A simple coculture system shows mutualism between anaerobic faecalibacteria and epithelial Caco-2 cells

Most gut bacteria are obligate anaerobes and are important for human health. However, little mechanistic insight is available on the health benefits of specific anaerobic gut bacteria. A main obstacle in generating such knowledge is the lack of simple and robust coculturing methods for anaerobic bacteria and oxygen-requiring human cells. Here, we describe the development of a coculture system for intestinal Caco-2 cells and an anaerobic symbiont, Faecalibacterium prausnitzii, making use of 50 mL culture tubes. F. prausnitzii was grown in 40 mL YCFAG-agar with glass-adhered Caco-2 cells placed on top in 10 mL DMEM medium. Grown for 18–36 h in a humidified incubator at 37 °C and 5% CO2, coverslip-attached Caco-2 cells promoted growth and metabolism of F. prausnitzii, while F. prausnitzii suppressed inflammation and oxidative stress in Caco-2 cells. F. prausnitzii did not compromise Caco-2 cell viability. Exogenously added porcine mucin also promoted growth of F. prausnitzii, suggesting that it may be part of the mechanism of Caco-2-stimulated growth of F. prausnitzii. This ‘Human oxygen-Bacteria anaerobic‘ (HoxBan) coculturing system uniquely establishes host-microbe mutualism of a beneficial anaerobic gut microbe in vitro and principally allows the analysis of host-microbe interactions of pure and mixed cultures of bacteria and human cells.

The human gut microbiome is increasingly recognized as an important determinant for human health, affecting a variety of gut, metabolic, neurological and psychological disorders 1 . Gut microbiota provide essential nutrients and anti-inflammatory compounds to the host and confine expansion of pathogens [2][3][4] . High-throughput sequencing techniques have uncovered the high complexity of the gut microbiome and the composition changes during ageing and disease 5,6 . The healthy gut microbiome contains 500-1,000 different bacterial species and their collective genomes (metagenome) encode at least a 100-fold more genes compared with the human genome [7][8][9] . Only a small fraction of these bacterial species are cultured in vitro and even more challenging is to coculture gut bacteria and human cells 10,11 . The main obstacle in a host-microbiome coculture system is that most (> 90%) gut bacteria are obligate anaerobes that die quickly when exposed to atmospheric conditions (21% O 2 ), while human cells depend on oxygen.
Faecalibacterium prausnitzii is an obligate anaerobe that may represent up to 25% of all bacteria in the healthy gut 12 . F. prausnitzii-excreted products and cell extracts suppress inflammatory signaling in intestinal epithelial (Caco-2) cells in vitro, as well as in 2,4,6-trinitrobenzenesulphonic acid (TNBS)-induced colitis in mice in vivo 13 . Moreover, human intestinal inflammation is associated with decreased numbers of F. prausnitzii and low numbers of this bacterium predispose for post-operative ileal recurrence of Crohn's disease [13][14][15][16] . Among anaerobes, F. prausnitzii has the unique ability to grow close to the intestinal epithelium in the oxic-anoxic interphase of the gut [17][18][19] . However, it remains elusive whether direct mutualism exists between this gut microbe and intestinal epithelial cells.

Results and Discussion
The 'Human oxygen-Bacteria anaerobic' (HoxBan) system developed in this study establishes coculturing of glass-adherent human cells in liquid medium and anaerobic bacteria in solid agar medium for over 24 h and allows the analysis of cell growth, transcriptome and exo-metabolome of cocultured cells. A detailed protocol is given in the supplementary/online section. In short, hand warm (~40 °C) agar medium was inoculated with F. prausnitzii starter cultures and aliquots of 40 mL were allowed to solidify in 50 mL Falcon centrifugation tubes. Caco-2 cells grown on coverslips were placed up-side-down on top of the F. prausnitzii-containing YCFAG agar and overlaid with DMEM medium. The HoxBan coculture tubes (schematically drawn in Fig. 1a) were placed either with a loose (air-open) or tightly-closed cap (air-closed) in a standard humidified incubator at 37 °C, 5% CO 2 and atmospheric O 2 . F. prausnitzii colony formation, transcriptional adaptations of Caco-2 cells and excreted metabolites in the liquid medium were analyzed after 18-36 h of coculture.
Within 18 h, F. prausnitzii formed colonies throughout the 40 mL YCFAG-agar. In the absence of Caco-2 cells, no F. prausnitzii colonies formed in the top agar layer, most likely due to penetration of intolerable amounts of oxygen (Fig. 1b, left 2 panels). In contrast, clear and bigger F. prausnitzii colonies appeared close to the coverslip-attached Caco-2 cells, both in air-closed and air-open tubes (Figs 1b, 2 panels on the right). Notably, F. prausnitzii continued to expand over a total coculture time of 36 h (Fig. 1b, 2 bottom panels on the right). In a similar way, human colonic DLD-1 cells promoted growth of F. prausnitzii in the top agar layer close to the coverslip, though bacterial growth was less pronounced compared to the cocultures with Caco-2 cells (Fig. 1c). In contrast to both intestine-derived cell lines, HepG2 (human hepatoma) cells did not stimulate F. prausnitzii growth (Fig. 1c), indicating that this effect is cell type-specific.
Caco-2 cells were viable and actively dividing even after 24 h of (co)culturing in the HoxBan system (Fig. 2a). Trypan blue staining revealed that over 80% of the Caco-2 cells were viable after overnight culture in the HoxBan system in the absence or presence of F. prausnitzii (Fig. 2b). Viability of Caco-2 cells in the HoxBan system without F. prausnitzii was significantly lower when compared to either Caco-2 cells grown in 12-wells culture plates or HoxBan system with PBS-buffered agar instead of YCFAG agar. These data imply that both the position of the glass-adherent Caco-2 cells and the contact with bacterial growth medium slightly reduce Caco-2 cell viability, but that this is not affected by the absence or presence of F. prausnitzii. Caco-2 cells harvested after overnight culture with and without F. prausnitzii revealed that both Il-1β and iNOS mRNA levels were significantly reduced (p < 0.05) in Caco-2-F. prausnitzii cocultures compared to Caco-2 mono-cultures (Fig. 2c). A similar effect was observed for the oxidative stress marker heme oxygenase 1 (HO-1; Fig. 2c), In contrast, mRNA levels of villin (a marker for intestinal epithelium), claudin-1 (marker for barrier function) and the multidrug resistance protein-1 (cytoprotective substrate pump) in Caco-2 cells were all similar when cultured in the absence or presence of F. prausnitzii (Fig. 2d). These results indicate that expansion of F. prausnitzii close to the Caco-2 cells has both anti-inflammatory and anti-oxidant effects, while intestinal barrier functions are maintained.
F. prausnitzii has been shown to be able to grow close to the intestinal epithelium adhered to, or even within the mucus layer 9,18,20 . Caco-2 cells do not express the most dominant intestinal mucin, mucin-2 (encoded by MUC2), but have been shown to express MUC3 and MUC5A/C 21,22 . To test directly whether mucin may stimulate F. prausnitzii growth, porcine mucin type II with 0.8% agar was added on top of the F. prausnitzii-containing YFCAG agar in the absence and presence of Caco-2 cells. Mucin alone strongly promoted F. prausnitzii growth in the upper layer of the YCFAG agar, which was not affected by the presence of Caco-2 cells (Fig. 3). This indicates that indeed mucin may contribute to the stimulation of F. prausnitzii growth by Caco-2 cells, which is a remarkable observation as this gut bacterium is not known to be able to use mucin as carbon and/or energy source 23 .
To obtain a comprehensive overview of the mutual metabolic effects of F. prausnitzii and Caco-2 cells, we performed a metabolome analysis on the liquid medium after 18 hours of (co)culturing Caco-2 cells and/or F. prausnitzii, which included short chain fatty acids (SCFAs), hydrocarbons, lipids and amino acids (Fig. 4a). Systematic exo-metabolic changes occurring in different culture condition were assessed using principal component analysis (PCA) (Fig. 4b). The first principal component (PC1) accounts for 49.7% of the total variance and separates the liquid HoxBan culture medium and the Caco-2 monoculture from the F. prausnitzii monoculture and the two Caco-2-F. prausnitzii cocultures. This indicates that F. prausnitzii has the strongest effect on the level of metabolites in the culture medium. Main determinants of PC1 are the SCFAs butyrate and formate that are produced by F. prausnitzii (Fig. 5a,b). Other metabolites that associate with PC1 are the essential amino acid methionine and the amino acid-derivative N-acetyl aspartate. Concentrations of all these metabolites are enhanced in HoxBan cultures with F. prausnitzii. The second principal component (PC2) accounts for 16.8% of the variance in the data and separates the F. prausnitzii monoculture from the two cocultures. Metabolites that contribute to PC2 are formate, adenine and inosine (Fig. 5b-d). Formate levels significantly increased after coculturing F. prausnitzii with Caco-2 cells, which is in line with the enhanced bacterial biomass under these conditions. However, butyrate levels did not change upon coculturing F. prausnitzii with Caco-2 cells. Butyrate is a preferred energy source of intestinal epithelial cells 24 and our data suggests that Caco-2 cells consume part of the butyrate produced by F. prausnitzii. In turn, this may contribute to the suppression of inflammatory and oxidative stress markers in Caco-2 cells (Fig. 2c).
Together with adenine and inosine, also concentrations of xanthosine and 5-methylthioadenosine were strongly reduced in medium of Caco-2-F. prausnitzii cocultures compared to the two monocultures (Fig. 5e,f). These compounds of purine metabolism are required for DNA (cell proliferation) and ATP (energy) synthesis. Their depletion from the medium is likely due to the stimulated growth of F. prausnitzii, although a pro-proliferative effect on Caco-2 cells cannot be excluded at this point. On PC2, a slight separation between the open and closed HoxBan Caco-2-F. prausnitzii cocultures was observed (Fig. 4b), however, this could not be attributed to significant changes in single metabolites. This indicates that even in the closed condition, 10 mL of DMEM medium carries sufficient oxygen to support the growth of the coverslip-attached Caco-2 cells, which is further supported by the suppressed expression of hypoxia-sensitive HO-1 25 in both coculture conditions (Fig. 2c). These data show that coculturing Caco-2 cells with F. prausnitzii leads to a unique profile of excreted and consumed metabolites that is not simply the cumulative result of the individual cell types, indicating that these cells modify each other's metabolism.
Thus, the HoxBan coculture system presented here demonstrates for the first time mutualism between oxygen-requiring intestinal epithelial cells (Caco-2) and an obligate anaerobic gut bacterium (F. prausnitzii) while both cell types are viable. Several other systems have been developed to study the interaction between human   13,20,26 . However, these systems did not allow the analysis of the effect of human cells on F. prausnitzii, which were evidently observed in the HoxBan system. The simplicity of the HoxBan coculture system lies in the use of solid agar medium for growth of the anaerobic gut bacteria, overlaid with liquid medium exposed to air for human cells. The HoxBan is a robust system that can be implemented in almost any molecular biology research laboratory with access to an anaerobic facility and a tissue culture cabinet and incubator. As such, it has great potential to support research to understand the communication between gut microbes and their host. The HoxBan system is readily adaptable to coculture   any other anaerobic gut bacterium, as well as complex mixtures of bacteria, with adherent -and potentially also non-adherent-human cell lines or primary cells. Applications of the HoxBan coculture system for other cells than Caco-2 and/or F. prausnitzii may require specific optimization in culture conditions, but it holds the universal principle of coculturing oxygen-requiring human cells together with obligate anaerobic bacteria that will foster our understanding of the role of gut bacteria in human health and disease.
Protocol for the HoxBan coculture. Starting in the anaerobic chamber, one (1) mL of the overnight F. prausnitzii preculture was used to inoculate 1000 mL of freshly-autoclaved and cooled-down (~40 °C) YCFAG agar medium containing 1% agar. Aliquots of 40 mL of this inoculum were transferred to sterile 50 mL falcon tubes and allowed to solidify for 30 minutes. The F. prausnitzii-inoculated Falcon tube cultures were transferred to a tissue culture cabinet at ambient air and Caco-2 cells on coverslips were placed (up-side-down) on top of the agar and overlaid with 10 mL pre-warmed (37 °C) DMEM medium (without antibiotics), after which the co-cultures were placed in a humidified incubator at 37 °C and 5% CO 2 for 18-36 hours. The screw caps of the Falcon tubes were either tightly closed (to maintain maximum anaerobic conditions for F. prausnitzii) or kept loosely tightened (to allow oxygen exposure for Caco-2 cells). Control conditions were: 1) YCFAG-agar without F. prausnitzii inoculum; 2) F. prausnitzii-inoculated cultures with coverslips without Caco-2 cells and 3) F. prausnitzii-inoculated cultures with human HepG2 cells on coverslips (growth conditions detailed below). Experiments were performed 3 times (N = 3) in triplicate tubes for each condition, with a total of 9 tubes per condition.
F. prausnitzii growth rim visualization. Visualization of F. prausnitzii colony formation in the YCFAG agar close to the coverslips with or without Caco-2 cells was performed using a digital camera (Canon EOS 450D) and the obtained images were processed using Digital Photo Professional software (Canon) without any qualitative and quantitative changes to the raw images.
Mucus layer preparation. The mucus layer was prepared essentially as described before by boiling Milli-Q water containing 5% porcine mucin type II (Sigma Aldrich, St. Louis, MO, USA) and 0.8% agar 20 . The solution was autoclaved after adjusting the pH to 6.8. One (1) mL of mucin solution was allowed to solidify on top of the YCFAG-agar, after which coverslips with or without Caco-2 cells and/or DMEM were applied followed by standard HoxBan culture conditions. Ki-67 staining). At the end of the coculture experiment, Caco-2 cells adherent to coverslips were removed from the HoxBan coculture tubes and total RNA was isolated using Trizol according to the suppliers protocol (Sigma-Aldrich). RNA concentrations were determined using a NanoDrop 1000 spectrophotometer (Thermo Fisher Scientific, Wilmington, DE, USA). Reverse transcription polymerase chain reaction (RT PCR) was performed as described 28 . Quantitative PCR (qPCR) for the inducible isoform of nitric oxide synthase (iNOS-NOS2), interleukin-1 beta (IL-1β ) and heme oxygenase 1 (HO-1) were performed. Primers (Invitrogen) and probes (Eurogentec) were designed using Primer Express 2.0 software (Applied Biosystems). Details of primers and probes are given in Supplementary Table S1. Q-PCR conditions were as described by Blokzijl et al. 28  staining using a rabbit polyclonal antibody directed against Ki-67 (dilution 1:1000 60 minutes at 25 °C) based on manufacturers protocol (Monosan; Netherlands).

Harvesting of Caco-2 cells and analysis (Q-PCR and
Trypan blue staining. Trypan blue solution (0.2%) was added for one minute to the Caco-2 cells adherent to coverslips. After removing the trypan blue solution, the cells were fixed immediately with 4% paraformaldehyde (10 minutes at 4 °C), rinsed 4 times with PBS and mounted on glass slides. Trypan blue-negative (viable) cells were quantified microscopically.
Metabolome analysis by Liquid Chromatography-Mass Spectrometry. At the end of the (co)culture experiments the liquid medium on top of the HoxBan cultures was collected (~10 mL) and polar metabolites were analyzed and quantified by ultra-performance liquid chromatography (UPLC Acquity, Waters, Manchester, UK) coupled in line with a quadrupole-time-of-flight hybrid mass spectrometer (Synapt G2, Waters, Manchester, UK) as previously reported 30 . All materials used in the UPLC-MS experiments were purchased from Sigma-Aldrich (Germany) and were of analytical grade or higher purity. For the analysis of targeted metabolites, data were processed using TargetLynx (Waters) while for untargeted analysis MarkerLynx (Waters) was used to integrate and align MS data points and convert them into exact mass retention time pairs. The identity of metabolites was established by comparison of accurate mass measurements and tandem mass spectrometry information against our in-house database and/or online databases 31,32 .
Short Chain Fatty Acid analysis. For short chain fatty acid (SCFA) analysis, including lactate, formate, butyrate and acetate, a HPLC Ion Chromatography system (Metrohm AG, Herisau, Switzerland) with a conductivity detector was used as described before 19 .
Statistics. Principal Component Analysis (PCA) was performed on all detected metabolites and SCFAs by using MetaboAnalyst 33 . Before PCA, data was normalized by the sum, log transformed and then scaled by using pareto scaling.
One-way Anova test was used to detect metabolites that were significantly different between groups. Hierarchical Clustering Analysis (HCA) was then performed by using MetaboAnalyst 33 on the top 25 metabolites ranked by the Anova Test. The similarity measure was obtained by applying the Spearman's rank correlation. The clustering algorithm used was the Ward's linkage. Differences in the dead/alive cell were calculated using Chi-square test. Differences in the gene expressions were assessed by using the Mann-Whitney U test, tests were two-tailed and p-values of 0.05 or lower were considered significant. Tests were performed with PASW Statistics 22 (SPSS, USA).