Bacteroides fragilis fucosidases facilitate growth and invasion of Campylobacter jejuni in the presence of mucins

Abstract The enteropathogenic bacterium, Campylobacter jejuni, was considered to be non‐saccharolytic, but recently it emerged that l‐fucose plays a central role in C. jejuni virulence. Half of C. jejuni clinical isolates possess an operon for l‐fucose utilisation. In the intestinal tract, l‐fucose is abundantly available in mucin O‐linked glycan structures, but C. jejuni lacks a fucosidase enzyme essential to release the l‐fucose. We set out to determine how C. jejuni can gain access to these intestinal l‐fucosides. Growth of the fuc + C. jejuni strains, 129,108 and NCTC 11168, increased in the presence of l‐fucose while fucose permease knockout strains did not benefit from additional l‐fucose. With fucosidase assays and an activity‐based probe, we confirmed that Bacteriodes fragilis, an abundant member of the intestinal microbiota, secretes active fucosidases. In the presence of mucins, C. jejuni was dependent on B. fragilis fucosidase activity for increased growth. Campylobacter jejuni invaded Caco‐2 intestinal cells that express complex O‐linked glycan structures that contain l‐fucose. In infection experiments, C. jejuni was more invasive in the presence of B. fragilis and this increase is due to fucosidase activity. We conclude that C. jejuni fuc + strains are dependent on exogenous fucosidases for increased growth and invasion.

bacteria have to compete for nutrients and space at this interface in order to proliferate and cause infection.
One bacterial phylum that is well-equipped to degrade polysaccharides is the Bacteroidetes that make up almost half of the bacteria found in the intestine (Eckburg et al., 2005). Genomic and proteomic analyses of Bacteroides thetaiotaomicron demonstrated that this bacterium possesses over 280 glycosidases, of which 11% are located on the outer membrane or released extracellularly (Eckburg et al., 2005;Comstock & Kasper, 2006, http://www.cazy.org/b5118.html).
Sialidases and fucosidases are the glycosidases that target the two major terminal epitopes found on mucin O-linked glycans, sialic acid and fucose. In the human intestine, the density of the fucosylated O-glycans decreases from ileum to colon while sialylated O-glycans show the reversed pattern (Thiele et al., 2015). Both fucose and sialic acid are reported in the literature to be correlated with the ability of pathogenic bacteria to thrive within the gut (Li et al., 2019;Ng et al., 2013;Pacheco et al., 2012). Enteropathogenic bacteria need to compete for nutrients with resident commensals to gain a foothold in our gut and it is thus important to understand how they access mucin-derived monosaccharides to establish infection.
They showed that two commonly studied C. jejuni strains, NCTC 11168 and RM1221, have a growth benefit in the presence of L-fucose. Previous studies have identified potential similarities between the L-fucose breakdown pathway of C. jejuni and the plant pathogen Xanthomonas campestris (van der Hooft et al., 2018). A recent study elucidated the L-fucose breakdown pathway in the C. jejuni NCTC 11168 fuc + strain by solving the structure of a putative dehydrogenase, FucX, that can reduce L-fucose and D-arabinose in vitro .
Besides the effects of L-fucose breakdown for basic metabolism, L-fucose utilisation has a broader impact on C. jejuni biology. Transcriptomics analysis of C. jejuni fuc + strains showed a large-fold change in transcript abundancies upon addition of L-fucose with 74 transcripts up-regulated and another 52 down-regulated (Stahl et al., 2011). For example, up-regulation was seen for the immunoreactive cstA (cj0917c), a carbon starvation protein A homologue (Nielsen et al., 2012;Stahl et al., 2011). However, explanations for these up-and down-regulations are not immediately clear. Furthermore, a recent metabolomics study has demonstrated that C. jejuni fuc + strains have an adaptive metabolome that changes in the presence of L-fucose (van der Hooft et al., 2018). Metabolites dependent on L-fucose, such as thiazolidine-containing metabolites, could be detected that demonstrate the activation of metabolic pathways generating bio-active compounds in C. jejuni (van der Hooft et al., 2018).
The fucose operon is not conserved universally among C. jejuni strains, but its presence has been linked to hyper invasiveness in in vitro virulence and transposon mutagenesis (Fearnley et al., 2008;Javed et al., 2010). The capacity of C. jejuni fuc + strains to utilise L-fucose correlates with colonisation and pathogenicity advantages in neonatal piglet model (Stahl et al., 2011).
Interestingly, the C. jejuni genomes that have been sequenced, so far, lack fucosidases that would be necessary to release L-fucose from host mucins. A recent publication verified a lack of C. jejuni fucosidases and showed increased growth of C. jejuni fuc + strains in the presence of fucosidases secreted by commensal bacteria . Complementary to this finding, we want to determine the effect of fucosidase activity of residing commensal bacteria on C. jejuni fuc + strains hyper invasiveness in the presence of mucins.
To investigate the dependence of C. jejuni on exogenous fucosidases activity and its implications for growth and virulence, we took an interdisciplinary approach by combining microbiology with an activity-based probe (ABP), competitive fucosidase inhibitors and infection assays. With the use of the ABP, we were capable of visualising active GH29 fucosidases secreted by B. fragilis. Using this diverse approach, we demonstrated that fucosidases, secreted by commensal Bacteroides fragilis, increased growth and invasion of the 2 | RESULTS 2.1 | The hyperinvasive C. jejuni 108 strain contains the genomic island required for fucose utilisation The fuc + operon, containing genes, cj0480c to cj0490, has been identified in C. jejuni NCTC 11168 and RM1221 to be required for L-fucose utilisation (Stahl et al., 2011). We sequenced the hyperinvasive C. jejuni 129,108 (108) strain and identified 11 genes to be homologous to the cj0480-cj0490 gene cluster of NCTC 11168 with a sequence similarity of 98.91%. Figure 1 shows the schematic representation of the fuc + operon. The genes encoded by the C. jejuni fuc + operon are predicted to include a transcriptional regulator (FucR), a synthase (dapA), a dehydratase (uxaA'), two major facilitator superfamily transporters (Cj0484 and FucP), two dehydrogenases (FucX and Cj0489), a hydrolase (Cj0487) and a mutarotase (Cj0488) Stahl et al., 2011). Cj0486 is homologous to fucose permeases found in other bacteria and was previously shown to be an essential component of the active L-fucose assimilation pathway in C. jejuni NCTC 11168 (Stahl et al., 2011). The predicted Cj0486 gene product in C. jejuni 108 is 99% identical to its NCTC 11168 homologue. Based on its sequence, we predict that the hyperinvasive C. jejuni 108 is a fuc + strain that has the ability to scavenge and metabolise L-fucose.

| L-fucose increases growth of C. jejuni 108
We next investigated the effect of L-fucose on growth of the C. jejuni 108 strain. Strains 108 and NCTC 11168 were grown in DMEM with and without L-fucose. In the presence of L-fucose, both strains reached a higher final optical density, but the growth increase was most pronounced in the 108 strain ( Figure 2a). In contrast with L-fucose, no significant increase in growth of C. jejuni 108 was seen in the presence of sialic acid ( Figure 2b). We generated deletion strains for the fucose permease cj0486 for both the 11,168 and 108 strains and tested their ability to grow on L-fucose.
The growth of the mutant strains was similar with or without addition of L-fucose, indicating that both mutants lost their ability to utilise L-fucose (Figure 2c). These results demonstrate that the hyperinvasive C. jejuni 108 strain contains a pathway for the uptake and metabolism of L-fucose and that L-fucose confers a growth benefit.
2.3 | Detection of fucosidase activity of commensal Bacteroides fragilis using chemical tools Our results demonstrate that C. jejuni fuc + strains can utilise the monosaccharide L-fucose, but the release of L-fucose from complex mucin O-linked glycans requires extracellular fucosidase activity (mucin O-glycan Figure 3a). We hypothesize that C. jejuni fuc + strains are dependent on fucosidase activity of residing commensals such as Bacteroides species. We selected Bacteroides fragilis and Bacteroides thetaiotaomicron for our experiments to induce extracellular fucosidases and detect their activity with a set of molecular tools (the structures of the tools used in our assays are depicted in Figure 3a). In a previous study, proteomics analysis showed the presence of secreted GH29 fucosidases in this specific B. fragilis strain (Elhenawy et al., 2014). To confirm the nature of the fucosidase enzymes secreted by B. fragilis in our experimental set-up, we used an activitybased fucosidase probe (JJB256) that was previously synthesised and applied by the Overkleeft group . When JJB256 is bound by a catalytically active GH29 fucosidase its reactive warhead (aziridine, blue, Figure  2.4 | Campylobacter jejuni is dependent on B. fragilis fucosidases for growth on mucin We next investigated if C. jejuni 108 can benefit from exogenous fucosidase activity for growth on mucin O-linked glycans. Porcine gastric mucin (PGM) was pretreated with or without purified fucosidase FucHS, and subsequently C. jejuni 108 was added and incubated for 24 hours. Pretreatment of the mucin with the fucosidase enzyme resulted in a significant increase in C. jejuni colony forming units (CFUs) compared to non-treated PGM (Figure 4a). Pretreatment of PGM with sialidase did not confer a significant growth benefit for C. jejuni 108 (Figure 4b), which is in line with our earlier observations in the growth assays with sialic acid.
Next, we set out to investigate the effects of the more complex B. fragilis supernatant. We normalised fucosidase activity in the supernatant fraction to the previously used FucHS activity by comparing their in-gel fluorescent signals with probe JJB256 using ImageJ software. PGM was pre-treated with the concentrated B. fragilis supernatant fraction and added to the C. jejuni 108 culture.
Campylobacter jejuni growth was significantly increased in the presence of the pre-treated PGM compared to the untreated PGM ( Figure 4c). Addition of 100 μM FNJ to the B. fragilis supernatant fraction resulted in a significant decrease in C. jejuni growth on PGM, demonstrating that the increase in C. jejuni growth is due to B. fragilis fucosidase activity (Figure 4d). As a control, we also investigated the effect of FNJ on the growth of C. jejuni 108. Growth curves of C. jejuni 108 were similar in the presence and absence of FNJ, indicating that this inhibitor does not directly impact growth (Figure 4d).
Together, these results demonstrate that for growth on glycosylated mucin, C. jejuni 108 is dependent on secreted fucosidases from other species.
2.5 | Effects of fucosidase activity on C. jejuni 108 invasion into intestinal epithelial cells  fucosidases activity and no cell membrane bound fucosidases (Elhenawy et al., 2014). Taken together, these results support our hypothesis that locally liberated L-fucose by secreted fucosidases from other species can increase growth and invasion of fuc + C. jejuni strains at the intestinal epithelial interface.

| DISCUSSION
When C. jejuni invades the mucosal lining of the intestinal epithelium it has to compete with residing intestinal microbiota for nutrients and space (Lee, O'Rourke, Barrington, & Trust, 1986). Although C. jejuni prefers growth on amino acids, some hyperinvasive strains possess the ability to metabolise L-fucose (Fearnley et al., 2008;Javed et al., 2010), which is an abundant terminal component of mucin O-glycans that cover the intestinal epithelium. The ability of C. jejuni fuc + strains to metabolise L-fucose confers a competitive advantage in infection models (Stahl et al., 2011). However, as C. jejuni lacks endogenous fucosidases, it does not have the capacity to release Lfucose from mucin O-glycans. Our results here show that secreted fucosidases of a commensal, Bacteroides fragilis, facilitate enhanced growth of the hyperinvasive C. jejuni fuc + strain 108 on glycosylated mucins (Figure 7). Furthermore, we used activity-based protein profiling (ABPP) and chemical competitive inhibitors to demonstrate the crucial contribution of these exogenous fucosidases to the increased invasion by C. jejuni 108 into intestinal epithelial cells. Our findings complement two recent publications on the topic of C. jejuni fuc + strains that investigated nutrient scavenging by C. jejuni fuc + strains and demonstrated how the presence of glycoproteins in human milk affects the selection of these specific strains Garber et al., 2020). Furthermore, it is interesting to speculate that our data are in line with a previous finding that individuals with higher proportions of Bacteroides species are more susceptible to C. jejuni infections (Dicksved, Ellström, Engstrand, & Rautelin, 2014).
Bacteroides species are known to tightly regulate the secretion of their extracellular fucosidases, which levels often appear lower   (Sonnenburg et al., 2005).

Fucosylated mucin O-glycans
ABPP is an established and powerful technique to label and detect catalytically active enzymes in their native environment (Cravatt, Wright, & Kozarich, 2008). Our successful application of chemical probe, JJB256, to label previously putative retaining GH29 α-L-fucosidases in the secretome of B. fragilis highlights the possible further application of this probe and potential future derivatives . One such application could be screens for secreted GH29 fucosidases in both microbiota and pathogens.
The L-fucose cross-feeding that we observe between the commensal Bacteroides fragillis, and invading pathogen Campylobacter jejuni is a strategy encountered more often among enteropathogenic bacteria. The virulence genes of Enterohemorrhagic E. coli (EHEC) are regulated by the fucose-activated FusKR signalling pathway. When EHEC is grown on mucins in the presence of Bacteroides, virulence genes are upregulated in a FusKR-dependent manner. These results suggest that EHEC uses L-fucose, liberated by Bacteroides fucosidases, to modulate its pathogenicity (Pacheco et al., 2012). This strategy is not limited to L-fucose, as two other enteric pathogens, Salmonella Typhimurium and Clostridium difficile, have both been shown to use sialic acids that were liberated by sialidases expressed by microbiota (Ng et al., 2013). Salmonella Typhimurium shows a significant upregulation of genes involved in the sialic acid catabolism pathway when infecting mice that contain Bacteroides species compared to germ-free mice. A similar effect is seen for C. difficile that upregulates its sialic acid catabolism in mice colonised with wild-type Bacteroides thetaiotaomicron compared to mice colonised with a sialidasedeficient Bacteroides mutant (Ng et al., 2013).
Differences in host tropism of C. jejuni strains and host mucin composition underscore the importance of choosing the right intestinal epithelial model to investigate virulence of C. jejuni fuc + strains. In a piglet model of human disease, the NCTC 11168 fuc + strain has been shown to possess a competitive advantage when colonising the intestinal tract (Stahl et al., 2011). In our studies, we used a confluent monolayer of human intestinal Caco-2 cells and observed that invasion of a C. jejuni fuc + strains was enhanced by fucosidase activity from a co-culture with B. fragilis. In chicken, C. jejuni colonises the intestinal tract as a commensal and hence does not invade. Chicken mucins have an inhibitory effect on C. jejuni invasion into epithelial cells and, in these animals, fuc + strains do not have a competitive advantage over other C. jejuni strains. However, when chickens were fed additional L-fucose, the C. jejuni fuc + wild-type strain was more effective in colonising the intestinal tract compared to a fucose permease knockout strain (Byrne, Clyne, & Bourke, 2007;Stahl et al., 2011). Compared to human mucins, chicken mucins contain a large amount of sulfate modifications on their mucin O-glycans (Struwe et al., 2015). One hypothesis is that the sulfate modifications block the function of exogenous fucosidases (Roberton & Wright, 1997). This hypothesis is supported by a mouse model where a decrease in sulfation enhanced intestinal penetrability by pathogens, including C. jejuni (Dawson et al., 2009). Interestingly, the occurrence of inflammatory bowel disease in humans, which is characterised by decreased mucus barrier function, has been correlated to an altered microbiota with an increased sulfate-reducing bacterial population (Ijssennagger, van der Meer, & van Mil, 2016). Therefore, differences in mucin O-glycan sulfation and surrounding microbiota could contribute to C. jejuni fuc + host tropism. The accessibility of terminal L-fucosides on mucin O-glycans for bacterial fucosidases in the presence or absence of specific sulfation patterns is, therefore, an interesting area for future studies.
Wild type and mutant strains were grown for 4 days on saponin agar plates supplemented with corresponding antibiotics at 42 C under microaerophilic conditions. Several colonies were picked and plated onto fresh saponin agar plate and grown for 24 hr at 37 C under microaerophilic conditions. One colony was picked and grown overnight in HI at 37 C. These cultures were used to inoculate the growth medium at an OD 600 of 0.1. Aliquots of this cell suspension were pipetted into a 96-well plate for use with a synergy HTX plate reader.
The plates were placed inside the hypoxic glove box (10% CO 2 , 5% O 2, 85% N 2 ) in a plate reader and incubated at 37 C with moderate, continuous shaking for 20 hr with OD measurements every 10 min.
Each growth condition was assessed in triplo and three biological replicates were performed. Statistical analysis was performed using a Student t test.

| 4-Umbelliferon fucopyranoside assay for fucosidase activity
The enzymatic activity of α-L-fucosidases was assayed at 37 C by

| Mammalian cells and culture conditions
The human gastrointestinal epithelial cell line, Caco-2 (ATCC-HTB-37), was routinely cultured in 25 cm 2 flasks in Dulbecco's modified Eagle's medium (DMEM), containing 10% fetal calf serum (FCS), at 37 C in 10% CO 2 . For C. jejuni gentamicin protection invasion assays, Caco-2 cells were split into six-well plates and grown for 5 days to form a monolayer before C. jejuni infection. For microscopy analysis, the cells were cultured on circular glass coverslips in 24-well plates.

| Campylobacter jejuni infection assays
Campylobacter jejuni cultures were grown in HI for 24 hr at 37 C under microaerophilic conditions and adjusted to OD 600 of 0.05 in 1 mL DMEM medium. Five day-grown Caco-2 cells were washed twice with DMEM without FCS. Bacteria were added to the cells in DMEM without FCS (+/− 10 mM L-fucose; +/− 1 μL FucHS) at MOI = 100 and incubated under microaerophilic conditions at 37 C for 3.5 hr. The cells were washed five times with DMEM without FCS and replaced with DMEM without FCS, containing 250 μg/ml gentamicin, and incubated for 3 hr at 37 C to kill extracellular bacteria. The cells were then washed three times with PBS and lysed with 0.5% Triton X-100 in PBS for 5 min at 37 C.
Serial dilutions were made and plated on saponin agar plates that contained appropriate antibiotics. Plates were incubated at 37 C under microaerophilic conditions and the number of colony-forming units was (CFU) determined. Statistical analysis was performed using a Student t test.
4.10 | Bacteroides fragilis and C. jejuni co-culture infection assay Caco-2 cells and C. jejuni 108 were prepared as mentioned above, B. fragilis (OD 600 1.0 of overnight culture; MOI 10) was added before addition of C. jejuni. FNJ (CAS 99212-30-3, Carbosynth) was added to the medium simultaneously with B. fragilis with a final concentration of 100 μM, to inhibit fucosidase activity. Invasion of C. jejuni was determined as described above.