Clostridioides difficile LuxS mediates inter-bacterial interactions within biofilms

The anaerobic gut pathogen, Clostridioides difficile, forms adherent biofilms that may play an important role in recurrent C. difficile infections. The mechanisms underlying C. difficile community formation and inter-bacterial interactions are nevertheless poorly understood. C. difficile produces AI-2, a quorum sensing molecule that modulates biofilm formation across many bacterial species. We found that a strain defective in LuxS, the enzyme that mediates AI-2 production, is defective in biofilm development in vitro. Transcriptomic analyses of biofilms formed by wild type (WT) and luxS mutant (luxS) strains revealed a downregulation of prophage loci in the luxS mutant biofilms compared to the WT. Detection of phages and eDNA within biofilms may suggest that DNA release by phage-mediated cell lysis contributes to C. difficile biofilm formation. In order to understand if LuxS mediates C. difficile crosstalk with other gut species, C. difficile interactions with a common gut bacterium, Bacteroides fragilis, were studied. We demonstrate that C. difficile growth is significantly reduced when co-cultured with B. fragilis in mixed biofilms. Interestingly, the absence of C. difficile LuxS alleviates the B. fragilis-mediated growth inhibition. Dual species RNA-sequencing analyses from single and mixed biofilms revealed differential modulation of distinct metabolic pathways for C. difficile WT, luxS and B. fragilis upon co-culture, indicating that AI-2 may be involved in induction of selective metabolic responses in B. fragilis. Overall, our data suggest that C. difficile LuxS/AI-2 utilises different mechanisms to mediate formation of single and mixed species communities.

Formation of adherent communities within the gut requires communication between bacteria. For many species, quorum sensing (QS) is important for the construction and/or dispersal of biofilm communities 31 , with bacteria utilising diverse QS systems 31,32 . Many bacteria possess the metabolic enzyme LuxS, which is involved in the detoxification of S-adenoslylhomocysteine during the activated methyl cycle. Whilst catalysing the reaction of S-ribosylhomocysteine to homocysteine, LuxS produces the bi-product 4,5-dihydroxy-2,3-pentanedione (DPD). DPD is an unstable compound that spontaneously cyclises into several different forms. These ligands are collectively known as autoinducer-2 (AI-2), a group of potent, cross-species QS signalling molecules 33 . In many bacteria, including C. difficile, AI-2 plays a role in biofilm formation, with luxS mutants showing a defect during biofilm formation and development 27,[34][35][36][37][38][39] . The precise mode of action for LuxS in C. difficile has remained elusive as a result of conflicting studies and the lack of a clear receptor for AI-2 34,40,41 .
Here we investigate the role of LuxS within C. difficile and mixed biofilm communities. Interestingly, we find that C. difficile LuxS/AI-2 mediates the induction of two putative C. difficile R20291 prophages within C. difficile biofilms. In mixed biofilms, we show that in the presence of B. fragilis, a gut bacterium, C. difficile growth is inhibited and this inhibition is alleviated in the absence of LuxS. Dual species transcriptomics show that distinct metabolic pathways are triggered in mixed cultures with the wild type (WT) and luxS mutant C. difficile strains.

LuxS mediates biofilm formation in vitro.
We previously reported that a R20291 C. difficile luxS mutant (luxS) was defective in biofilm formation as measured by crystal violet (CV) staining 27 . However, in our subsequent studies, although we see a reduction in biofilms at 24 h, we observe a high variability in the WT biofilms formed at 24 h and in the reduction in the luxS mutant between experiments (Fig. 1A). Nevertheless, the biofilm defect for luxS was very consistent at later time points (72 h) (Fig. 1A,B). In spite of differences in total biofilm content, colony counts from the WT and luxS biofilms were similar at both time points (Fig. 1C). There were also no differences in the spore content of WT and luxS biofilms (~0.003% at 24 h and ~0.1% at 72 h) (Fig. S1).
To determine if AI-2 signalling is involved in biofilm formation, we first performed an AI-2 assay from both planktonic and biofilm supernatants as described by Carter et al. 2005 (Fig. S2A). AI-2 is produced maximally in mid-log and stationary phases as previously reported 40 . The WT strain produced less AI-2 in 24 h biofilms compared to log phase culture, while the luxS strain did not produce AI-2 as expected (Fig. S2B). To study if the reduction in biofilm formation in luxS could be complemented by chemically synthesised 4,5-dihydroxy-2, 3-pentanedione (DPD), the precursor of AI-2, was supplemented in the culture medium. Whilst high concentrations (>1000 nM) appeared to have only a partial effect on biofilm formation (Fig. S3), a concentration of 100 nM was capable of restoring the WT phenotype (Fig. 1D), indicating that AI-2 may be involved in signalling within C. difficile biofilms.

RNA-seq analysis reveals LuxS-mediated prophage induction.
To investigate mechanism of action of LuxS/AI-2 in C. difficile, transcriptional profiles of C. difficile WT and luxS strains planktonically cultured (OD 600 0.8 in BHI) were first compared using RNA-seq. However, surprisingly, no differential transcriptional changes were observed (Accession number E-MTAB-7486). Following this, an RNA-seq analysis was performed with total RNA isolated from C. difficile WT and luxS biofilms grown in BHIS +0.1 M glucose [(BHIS-G) conditions used for biofilm assays in Fig. 1] for 18 h (Accession number E-MTAB-7523). Both strains show similar planktonic culture growth rates in BHIS-G as reported previously 27 .
The DESeq2 variance analysis package 42 was used to identify genes that were differentially expressed in luxS ≥1.6-fold relative to the WT strain, with a p -adjusted value ≤0.05. This pairwise analysis identified 21 differentially expressed genes ( Fig. 2A,C) (Table 1). Interestingly, all 18 down-regulated genes correspond to two prophage regions located within the C. difficile R20291 genome, CDR20291_1415-1464 and CDR20291_1197-1226 ( Fig. 2B), as identified using the online phage search tool, Phaster 43,44 . A Fisher's exact test (p-value < 0.001 for both prophage regions) further confirmed an enrichment of differently regulated genes in prophage regions. There were only three genes upregulated in the luxS biofilms compared to the WT; two of these were involved in trehalose utilisation, while the third was a phosphotransferase system glucose-specific transporter.
To demonstrate the presence of phage in the biofilm, cell-free supernatants were treated with DNase, before a subsequent DNA extraction was performed. As the bacterial cells were already removed, only DNA within intact bacteriophages would be protected from DNase. A 16S rRNA gene PCR was performed to confirm digestion of all free extracellular genomic DNA from the biofilm (Fig. 3A). PCRs with primers corresponding to genes specific to each prophage, confirmed that the DNA extracted had come from the phage (Fig. 3B). Since cell lysis is linked to phage release, we quantified and compared the total extracellular DNA (eDNA) content of luxS mutant and WT biofilms. The WT biofilms contained more eDNA compared to the luxS mutant at 24 h and 72 h (Fig. 3C). Overall, these data suggest that AI-2 may play a role in inducing prophages in C. difficile biofilms, which leads to phage-mediated host cell lysis and eDNA generation, which may contribute to subsequent biofilm accumulation.
C. difficile is inhibited when cultured with B. fragilis in mixed biofilms. Given the high microbial density within the gut, C. difficile likely needs to interact with other members of the gut microbiota to establish itself within this niche. As an inter-species signalling function has been previously proposed for AI-2 45 , we sought to investigate the interactions between a gut-associated Bacteroides spp and C. difficile. We examined C. difficile interactions with Bacteriodes fragilis, a gut commensal and pathogen, that has been previously reported to negatively correlate with CDI infections 46 .
C. difficile formed significantly more biofilms in vitro compared with B. fragilis in monocultures, as measured by CV staining (Fig. 4A). When both organisms were co-cultured, less biofilm was formed compared to www.nature.com/scientificreports www.nature.com/scientificreports/ increase in the actual spore numbers (Fig. S7). These data suggest that the presence of B. fragilis in biofilms results in inhibition of C. difficile growth.
To understand if the inhibitory effect was due to a factor secreted by B. fragilis, C. difficile and B. fragilis were co-cultured under planktonic conditions for 6 h and 10 h (Fig. 4C). However, there were no significant differences in C. difficile bacterial numbers between mono and co-culture. Additionally, supplementing biofilms with B. fragilis planktonic or biofilm culture supernatants did not cause C. difficile growth inhibition (Fig. S8), indicating that the observed inhibitory effects were specific to adherent biofilms i.e. when they are in close proximity to each other.

Luxs is involved in the B. fragilis inhibition of C. difficile.
To study the role of LuxS in C. difficile-B. fragilis interactions, WT C. difficile and luxS strains were co-cultured with B. fragilis in mixed biofilms. CV staining of biofilms showed that there was less luxS biofilm formed compared to the WT when co-cultured with B. fragilis (Fig. 5A). While colony counts of C. difficile in monocultures were similar for both WT and luxS strains (Fig. 1B), when co-cultured with B. fragilis the bacterial counts for both C. difficile strains were significantly reduced, although the reduction was significantly higher for the WT than luxS (Fig. 5B). Colony counts for B. fragilis increased significantly during co-culture, with similar levels observed in both co-culture conditions (Fig. 5B). There was no increase in the spore numbers in the luxS in mixed biofilms (Fig. S7). These data suggest that AI2/ LuxS is involved in mediating the B. fragilis-induced inhibition of C. difficile, when they are within adherent communities. (C) Heat map representation of the genes that were differentially expressed in luxS, red and green indicate down-and up-regulation respectively when compared to WT. 18 prophage genes were found to be down-regulated in luxS relative to WT, whilst two genes involved in trehalose metabolism were up-regulated in luxS relative to WT. Data shown is the mean of 3 independent experiments in triplicates. Differential expression was defined as ≥1.6-fold change relative to WT with an adjusted p-value ≤ 0.05. (   (B) Colony counts for both C. difficile (vegetative cells) and B. fragilis from mono and co-culture biofilms after 24 h. (C) Colony counts for both C. difficile (vegetative cells) and B. fragilis from mono and co-culture during planktonic growth. Data shown is the mean of 3 independent experiments in triplicates and error bars indicate SD, **p < 0.005, ****p < 0.0001 as determined by one-way ANOVA, Tukey's multiple comparison test, ns -not significant (significant differences were determined for C. difficile or B. fragilis mean CFU counts between single and mixed biofilms). Figure 5. B. fragilis-mediated inhibition of C. difficile is more prominent for WT than LuxS. (A) Biofilms for mono and co-cultures of C. difficile WT and luxS with B. fragilis were grown for 24 h and stained with 0.2% CV and were quantified using a spectrophotometer OD 570 . (B) Colony counts for C. difficile WT, C. difficile luxS (vegetative cells) during co-culture with B. fragilis were performed at 24 h. Data shown is the mean of 3 independent experiments in triplicates and error bars indicate SD, **p < 0.01, ***p < 0.001 as determined by one-way ANOVA, Tukeys multiple comparison test (significant differences were determined for C. difficile or B. fragilis mean CFU counts between single and mixed biofilms).
www.nature.com/scientificreports www.nature.com/scientificreports/ biofilm co-cultures of C. difficile WT or luxS with B. fragilis (Accession number E-MTAB-7523). Differentially expressed genes were defined as having ≥1.6-fold relative to their respective control (mono-cultures of either B. fragilis or C. difficile WT), with an adjusted p-adjusted value ≤ 0.05.
We compared the expression profiles of C. difficile WT biofilms to C. difficile WT-B. fragilis mixed biofilms and C. difficile luxS biofilms to C. difficile luxS-B. fragilis mixed biofilms. A total of 45 genes were up-regulated (21) or down-regulated (24) in C. difficile WT (Fig. 6, Table 2), while 69 genes were differentially expressed in C. difficile luxS of which 34 were down-regulated and 35 up-regulated, during co-culture with B. fragilis (Fig. 6A, Table 3).
Eight up-regulated genes were specific to C. difficile WT co-culture (Table 2). Of these, four genes involved in fatty acid biosynthesis and metabolism: fabH encoding 3-oxoacyl-[acyl-carrier protein] synthase III, fabK encoding trans-2-enoyl-ACP reductase, accC encoding a biotin carboxylase (acetyl-CoA carboxylase subunit A), and accB encoding a biotin carboxyl carrier protein of acetyl-CoA carboxylase ( Table 2). 11 genes were down-regulated exclusively in WT however, these genes do not coincide with a specific metabolic pathway.
A total of 26 genes were differentially expressed in both C. difficile WT and C. difficile luxS when co-cultured with B. fragilis (Fig. 6A, Table S1). These include six up-regulated genes (accB, abfH, abfT, abfD, sucD and cat1) involved in carbon and butanoate metabolism, with cat1, which encodes succinyl-CoA:coenzyme A transferase, being the highest up-regulated gene for both C. difficile strains.
When C. difficile WT-B. fragilis mixed biofilms and luxS-B. fragilis mixed biofilms were compared to B. fragilis single biofilms, in contrast, a higher number of genes (266) were differentially expressed in B. fragilis when co-cultured with C. difficile (WT and luxS) ( Table S4, Fig. 6B). A total of 114 B. fragilis genes were found to be specific to C. difficile WT co-culture, with 56 of these up-regulated and 58 down-regulated (Table S2). Similarly, 91 genes were found to be specific to C. difficile luxS co-culture, with 56 of these up-regulated and 35 down-regulated (Table S3). Although distinct B. fragilis expression profiles were observed with the WT and luxS (Fig. 6B), there were no clear pathways identified in the datasets.
Whilst the highest up-regulated gene in both WT and LuxS co-cultures encodes a putative virus attachment protein, no other viral genes were shown to be up-regulated (Table S4). Genes encoding iron containing proteins desulfoferrodoxin and rubrerythrin were highly up-regulated in both conditions. Interestingly, multiple copies of fecR, a key regulator for the ferric citrate transport system 47 , and ferrous iron transport protein B were also up-regulated. Additionally, a number of other metabolic pathways were up-regulated, including four genes (encoding 3-isopropylmalate dehydratase small subunit, 3-isopropylmalate dehydrogenase, Galactokinase and 3-isopropylmalate dehydratase large subunit) involved in valine, leucine and isoleucine biosynthesis and C5-branched dibasic acid metabolism. It should be noted that many of the up-regulated genes were hypothetical proteins of unknown function. Similarly, several metabolic pathways were down-regulated in both co-culture conditions. These include six genes involved in carbon metabolism, four genes involved in alanine, aspartate and glutamate metabolism, and four genes involved in the biosynthesis of amino acids, although these appear to be single genes rather than specific pathways.

Discussion
Inter-bacterial interactions within gut communities are critical in controlling invasion by intestinal pathogens. Quorum sensing molecules such as AI-2 are instrumental in bacterial communication, especially during formation of bacterial communities 31,36,37,[48][49][50] . C. difficile produces AI-2, although the mechanism of action of LuxS/ AI-2 in C. difficile, particularly within a biofilm community is unclear. We report here that the C. difficile LuxS/ AI-2 plays an important role in the formation of single and multi-species communities. In C. difficile biofilms, LuxS mediates the induction of prophages, which likely contributes to the biofilm structure. Whereas, in a mixed biofilm of C. difficile and the intestinal commensal and pathogen, B. fragilis, LuxS likely triggers the induction of differential metabolic responses in B. fragilis, that leads to growth inhibition of C. difficile. To our knowledge, this is the first time dual species RNA-seq 51 has been applied to analyse interactions between anaerobic gut bacteria in an adherent biofilm community.
Bacterial biofilms contain a number of extracellular components that make up their complex structure including extracellular DNA (eDNA), a key component that binds together bacteria within a community. Autolysis is a common mechanism by which eDNA is released from bacterial cells 52 . In bacteria such as Staphylococcus aureus and Pseudomonas aeruginosa, eDNA is generated through the lysis of subpopulations within a biofilm, under the control of quorum sensing [52][53][54][55] . In C. difficile luxS mutant (luxS) biofilms, we observe reduced induction of two C. difficile prophages compared with the WT. These phage loci were conserved in several C. difficile strains with the Region 2 encoding a phiC2-like, phi-027 phage [56][57][58] . Given that it has previously been shown that eDNA is a major component of C. difficile biofilms 27,30 , it is likely that phage-mediated bacterial cell lysis and subsequent DNA release help build a biofilm. Indeed, given the detection of eDNA in the luxS biofilms, there are likely other unknown eDNA release mechanisms during C. difficile biofilm formation.
Phages are also known to control biofilm structure in some organisms: a filamentous phage of P. aeruginosa was reported to be a structural component of the biofilm 59 and an AI-2 induced phage mediated the dispersal of Enterococcus faecalis biofilms 60 . Although attempts to visualise phages from C. difficile biofilms with transmission electron microscopy (data not shown) were unsuccessful, we cannot rule out the possibility that C. difficile phages may directly influence the biofilm structure. While the precise mechanisms by which LuxS/AI-2 controls phage www.nature.com/scientificreports www.nature.com/scientificreports/ induction are yet to be elucidated, AI-2 appears to be signalling through a yet unidentified AI-2 receptor in C. difficile. Phage-mediated control of biofilms may in part explain the variation observed in biofilm formation between different C. difficile strains 61 .
The human gut hosts a variety of bacterial species, which compete or coexist with each other. It is likely that the bacteria occupying this niche form multi-species bacterial communities in association with the mucus layer. Interactions within such communities are important in gaining a better understanding of phenomena such as www.nature.com/scientificreports www.nature.com/scientificreports/ 'colonisation resistance' which prevents pathogens such as C. difficile from establishing an infection 14 . Whilst sequencing studies have identified members of the Bacteroides genus as being associated with gut colonisation resistance to C. difficile, the mechanisms have remained elusive 10 . A recent study demonstrated that production of the enzyme: bile salt hydrolase, is responsible for the inhibitory effect of B. ovatus on C. difficile 62 . This study reported that in the presence of bile acids, cell free supernatants for B. ovatus were capable of inhibiting the growth of C. difficile whereas in the absence of bile acids, C. difficile growth was promoted. Since bile acids are not supplemented into our media, a different mechanism is likely responsible for B. fragilis mediated inhibition of C. difficile. Also, the growth restraining effects of B. fragilis on C. difficile were evident only within mixed biofilms, not in planktonic culture or with culture supernatants. While it is likely that cell-cell contact is essential for the inhibitory effect, we cannot exclude involvement of an inhibitory secreted molecule that accumulates to a higher concentration within a biofilm environment, or that B. fragilis has a competitive growth advantage in a biofilm environment.
A dual species RNA-seq analysis performed to understand the interactions between the two bacterial species, showed that largely all the differentially expressed genes mapped to distinct metabolic pathways. Overall, a higher number of genes were modulated in B. fragilis as compared to C. difficile strain during co-culture, which is in line with the growth characteristics observed. Carbon and butanoate metabolism pathways were induced in C. difficile strains in response to co-culture (accB, abfH, abfT, abfD, sucD and cat1). An up-regulation of the succinate utilisation genes was also recently reported in C. difficile 630 microfermenter biofilm cells as compared to planktonic cells 63 . As B. fragilis is known to produce succinate 64 , it is likely that the upregulation in these pathways results from the increased levels of succinate in the culture medium. However, since gut microbiota-produced succinate promotes C. difficile growth in vivo 17 , it is unlikely that these changes are directly responsible for the observed inhibition of C. difficile. However, bacteria utilise carbohydrates in a sequential manner 65 . Consistent with this, we observed a down-regulation of genes important for the utilisation of pyruvate such as bcd2 and idhA encoding for butyryl-CoA dehydrogenase and (r)-2-hydroxyisocaproate dehydrogenase respectively. A down-regulation of sugar fermentation pathway genes was also observed by Poquet et al. in single species C. difficile biofilms compared to planktonic cultures 63 . Such a shift in metabolism could allow B. fragilis to fully consume other metabolites, and thus enabling it to outcompete C. difficile.
Additionally, it is interesting to note that a number of copies of the ferric citrate transport system regulator, fecR, are up-regulated in B. fragilis during C. difficile co-culture. The ferric citrate transport system is an iron uptake system that responds to the presence of citrate 47,66 . Analysis of the C. difficile genome using BLAST (NCBI) showed that C. difficile does not possess this iron uptake system. Given the evidence that ferric citrate is an iron source in the gut 47,67 , B. fragilis may have an advantage over C. difficile in sequestering iron, and thus preventing C. difficile colonisation. Although the clear modulation of metabolic pathways strongly suggest a competitive advantage of B. fragilis over C. difficile, it is possible that the genes with unknown functions that are differentially expressed in B. fragilis (Table S1), encode pathways for the production of yet to be identified small inhibitory molecules.
The LuxS/AI-2 quorum sensing system is known to have a cross-species signalling role in many bacteria 45,68 . While all sequenced C. difficile strains produce AI-2, only selected strains of B. fragilis have the ability to produce AI-2 69 . A recent study showed that Ruminococcus obeuem inhibited Vibrio cholerae in the gut via LuxS/AI-2 mediated downregulation of V. cholerae colonisation factors 70 . Also, AI-2 produced by engineered E. coli was No Gene ID log2FoldChange Gene Annotation  www.nature.com/scientificreports www.nature.com/scientificreports/ reported to influence firmicutes/bacteroidetes ratios in microbiota treated with streptomycin 71 . Our data show the involvement of LuxS/AI-2 in the B. fragilis-mediated C. difficile growth inhibition. As the B. fragilis strain used in this study does not produce AI-2, it is likely that B. fragilis responds differentially to AI-2 produced by C. difficile. Similar to the WT, most transcriptional changes were in metabolic pathways, although specific sets of genes were modulated in the luxS mutant. Modulation of prophage genes was not seen, unlike single biofilm cultures, indicating different dominant mechanisms at play in a multi-species environment.
It was interesting to note that the trehalose utilisation operon, which provides a growth advantage to C. difficile against other gut bacteria 72 , was upregulated in both the single luxS biofilms (Table 1) and luxS co-cultured with B. fragilis ( Table 3). The upregulation of a phosphotransferase system component and treA (both in the same operon), likely enables increased utilisation of trehalose, providing an additional carbon source. Like glucose, trehalose acts as an osmoprotectant 73 and its presence within the cell may help maintain protein conformation during cellular dehydration. It is possible that trehalose plays a role in building C. difficile biofilms, as reported for Candida biofilms 74 . Our preliminary studies with exogenous trehalose levels similar  Table 3. Genes up and down-regulated in luxS during co-culture with B. fragilis relative to the WT C. difficile monoculture. Note: P-adjusted value ≤ 0.05.
www.nature.com/scientificreports www.nature.com/scientificreports/ to those used by Collins et al. (2018) showed an inhibitory effect on biofilm formation by both WT C. difficile and luxS, although no differential effects were observed between luxS and WT (Fig. S9). However, further investigations into accumulation of trehalose within biofilms are required to clarify the role of trehalose in luxS mediated biofilm formation.
In conclusion, we report that C. difficile LuxS/AI-2 may play a key role in building C. difficile communities through mediating prophage induction, and subsequent accumulation of eDNA. In mixed communities, C. difficile AI-2 likely signals to B. fragilis to induce an altered metabolic response, enabling it to outgrow C. difficile. Further studies are required to understand the precise AI-2 sensing pathways involved.

Materials and Methods
Bacterial strains and media. Two bacterial species were used in this study -C. difficile strain: B1/ NAP1/027 R20291 (isolated from the Stoke Mandeville outbreak in 2004 and 2005), and Bacteroides fragilis (a clinical isolate from a biliary stent kindly provided by Dr Claudia Vuotto and Dr Gianfranco Donelli, Rome). A luxS Clostron R20291 mutant described previously in Dapa et al. 27 was used in this study. Both species were cultured under anaerobic conditions (80% N 2 , 10% CO 2 , 10% H 2 ) at 37 °C in an anaerobic workstation (Don Whitley, United Kingdom) in BHIS, supplemented with L-Cysteine (0.1% w/v; Sigma, United Kingdom), yeast extract (5 g/l; Oxoid) and glucose (0.1 M).
Biofilm formation assay. Biofilms were grown as per the previously published protocol 27 . Overnight cultures of C. difficile were diluted 1:100 in fresh BHIS with 0.1 M glucose. 1 ml aliquots were pipetted into 24-well tissue culture treated polystyrene plates (Costar), and incubated under anaerobic condition at 37 °C, for 6-120 h. Tissue culture plates were pre-incubated for 48 h prior to use. The plates were wrapped with parafilm to prevent liquid evaporation.

Measurement of biofilm biomass.
Biofilm biomass was measured using crystal violet (CV) 27 . After the required incubation, each well of the 24-well plate was washed with sterile phosphate buffer saline (PBS) and allowed to dry for a minimum of 10 mins. The biofilm was stained using 1 ml of filter-sterilised 0.2% CV and incubated for 30 mins at 37 °C, in anaerobic conditions. The CV was removed from each well, and wells were subsequently washed twice with sterile PBS. The dye was extracted by incubated with 1 ml methanol for 30 mins at room temperature in aerobic conditions. The methanol-extracted dye was diluted 1:1, 1:10 or 1:100 and OD 570 was measured with a spectrophotometer (BMG Labtech, UK).
For bacterial cell counts from the biofilm, the planktonic phase was removed and wells were washed once using sterile PBS. The adherent biofilms were then detached by scrapping with a sterile pipette tip and re-suspended into 1 ml PBS. Serial dilutions were made and plated onto BHIS plates to determine the CFU present in the biofilm.
Co-culture biofilm assay. For generation of co-culture biofilms, both C. difficile and B. fragilis were diluted to an OD 600 of 1. Both species were diluted 1:100 into fresh BHIS with 0.1 M glucose. Biofilms assays were performed as described above and measured by a combination of CV staining and CFU. To distinguish between C. difficile and B. fragilis, serial dilutions used for determining CFU were plated on BHIS plates additionally supplemented with C. difficile selective supplement (Oxoid, UK). Colonies can be differentiated by size and colony morphology as B. fragilis form very small colonies. exogenous addition of DpD. To analyse the potential signalling role of AI-2, biofilm assays were performed as described above in BHIS with 0.1 M glucose containing 1 nM, 10 nM, 100 nM, or 1 µM of chemically synthesised, exogenous 4,5-Dihydroxy-2,3-pentanedione (Omm Scientific, Texas USA) for both C. difficile WT and LuxS. BHIS with 0.1 M glucose was used as a control. Samples were washed and stained with 0.2% CV at either 24 h or 72 h. AI-2 Assay. The AI-2 bioluminescence assay was carried out essentially as described by Bassler et al. 1993 75 .
The V. harveyi reporter strain BB170 was grown overnight in LB medium before being diluted 1: 5000 in Autoinducer Bioassay (AB) medium containing 10% (v/v) cell-free conditioned medium collected from either planktonic or biofilm C. difficile cultures (in BHI) and allowed to grow at 30 °C with shaking. AB medium containing 10% (v/v) from V. harveyi BB120 was used as a positive control, and 10% (v/v) sterile BHI medium as a blank. Luminescence was measured every hour using a SPECTROstar Omega plate reader. Induction of luminescence was taken at the time when there was maximal difference between the positive and negative controls (usually 2-5 h) and is expressed as a percentage of the induction observed in the positive control.
RNA-seq. Biofilms were grown for 18 h in BHIS + glucose, supernatants were removed and attached biofilms were washed with 1 ml PBS. Biofilms were disrupted and RNA was extracted using Trizol (Invitrogen, UK). 5 µg of extracted RNA was treated with RiboZERO TM (Illumina, UK) according to the manufacturer's protocol to deplete rRNA. cDNA libraries were prepared using TruSEQ LT (Illumina, UK) according to the manufacturer's instructions. Briefly, samples were end-repaired, mono-adenylated, ligated to index/adaptors. Libraries were quantified by bioanalyzer and fluorometer assay. The final cDNA library was prepared to a concentration of 10-12 pM and sequenced using paired end technology using a version-3 150-cycle kit on an Illumina MiSeq TM (Illumina, UK). (2019) 9:9903 | https://doi.org/10.1038/s41598-019-46143-6 www.nature.com/scientificreports www.nature.com/scientificreports/ RNA-seq analysis. The paired-end sequencing reads from RNA-seq experiments were mapped against the appropriate reference genome (NC_013316 for C. difficile and a de novo assembly using RNA SPAdes v3.9 with default settings 76 from the RNA-sequence reads for B. fragilis [Accession number PRJEB29695). The first read was flipped into the correct orientation using seqtk v1.3 (https://github.com/lh3/seqtk) and the reads were mapped against the reference genome using BWA v0.7.5 with the 'mem' alignment algorithm 77 . BAM files were manipulated with Samtools v0.1.18 using the 'view' and 'sort' settings 77 . Sorted BAM and GFF (general feature format) files were inputted into the coverageBed tool v2.27.0 with default settings 78 to gain abundance of each genomic feature. The R package DESeq2 was used with default settings to calculate differential gene expression using a negative binomial distribution model 42 . The data was filtered by applying a cut-off of 1.6 for the fold change and 0.05 for the adjusted p-value. All sequencing reads were submitted to the European Bioinformatics Institute (Accession numbers E-MTAB-7486, E-MTAB-7523 and PRJEB29695).
As analysis by BLAST (NCBI) demonstrated species specificity for mapping, co-culture samples were mapped to each species reference separately. Initial mapping of the B. fragilis strain to a published reference proved unsuccessful, offering a poor rate of alignment of 60%. As the B. fragilis strain has not been previously sequenced, and because we were not successful in generating high quality genome sequence, a reference was generated from RNA library of B. fragilis using the software rnaSPAdes v3.9 79 and annotated using Prokka v1.11 (default settings) 80 . The reads from each condition were mapped to their respective reference sequence using BWA v0.7.5 ('mem' algorithm) 77,81 and counted using coverageBed v2.27.0 78 . Metabolic pathways in C. difficile were identified using the KEGG mapper 82 , a tool that identifies the function of genes in a published genome. As the B. fragilis strain used in this study does not have a published reference genome, blastKOALA 83 was used to search for gene homology within metabolic pathways. Heatmaps were generated from normalised gene expression data outputted from DESeq2, using the online tool Heatmapper 84 using the default settings. pCR analysis. 16S PCRs were performed using the universal 16S rRNA bacterial primers 27F and 1392R (Table S2). Primers were constructed for prophage genes CDR20219_1208 and CDR20291_1436 (Table S2) to confirm the presence of prophage within C. difficile biofilms. PCR was carried out using Fusion High-Fidelity DNA polymerase (NEB, USA) following the manufacturer's protocol. Samples were heated to 95 °C for 5 mins followed by 35 cycles of: 95 °C for 30 seconds, 51 °C for 30 seconds and 72 °C for 30 seconds, after which samples were heated to 72 °C for 10 mins. eDNA quantification. eDNA was extracted from C. difficile biofilms grown in a 24-well plate as described above, using a protocol described in Rice et al. 53 . Briefly, the plate was sealed with parafilm and chilled at 4 °C for 1 hour. 1 μl 0.5 M EDTA was added to each well and incubated at 4 °C for 5 mins. The medium was removed and biofilms were resuspended in 300 μl 50 mM TES buffer (50 mM Tris HCl/10 mM EDTA/500 mM NaCl). The OD 600 was measured to determine biofilm biomass and the tubes were centrifuged at 4 °C at 18,000 g for 5 mins, and was used to normalise the eDNA values. 100 μl of supernatant was transferred to a tube of chilled TE buffer (10 mM Tris HCl/ 1 mM EDTA) on ice. DNA was extracted using an equal volume of phenol/chloroform/isoamyl alcohol three times. 3 volumes of ice-cold 100% ethanol and 1/10 volumes 3 M sodium acetate were added to the aqueous phase to precipitate the DNA. The DNA pellet was washed with 1 ml ice-cold 70% ethanol, dissolved in 20 μl TE buffer, quantified by Qubit fluorometer (Thermo Fisher). spore counts. To determine the number of spores, adherent biofilms were resuspended in PBS and treated at 65 °C for 25 mins as previously described 85 . Untreated and heat-treated samples were serial diluted and plated on BHIS and BHIS-TC agar (supplemented with 0.1% sodium taurocholate, Sigma-Aldrich, UK). No bacteria were obtained from the heat-treated samples plated on BHI (without sodium taurocholate). The CFU/ml obtained from heat-treated samples plated on BHIS-TC plates represent heat-resistant spores, and the CFU/ml obtained from untreated samples plated on BHIS plates represent the total cell counts. statistical analysis. All experiments were performed in triplicate, with at least three independent experiments performed. Paired or unpaired student's t-test was used to determine if differences between two groups were significant, and one way-ANOVA was used to compare multiple groups. Mann-Whitney U tests were used to compare non-parametric data. Fisher's exact t-test was used to confirm the enrichment of differently regulated genes in prophage regions.