these data demonstrate that the ABC-F4D5 transporter possess unique features (its amino acid sequence and structure) that could account for its biologic activity.

Discussion

Screening of metagenomic libraries is a powerful tool for the identification of molecules that could be further developed into new drugs. Through the screening of 5,000 clones from a human metagenomic library using NF-κB reporter cells, a bacterial muropeptide precursor transporter has been identified. It allows the secretion of at least three muropeptide precursors that accumulate in huge amount in the supernatant. The identification of a membrane transporter from a metagenomic library correlates with the observation of Estrela and coworkers (44), highlighting that most of the E. coli genes able to induce an NF-κB response in the host encode proteins related to membrane transporters, cell wall hydrolases, lipopolysaccharide biosynthetic enzymes and proteins of unknown function. The newly identified heterodimeric transporter is responsible for the secretion of unusual intermediates that all together enhance IL-10 secretion from human moDC and contribute in promoting an anti-inflammatory environment.

According to the bioinformatic analysis on the 204,938 MAGs (17), the 40-kb insert of F4 clone split on two different MAGs, GUT_GENOME259254 and GUT_GENOME139568, (the latter hosting the heterodimeric transporter) from the same taxonomical family (Acutalibacteraceae); this finding is consistent with the identification of a new unidentified bacterial strain. This hypothesis is supported, on the one side, by the mapping of the metagenomic sequencing data from 169 human gut microbiota samples which shows a homogeneous coverage on the whole length of the insert. On the other side, it is supported by phylogenetic analysis showing that the proteins forming the heterodimer are close to protein sequences of an unannotated Firmicutes bacterium. It should also be highlighted that the F4D5-ABC transporter possesses unique properties to accumulate muropeptides precursor as demonstrated by the experiences performed with the corresponding proteins from A. muris and the Hungateiclostridiaceae bacterium, with which it shares high sequence identity but whose supernatants are unable to induce IL-10 secretion from moDC.

The three identified compounds, M-TriDAP-monophosphate, UDP-M-TetraDAP, and UDP-M-TriDAP, are cell wall components that, under ordinary conditions, are not expected to be secreted in the supernatant. It is noteworthy that peptidoglycan precursor cannot accumulate in the supernatant and that UDP-M-Tetrapeptide does not exist as natural intermediate of peptidoglycan biosynthesis. Indeed, UDP-M-TriDAP should directly be converted into UDP-M-Pentapeptide (which, as the UDP-M-tetrapeptide, lacks the capacity to stimulate NOD receptors) through the addition of two amino acids (most often d-Ala-d-Ala) by the MurE synthetase. UDP-M-TetraDAP is only produced following recycling of the turnover peptidoglycan fragments. The recycling pathway can transport back into the cytoplasm the products of lytic transglycosylases through the AmpG permease and the stem peptides generated by periplasmic amidases through the oligopeptide permease (Opp) system. TetraDAP stem peptides can thus be recycled and incorporated in the de novo precursor synthesis by the action of the Mlp protein leading to the synthesis of UDP-M-TetraDAP. The ldcA gene, coding for a cytoplasmic carboxypeptidase, then converts UDP-MurNAc- TetraDAP into UDP-M-TriDAP following hydrolysis of d-Ala.

As formerly observed (45), lack of ldcA is lethal when E. coli enters into the stationary growth phase, as the incorporation of atypical tetrapeptide precursor into the peptidoglycan may result in a lethal cross-linkage defect causing bacteriolysis. The biological reason for having a peptidoglycan precursor exporter is unclear since the synthesis of precursors is an energy-consuming process and their export would constitute a waste in ATP. Interestingly, there is a precedent with the activation of the AlpK1/TifA pathway by the cytosolic ADP-heptose precursor, an intermediate in the synthesis of lipopolysaccharide (LPS) (46). In the case of ADP-heptose, this metabolite seems to be secreted via type four secretion systems (47, 48) encoded by pathogenicity islands by a yet unknown mechanism. The role of the peptidoglycan precursor exporter in its natural host(s) might also be to enhance its fitness in the intestine either by reducing the risk of accumulating toxic intermediates such as UDP-M-TetraDAP or by modulating the host immune system to create a more favorable niche. The two hypotheses are not mutually exclusive.

To date, we do not have any result justifying such accumulation of precursor in the supernatant of F4 clone. Nevertheless, it is extremely intriguing that a bacterial heterodimeric transporter is indirectly responsible for the protective (or preventive) effect during an inflammatory process. Indeed, ex vivo data indicate that F4-derived muropeptide precursors improve epithelial strength at both TEER, histology, and cytokines levels. Data from the DSS model also support a protective role of these compounds. It should be highlighted that, unlike the human counterpart (hNOD1), murine NOD1 (mNOD1) recognizes M-TriDAP only at high concentrations (38); thus, the effect of F4 supernatant could be partially lowered and could explain the moderate effect observed using the mouse model. Taken together, our findings suggest that F4 clone could be exploited as a Live Biotherapeutic Product once vectorized in a probiotic strain. We believe that F4 transporter could be useful to prevent or treat leaky gut syndrome (LGS), a pathology characterized by a chronic increase of intestinal permeability (49). Altered gut microbiota composition (dysbiosis) can be responsible for LGS through release of metabolites, mainly lipopolysaccharides. For instance, the spontaneous elevation of Gram-negative bacteria, generally indicated as E. coli blooming, is a potential trigger of LGS, but also of Crohn’s disease, the latter characterized by high prevalence of Adherent Invasive E. coli AIEC (up to 60%) in the intestinal mucosa of the patients (50). We speculate that high LPS levels in the gut can synergize with muropeptide precursor, secreted by means of F4 transporter, as observed in vitro, and help to reduce inflammation through reduction of both epithelial permeability and proinflammatory cytokines secretion.

Materials and Methods

High-Throughput Screening of Metagenomic Clones.

HT-29 NF-κB/SEAP-25 cells (19) were used to screen a metagenomic library issue from the intestinal microbiota of healthy patients (ACXLAB library).

Human feces sample collection was approved by CODECOH. Samples were obtained from Danish patients recruited in the frame of the MetaHIT project as described by Qin et al. (15). All individuals gave written informed consent before participating in the study.

The library was generated in E. coli Epi300 strain. Each clone carrying a fosmid with a metagenomic insert of 40 kb and the chloramphenicol gene (for antibiotic selection). E. coli bearing a fosmid without a metagenomic insert was used as control (referred as metagenomic control). Metagenomic clones were cultured for 24 h in lysogeny broth medium in 96-well plates. Then, optical density at 600 nm (OD600) was measured and given as a raw value. Bacterial cells were broken following two freeze (−80 °C) thaw (RT) cycles. The suspension was filtered by centrifugation at 4 °C through a 0.2-µm microplate filter and added to NF-κB/SEAP-25 reporter clone at 10% vol/vol to a final volume of 100 µL. Seeding of the reporter cells, addition of lysates, and SEAP activity measurement were performed using a robotic pipetting workstation (Microlab Star, Hamilton). Selected metagenomic clones were grown further in independent cultures and tested to validate their effect.

Compound Purification.

Frozen supernatants were thawed at 4 °C. Pure chilled acetone was added (70% v/v acetone) to the <10 kDa fraction and incubated for 2 h 30 at 4 °C. Solutions were gently mixed every 30 min by hand and then centrifuged at 3,500 g for 20 min at 4 °C to precipitate insoluble compounds. Supernatants were then transferred to new tubes, and pure chilled acetone was added (85% v/v acetone) and treated as described before. Pellets obtained from centrifugation were dried by blowing N2 over them during 5 to 10 min. Dry pellets were solubilized in ultrapure water, cooled to −60 °C, and lyophilized for 36 h to remove the solvent. Acetone pellets were solubilized in ultrapure water and loaded on Strata-X cartridges (Phenomenex). The flowthrough recovered from Strata-X Solid Phase Extraction, containing the active compounds, was dried overnight, then solubilized in ultrapure water and purified onto preparative diol-HILIC column (Kromasil HILIC-D, 10 µM, 50 × 250 mm) using water (solvent A)/acetonitrile (solvent B) mobile phases (isocratic for 20 min at 75% B; linear gradient 75 to 58% B in 50 min; linear gradient 58 to 5% B in 10 min; 40 mL/min) with UV detector (215 nm). Active fractions, selected for their ability to induce IL-10 secretion in moDC, were purified through a second HPLC dimension using preparative graphitized carbon column (Phenomenex Hypercarb 2 mm, 5 µM, 21 × 150 mm) with solvent A: water + 0.3% formic acid and solvent B: Acetonitrile + 0.3% formic acid (isocratic 7 min at 10% B; linear gradient 10 to 50% B in 46 min; isocratic at 50% for 6 min; linear gradient 50 to 90% in 1 min; isocratic at 90% for 10 min; 18 mL/min) with a detection at 215 nm. These two HPLC steps allowed the identification of UDP-M-TriDAP and UDP-M-TetraDAP (corresponding to compounds present in F24). An additional purification step (3D-HPLC) using the same graphite column was necessary to complete the purification of M-TriDAP-MP (present in F13): Solvent A: water + 0.1% TFA; Solvent B: acetonitrile + 0.1% TFA. (215 nm; linear gradient 5 to 15% B in 43 min; linear gradient 15 to 100% B in 7 min; isocratic for 10 min at 100% B; 18 mL/min).

Ex Vivo Experiments on Human Ileal Explants.

Human ileal explants were left untreated or were pretreated (apical compartment of the Ussing chambers) for 1 h at 37 °C with the prepurified fractions of F4D5:ABC or F4D5-pBAD clones diluted 1:10 (v:v).

After the 1 h preincubation period at 37 °C in a CO₂ incubator, LF82-gfp bacteria were added into the apical compartment of the Ussing chambers at 1 × 10⁹ bacteria/mL, and the incubation was prolonged for 4 h. TEER was measured each hour with a Millicell-ERS (Electrical Resistance System) Voltohmmeter (Millipore).

IL-8 was quantified by ELISA (OptiEIA IL-8, BD Biosciences) from the spent medium in the basolateral side.

Human explants were then washed 6 times with 2 mL of PBS (Gibco), removed from Ussing chambers, and fixed using PBS PFA 4% solution (Merck) at 4 °C during 24 h before being processed for microscopy and histological analysis (H&E staining).

Animal Experiment.

C57bl/6 mice from Charles River (France) were housed in a room with a 12-h light/dark cycle and a temperature of 20 ± 5 °C they were provided free access to diet and water. The mice were acclimated to the laboratory conditions and used for the animal experiment for 2 wk. Mice were divided into 5 groups of 10 animals each except for healthy control group (negative control) where 5 mice were used. The other groups included the DSS-treated mice (positive control), the F4D5:ABC- DSS and the F4D5-pBAD- DSS cotreated mice. A last group, Pentasa-DSS, was added as mice recovery control. All the animals except the healthy control group were fed with 2.5% DSS instead of drinking water for 5 d. F4D5:ABC-DSS and the F4D5-pBAD-DSS groups were treated with 0.2 mL of supernatant by oral gavage once a day starting from 2 d before DSS till killing. The healthy control and IBD control were provided 0.2 mL of PBS. Animal experiments were conducted according to governmental guidelines (articles R214-87 à R214-137 code rural update 13 February 2013 according to European directive 2010/63/UE) and those of the Nord-Pas de Calais Ethical Committee for animal use.

Statistical Analysis.

The results are given as mean ± SEM. Statistical analysis was performed with GraphPad Prism 9 software using either paired or unpaired t test. P ≤ 0.05 was considered significant.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

Some study data are available: the informed consent does not allow us to share on a public domain the metagenomic data. However, these data are available from the corresponding author upon reasonable request. Previously published data were used for this work (Bioinformatic analysis: ENA repository PRJEB32631, PRJEB7369, PRJEB9576, SRP057027, PRJNA400072) (51–55). Microarray dataset (Log2 transformed intensity) has been deposited in the Figshare database (https://figshare.com/account/home) under accession no. 10.6084/m9.figshare.24612663 (56). All other raw data, included in the manuscript and/or SI Appendix, are available upon request.