A microfluidics-based in vitro model of the gastrointestinal human–microbe interface

Changes in the human gastrointestinal microbiome are associated with several diseases. To infer causality, experiments in representative models are essential, but widely used animal models exhibit limitations. Here we present a modular, microfluidics-based model (HuMiX, human–microbial crosstalk), which allows co-culture of human and microbial cells under conditions representative of the gastrointestinal human–microbe interface. We demonstrate the ability of HuMiX to recapitulate in vivo transcriptional, metabolic and immunological responses in human intestinal epithelial cells following their co-culture with the commensal Lactobacillus rhamnosus GG (LGG) grown under anaerobic conditions. In addition, we show that the co-culture of human epithelial cells with the obligate anaerobe Bacteroides caccae and LGG results in a transcriptional response, which is distinct from that of a co-culture solely comprising LGG. HuMiX facilitates investigations of host–microbe molecular interactions and provides insights into a range of fundamental research questions linking the gastrointestinal microbiome to human health and disease.

. Global pathway analysis (GeneGo MetaCore™ pathway analysis) of the genes that were differentially expressed in Caco-2 cells co-cultured with LGG under anaerobic conditions versus their corresponding LGG-free controls. The highlighted pathways are the top 50 altered pathways that are differentially expressed. The threshold parameters used were FC > 1.5 and P 85 < 0.01 (BtS). 'Total' depicts the total number of genes that are linked to the indicated pathway, 'In data' highlights the number of genes linked to the indicated pathway that were found to be differentially expressed, and 'FDR' presents the False Discovery Rate.

Rank Enriched Pathways
Total In Data FDR Supplementary Table 3. Gene Ontology (GO)-based pathway analysis highlighting altered biological processes in Caco-2 cells co-cultured with LGG growing under anaerobic conditions compared to LGG-free anaerobic controls. In addition to the top 100 altered GO processes, the statistically 95 significantly altered processes that exhibit altered expression in our data compared to the available in vivo data from human subjects following the administration of LGG are highlighted. For our investigation, we only included the GO processes covered by at least 10 genes in the microarray dataset, and we then computed the median gene expression levels across the process members for each of the identified GO processes. The statistical significance was ascertained using the empirical Bayes t 100 statistic (BtS). LGG grown under anaerobic or aerobic conditions in comparison to their respective controls. For our 155 investigation, we only included the GO processes covered by at least 10 genes in the microarray dataset, and we then computed the median gene expression levels across the process members for each of these GO processes. The significance was ascertained using the empirical Bayes t statistic (BtS).

EGR1
Transcription regulation, transcription factor activity for the regulation of cell proliferation and apoptosis 1,2 , anti-cancer effect 3-5 and IL-8 suppression 6

CCL2
Chemotactic factor that attracts monocytes and basophils and binds to the chemokine receptors CCR2 and CCR4 7

SLC9A1
Signal transduction, regulation of pH homeostasis, cell migration, cell volume 8 , and antiinflammatory effect 9

MYBL2
Anti-apoptopic function 18 , regulation of cell cycle and transcription 19 , and epithelial cell differentiation 19

PIM1
Cell survival, cell proliferation, cell growth, and signal transduction 20

PILRB
Receptors involved in the regulation of the immune system and cellular signaling 24,25

CDK9
Cell proliferation, regulation of the cell cycle, and transcription elongation factor 26,27

SOX4
Transcription factor, regulation of cell fate and apoptosis pathway, and prognostic marker in colon and gastric cancer 28,29 CEBPA Transcription factor, cell cycle regulation, and regulation of metallothioneins 30,31

IGFBP2
Regulation of IGF-mediated growth and developmental rates 33

GSTA1
Detoxification of carcinogens, drugs, environmental toxins, and products of oxidative stress 34,35 CTNNB1 Regulation of cell growth in addition to the creation and maintenance of epithelial layers [36][37][38] TPD52 Molecular marker in human cancer, and target for immunotherapy 39,40 Supplementary Notes: and this finding is analogous to the previous independent in vivo data 43 (Figure 3a and S5a,b and Table   1). Our RT-qPCR results corroborated this finding (Supplementary Figure 5b, P = 0.029, Student's t test, n = 3). ccl2 has previously been found to be differentially expressed following the exposure of 198 Caco-2 cells to live and heat-killed LGG 44 as well as after the stimulation of Caco-2 cells with another probiotic strain, i.e., Lactobacillus acidophilus NCFM 45,46 . Furthermore, the upregulation of ccl2 in bladder cells after the administration of LGG has also been identified in murine studies 47 Table 1 and Supplementary Table 1 Table 1) 19 . Therefore, our HuMiX-based results underpin the notion that molecular factors secreted by probiotic bacteria play beneficial roles in the regulation of 225 immune system responses and the maintenance of the epithelial barrier.
egr1, a gene that was found to be downregulated in vivo following LGG administration 54 , was also among the most significantly downregulated genes following the co-culture of Caco-2 cells with

228
LGG grown under anaerobic conditions in the HuMiX model (Figure 3a and Supplementary Figure   5a, c, Table 1 and Supplementary Table 1, FC > 2.5, P < 0.001, BtS). The RT-qPCR results for egr1 expression support the microarray data (Supplementary Figure 5b). Reduced egr1 expression has been 231 found to suppress tumor growth, proliferation and differentiation in colorectal carcinoma 56 , and LGG administration has been shown to reduce the incidence of chemically induced tumors in the large bowel of rodents 57 . Interestingly, reduced expression of IL-8, which was also corroborated by cytokine 234 profiling (Figure 3b), has been previously linked to a negative regulation of egr1 6, 58 . egr1 has been shown to bind to NF-κβ to form a protein complex which competitively binds to the IL-8 promoter and thereby leads to the suppression of IL-8 expression 6 . Our transcriptomic results, which highlight 237 the downregulation of egr1 in the colonic adenocarcinoma cell line Caco-2, support the notion that LGG exerts anti-cancer effects [3][4][5]  has become a key target gene for anti-cancerogenic therapeutic compounds in colon cancer cells 66 .

252
Finally, pim1, another gene that was identified to be upregulated after LGG administration to human subjects 43 and Caco-2 cells (FC > 1.5, P < 0.005, BtS), is a proto-oncogene used as a prognosis marker for colon cancer 67 . It has been described to regulate pathways related to cell growth 20 (Table 1). This gene is known to be activated through bacterial surface structures, such as lipoteichoic acid, or via Toll-like receptor 2 signaling 70 . It also plays a significant role in xenobiotic uptake and drug metabolism 22 Table 3, P < 0.05, BtS). Analogous to the global pathway expression changes found within healthy human mucosa exposed to LGG 43 , we also found alterations in pathways related to the 294 interferon response, calcium signaling and ion homeostasis. Among the top-ranked pathways, we also identified pathways linked to cellular homeostasis, aminopeptidase activity, desmosomes and tightjunction complexes, which together provide support for a beneficial role of LGG in enhancing 297 epithelial barrier function 73 . Furthermore, a decrease in the expression of genes related to inflammatory pathways (response to tumor necrosis factor and LPS) further reinforce the potential anti-inflammatory effects of LGG when co-cultured with human epithelial cells (Supplementary Table   300 3). In addition to the LGG strain tested in human clinical trials, the effect of the administration of three distinct Lactobacilli probiotic strains, namely Lactobacilli acidophilus L10, Lactobacilli casei CRL-431 and Lactobacilli plantarum, to human subjects has also been investigated previously 43,74 .

303
Interestingly, the comparison of the results of our global pathway analysis (Supplementary Table 2) with the global pathways that were found to be altered in human intestinal biopsies before and after  (Table 1). Interestingly, cebpa (P < 0.03), which has been directly linked to the regulation of mt2a (Figure 3a and Supplementary LGG grown under anaerobic conditions. Moreover, elf3 (P < 0.03), which plays a role in epithelial cell differentiation, has also been linked to the regulation of CCL20 in Caco-2 cells, which further corroborates our results (Figure 3b) 76 . elf3 is also a known direct positive regulator of ptgs2 (FC > 1.3, 324 P < 0.05, BtS), which has been previously shown to be induced by LGG in colonic epithelial cells 77 .
Both ptgs2 and elf3 were found to be upregulated in our data, but the expression of ptgs2 was distinct from the transcriptional response reported in the gnotobiotic piglet study, likely resulting from 327 differences between a single epithelial cell type and complex epithelial tissues 75 . to hypoxia to promote angiogenesis and cell growth 78,79 . hmgcs2 (FC > 1.5, P < 0.05), primarily expressed in differentiated colonic epithelium was also down-regulated and has also been shown to be down-regulated in colon cancer 80 . hmgcs2 regulates ketogenesis and energy metabolism 81 and has been 339 previously highlighted to be influenced by gut microbiota 82 . Additionally, hmgcs2 has been proposed as a prognostic marker for rectal cancer 83 . cyr61 (FC > 2, P < 0.08) has been linked to probioticmediated apoptosis during Salmonella infection in chicken cacae 84 and it has also been proposed as a 342 prognostic marker for colon cancer 85 . Additionally, cyr61 has also been suggested to mediate angiogenesis in colitis via the Substance P 86 pathway and this inflammatory disease-linked pathway 87 was also found to be among the top differentially regulated pathways in Caco-2 cells when co-cultured 345 with LGG growing under anaerobic conditions (Supplementary Note 6, Supplementary Table 6).
These results highlight the need for exploring the combinatorial effects triggered when human cells are co-cultured with microbial consortia. The HuMiX model is ideally suited to deconvolute such effects. Interestingly, igfbp7, a tumor-suppressor gene that has been found to be involved in the 399 suppression of colorectal tumors and linked to a favorable prognosis in colorectal cancer patients 102 , was upregulated when LGG was co-cultured under anaerobic conditions but downregulated under aerobic conditions. Furthermore, in the context of gastrointestinal cancers, the stimulated slc19a3 has 402 been suggested to represent a signal of tumor growth arrest 103 . In this context, it is interesting to note that we found the upregulation of slc19a3 in Caco-2 cells exclusively following their co-culture with

Supplementary
LGG grown under anaerobic conditions (Supplementary Figure 8). Downregulation of the irs2 gene 405 was observed in Caco-2 cells co-cultured with LGG growing under anaerobic conditions, whereas the opposite expression pattern was observed in Caco-2 cells co-cultured with aerobically growing LGG ( Figure 5a). irs2 is a known oncogene in colorectal cancer and an indicator of cancer cell growth 104,105 . 408 Regulation of irs2 expression by probiotic Lactobacilli has also been highlighted previously in studies involving human subjects 43 . In this context, the expression of sulf2, which is known for its role in colon cancer progression, migration and invasion 106 , was found to be downregulated in Caco-2 cells  Table 5). The mechanisms underlying the apparent positive effects of LGG and 414 potentially other probiotics in relation to colorectal cancer 107 remain elusive, but the HuMiX model provides an essential tool for systematically exploring these in the future. grown in two distinct oxygen conditions, a GeneGo MetaCore™ pathway enrichment analysis was conducted using only the Caco-2 genes that exhibited contrasting gene expression patterns, i.e., genes that were either up-or downregulated when co-cultured with LGG under either anaerobic or aerobic 420 conditions relative to their respective controls (the threshold parameters used were FC > 1.5 and P < 0.05, BtS, Supplementary Table 6). Interestingly, the top two pathways that exhibited contrasting gene expression patterns (Supplementary Table 6) were linked to the regulation of signaling cascades 423 related to eNOS (endothelial nitric oxide synthase) and substance P (see also Supplemenary Note 5), both of which play major roles in gastrointestinal peristalsis 108, 109 . eNOS is expressed in the gastrointestinal tract and results in the secretion of very low concentrations of nitric oxide, which 426 mediates the relaxation of smooth muscle and thereby plays a major role in mucosal blood flow, permeability, motility and protection 110 . In this context, it is interesting to note that probiotics, particularly Lactobacillus rhamnosus strains, have been found to alter gut motility in vivo 43,111,112 , 429 which in turn may be linked to eNOS and substance P pathway activation. Of particular interest in this context are the opposing expression patterns of G-protein signaling pathways involved in proinsulin C-peptide signaling according to LGG culture conditions (Supplementary Table 6). C-peptide is 432 typically associated with insulin release but it has also been identified as a diagnostic marker for colorectal cancer in mucosal biopsy samples 113 . The inferred contrasting expression patterns of Cpeptide signaling pathways highlight the ability of HuMiX to sustain co-cultures that trigger known 435 responses in epithelial cells and may allow for validation of diagnostic markers identified in vivo.
Furthermore, the enriched pathways that exhibited contrasting gene expression patterns were linked to immune response, cell cycle, cell adhesion, apoptosis, cytoskeleton remodeling, lipid metabolism 438 regulation, signal transduction and developmental signaling pathways (Supplementary Table 6). An additional data-driven pathway analysis using the Gene Ontology database revealed that the top enriched pathways which exhibited contrasting gene expression patterns under anaerobic or aerobic 441 conditions were related to metabolism (more specifically lipid, protein and carbohydrate metabolism), cellular homeostasis, amino acid transporters, and particularly adaptive immune responses (Supplementary Table 7). Intriguingly, among the top 100 pathways, we also found pathways linked to nitric oxide biosynthetic process, which is analogous to the previously described eNOS pathway enrichment which may be linked to the regulation of gut peristalsis.
Overall, these additional results reinforce the importance of co-culturing human cells with 447 bacteria growing under oxygen-conditions to elicit physiologically relevant read-outs.

Supplementary Note 7: HuMiX-based co-cultures of Caco-2 cells with
LGG allow the study of 450 uncharacterized sRNAs. In addition to the miRNA microarrays, we also evaluated the differential expression of sRNAs included in the mRNA microarrays following co-culture of Caco-2 cells with LGG grown under anaerobic conditions (Figure 3a).