Fabrication of a human gut on-chip model and CFD simulation analysis
To investigate the interactions between microbiomes and human Caco-2 cells in a human intestinal microenvironment, we developed a microfluidic-based gut-on-a-chip embedding with microelectrode arrays (Fig. 1A). The gut-on-a-chip consisted of three parallel microchannels (500 μm wide, 10 mm long, 150 μm high) separated by the hexagonal-shaped micropillars: two stromal cell culture channels which could mimic the epithelium or endothelium layer and one central channel filled with collagen type I gel (Fig. 1A, right). We used the three parallel microchannels separated by collagen gel channel without any polymeric membrane, showing that it was relatively easy to fabricate. By using an osmotic pump, the culture medium was perfused at simulated flow rates through each microchannel, mimicking the fluidic flow and associated shear stress on the cell surface in the human intestinal lumen and the blood vessels in vivo. We co-cultured Caco-2 cells and HUVECs separated by collagen gels in a gut-on-a-chip (Fig. 1B). Prior to the experiment, the flow dynamics of the gut-on-a chip were performed in CFD simulation to assess the correlation between wall shear stress and the dimensions of the microfluidic channel (Fig. 2). Within the microchannel, the flow velocity \(u\) is dominated by the homogeneous, incompressible Naiver-Stokes equation and continuity equation:
$$\rho \left(u\bullet \nabla \right)u=\nabla \bullet \left[-p{\rm I}+\tau \right]+F$$
$$\rho \nabla \bullet u=0$$
where \(\rho\) is the fluid density, \(u\) is the velocity field, \(\nabla\) is the divergence, \(p\) is the pressure, \(\tau\) is the Stoke’s stress (\(\tau =\mu (\nabla u+\nabla {u}^{T})\)), \(\mu\) is the dynamic viscosity, and \(F\) is the volume force. In consequence, the wall shear stress acting on the epithelial cell channel was derived from the previously obtained velocity field. Assuming an isotropic Newtonian flow in the channel, the shear stress inside the shear stress inside the microfluidics can be calculated by the following equation:
$$\tau \left(\overrightarrow{u}\right)=\mu \nabla \overrightarrow{u}$$
where μ is the dynamic viscosity, and \(\nabla \overrightarrow{u}\) is the gradient of velocity field which is also called as wall shear rate. The average value of the wall shear stress was acquired from the boundary of the epithelial cell channel in which the lactic acid bacteria were cultured. The height and width of epithelial cell channel and the distance between the micropillars are set as parameters. The dimensions of the parameters were set in the ranges of 50 to 250 μm, 600 to 1000 μm, and 50 to 250 μm, respectively. It is important to find the optimal microchannel dimension for the well growth of lactic acid bacteria. Since the wall shear stress acting at the boundary surface is a crucial factor for the cell growth [36], the correlation between channel dimension and wall shear stress needs to be considered to determine optimal channel dimension. Simulating with respect to the microchannel height, the wall shear stress rapidly decreased as the channel height was increased (Fig. 2A). The average wall shear stress values for the channel height of 0.05, 0.2, 0.35, 0.5, and 0.65 μm at a flow rate of 0.35 μL/min were 16.16, 4.28, 1.99, 1.17, and 0.76 10–3 Pa, respectively. The lower the height of microchannels, the higher the wall shear stress. From the results of CFD simulation for the microchannel width, the wall shear stress was proportional to the microchannel width (Fig. 2B). Within a fixed flow rate, the changes in microchannel height and width leaded to variation in channel volume, which could affect the flow velocity. Simulating with respect to micropillar distance, the wall shear stress was increased with the micropillar distance (Fig. 2C). The flow rate was simulated in the range between 0.05 and 0.65 μL/min for each parameter. All three simulations showed the same tendency for wall shear stress to increase in proportion to the flow rate, because the laminar flow shared the same streamline in the identical geometry [37]. The distribution of the flow velocity in the channel at the different concentrations of PEG solutions were calculated (Fig. 2D). In the case of flow driven by 0.18 M PEG solution, the velocity was relatively uniform over most of the channel, whereas the flow with 0.36 M and 0.72 M PEG solutions exhibited wide variations between the upper and lower flows. The velocity at a position of A, B, C, and D is plotted in Fig. 2E, which shows that the mean velocities driven by 0.18 M, 0.36 M, and 0.72 M PEG were 99.55 mm/h, 197.16 mm/h, and 394.21 mm/h, respectively. At 0.36 M PEG concentration, the flow speed was increased and the colonies of microbiome were detached from the differentiated Caco-2 cells. In contrast, the flow speed was decreased at a 0.09 M PEG concentration, showing that the colonies of microbiome were rapidly grown. On the basis of these results, we optimized a flow speed of 0.18 M PEG solution, which could produce the most stable fluidic culture model in subsequent experiments.
Effect of fluidic flow and endothelial cell on differentiation of epithelial cells in gut on-chip
We first explored whether the shear stress generated by luminal fluidic flow above the epithelium is responsible for induction of the epithelial morphogenesis. To do this, Caco-2 cells were grown either in a static culture or microfluidic culture condition (Fig. 3). After 5 days of cultivation, the Caco-2 cells remained viable in the both fluidic and static conditioned gut-on-a-chip, which was higher cellular density than in the static conditions (Fig. 3A, D). We analyzed the fluidic flow effect on epithelial barrier integrity in Caco-2 cells confirmed by the immunostaining of junctional ZO-1 and the labeling of the actin cytoskeleton, F-actin. When cultured in a fluidic condition, the actin cytoskeleton showed a continuous ring appearance between adjacent cells, whereas the actin staining appeared the discontinuous and less ordered in a static condition (Fig. 3B, E). In addition, the immunostaining data showed the bright signals of the tight junction protein (ZO-1) at the edge of cells cultured in the fluidic culture condition, suggesting that the Caco-2 cells could form the confluent polygonal epithelial monolayers with well-developed tight junctions in a fluidic culture condition, much tighter than cells in a static culture condition. (Fig. 3C, F). Nevertheless, the cultured human intestinal epithelial cell alone spontaneously was not formed highly polarized epithelium and mucus secretion cells. To more effectively ameliorates the mimic in vivo intestinal system and differentiation of the Caco-2 cells, we co-cultured Caco-2 and HUVECs in a gut-on-a chip which separated with 2% collagen gels (Fig. 4). In the same flow conditions, one of the most noticeable changes in a co-culture model, the expression of polarized and differentiated columnar epithelium (Fig. 4A down panel, F-actin) appeared similar form to living in vivo intestinal villi, as previously described [38]. To investigate the effect of fluidic flow on the differentiation, the 3D projections in the Z-stack of the confocal images were used for quantifying the epithelial layer height (Fig. 4B). The epithelium layer differentiated by Caco-2 cells cultured in a gut-on-a-chip showed the 37.04 ± 2.38 μm height of the villi, while those under static conditions were only 14.11 ± 0.74 μm, respectively (Fig. 4C). Another important characteristic to consider when developing in vitro models of the gut is the presence of a mucus layer, which is most abundant structural protein of the gastrointestinal mucus layer [39]. We evaluated Mucin 2 (MUC2) visualized via immunofluorescence (Fig. 4A down panel, MUC2). We confirmed that the Caco-2 cells cultured in a fluidic culture condition generally produced more MUC2 expression as compared to cells grown in a static culture condition. They were located at the tips of the villi-like structures. Mucin are known to protect the underlying epithelium from mechanical stresses [40], which is expected to be highest at the tip of the villi-like structures under fluidic culture conditions. These results are consistent with previous studies combining Caco-2 and other gastrointestinal cell lines with fluidic flow that reported improved mucus production in response to mechanical stimulation [31, 39]. The Caco-2 cells acted as absorptive enterocytes [41, 42], and formed a continuous, planar epithelial monolayer in transwell inserts after 3 weeks of culture, as previously described [14]. However, they could not exhibit similar in vivo intestinal cell differentiation when grown under static culture conditions. In contrast, in our gut-on-a-chip system, the Caco-2 cell monolayer co-cultured with HUVECs spontaneously initiated villus morphogenesis within 5 days when cultured in the presence of fluidic flow and other supporting cells mimicking the physical microenvironment experienced by in vivo intestinal system. Glycocalyx, an efficient defense system for protecting the epithelium from pathogens [43], was evaluated by expression of WGA-Alexa488, which could bind to sugar residues on cellular surface. A marked increasing in WGA-Alexa488 binding was observed in a flow culture condition as compared to a static condition, suggesting that the flow culture condition could improve the epithelial cell differentiation. As expected, the expression of glycocalyx was influenced by a fluidic condition, indicating that flow condition promoted the differentiation of Caco-2 cells. Thus, our results suggest that the mechanical factors of the fluidic flow and cellular components are the crucial microenvironment cues that can drive more complete intestinal differentiation process than local stromal factors [44]. Furthermore, these results indicated that Caco-2 cells cultured in the presence of continuous flow required a shorter time to polarize and differentiate.
Effect of fluid flow and endothelial cell on barrier integrity of epithelial cells in gut on-a-chip
Caco-2 cell only cultured model and co-cultured model were exposed to fluidic flow for 4 days in a gut-on-a-chip and the differentiation process was evaluated using impedance spectrometry (Fig. 5). The impedance spectra of the Caco-2 cell only cultured model and co-cultured model in a gut-on-a-chip showed that the maximum difference in the impedance spectra was observed at 10 kHz (Fig. 5A, B). The measurement of day 0 before cell seeding was subtracted from all subsequent measurements (resulting in the |Zrelative|) to confirm the change in impedance attributed to the cell layer. The relative impedance |Zrelative| at 10 kHz was monitored in a Caco-2 cell only cultured model and co-cultured model on a gut-on-a-chip over 4 days (Fig. 5C). This non-invasive method can be applied to living cells and allows them to be monitored during growth and differentiation, since their morphological changes can be described by variations in impedance measurements [45]. In particular, it is important for monitoring the growth of cellular extrusions like microvilli [46]. In a gut-on-a-chip with Caco-2 cell only cultured model and co-cultured model, the measured impedance kept increasing during all 4 days, showing that the Caco-2 cell only cultured model and co-cultured model was 13 kΩ and 16 kΩ, respectively. The shear stress induced by the fluidic flow has the effect of the mechanotransduction on several endothelial molecular pathways through activation of membrane-bound receptors, leading to the production of the tight junction proteins (e.g., ZO-1). It modulated the cytoskeletal structure to promote the cell reorientation and restructuring [47, 48]. Hereby, the resistance of the sensor surface covered by the cells tends to be increased due to cell proliferation and spreading [49]. Since the current has to flow through the cells, the resistances between the measuring electrodes can keep increasing and the TEER barrier resistance can keep increasing [49]. The relative impedance |Z| and TEER were characterized by an increase in measured cell layer resistance for co-cultured model on a gut-on-a-chip (Fig. 5D). The calculated TEER value was reached up to 59 Ω∙cm2 over 4 days of culture. The previous study has reported a time-dependent barrier formation of human intestinal cells developing from 20 to 60 Ω∙cm2 in a static culture [50]. This value was comparable to our TEER-measurements. However, our co-cultured model on a gut-on-a-chip showed increase in the barrier strength within 5 days of fluidic flow culture. These results demonstrated that the application of fluidic shear stress and endothelial cells enhanced the cellular differentiation and barrier formation of the intestinal cells.
Host cell and probiotic co-culture on gut-on-a-chip
The adhesion of commensal bacteria to host cells is considered as an appropriate parameter to determine the colonization potential of a probiotic strain [51]. However, as the bacterial overgrowth occurs rapidly compromising the epithelium, it is impossible to expose these cells to living microbiome in long-term culture [30]. Thus, the establishment of the stable symbiosis between the epithelium and resident gut microbiome as observed in the normal intestine is crucial to maintain the normal epithelial differentiation and restrain microbial overgrowth in the intestine in vivo [52]. In a present study, we leveraged our gut-on-a-chip to maintain the probiotics. In this study, a commercial probiotic, Lactiplantibacillus plantarum was used as a control, Lactiplantibacillus plantarum HY7715 probiotic and Bifidobacterium animalis spp. lactis HY8002 probiotic were employed. Blocking the fluidic flow for the first 2 h, bacterial cells were allowed to adhere on the apical surface of villi. After 2 h, the physiological relevant flow was resumed through the microchannels to remove un-colonized gut bacteria and supply nutrients to both bacterial and villus epithelial cells. When a non-pathogenic laboratory strain of green fluorescent stained probiotics was allowed to adhere to the apical (luminal) surface of villi for 2 h under static conditions, these bacteria cells were subsequently colonized and spontaneously inhabited regions (Fig. 6). When the bacteria were cultured on the villus epithelium layer under a flow condition (21 μL/hour), we observed the colonized stable form until day 1 in all probiotic groups. However, Lpb. plantarum and HY8002 probiotic seemed detached from the villi and washed out after 3 days when cultured under flow conditions as compared to HY7715 probiotic, although the luminal flow was maintained constant. All species showed adhesion to the used epithelial layer, however, the adhesion level of HY7715 probiotic was greater to epithelial layer even in day 5. These results were concurred with the finding by Schillinger et al., who showed that the adherence of diverse probiotic strains varied among strains [53]. In addition, Gopal et al., has reported the higher affinity of L. acidophilus and L. rhamnosus strains to HT29-MTX cells than HT29 and Caco-2 cells [54]. Nevertheless, we need to optimize the incubation time and volumetric flow rate for attachment to the surface of villi.
LPS-induced intestinal damage responses and evaluation of barrier protection effect of probiotics
We further explored whether our gut-on-a-chip system could be used to mimic the human intestinal inflammation in vitro. Following the establishment of the physiologically relevant gut-on-a-chip, we evaluated whether our model could recapitulate the main characteristics of intestinal inflammation. We focused on LPS as alternative stimulus and chose to directly expose the 5 days cultured gut-on-a-chip model to induce an inflammation-like response for 24 h. LPS, is a heat-stable toxin associated with the outer membranes of gram-negative bacteria, belongs to the most studied pathogen-associated molecular patterns and generally presents in the intestinal lumen known for its involvement in intestinal inflammation [55]. In a co-cultured model on a gut-on-a-chip exposed to LPS, we observed more disorganized structure and fainter staining of F-actin structures as compared to non-LPS treated models (Fig. 7A, F-actin, white arrows). The location and distribution of the ZO-1 protein was determined by immunofluorescence assay. Under normal conditions, ZO-1 proteins were localized at the cell membrane and appeared as a continuous band encircling the cells at the cellular borders (Fig. 4.) LPS (15 μg/mL) disturbed the distribution of ZO-1 proteins at the cellular borders. LPS also induced obvious cytoplasmic accumulation of ZO-1 in Caco-2 cells (Fig. 7A, ZO-1 staining). We measured villus heights after LPS administration, because the villus contraction was typically utilized as a measure of small intestinal damage. In co-cultured intestinal models on a gut-on-a-chip at 24 h after LPS administration, the mean villus height was reduced by 32.9%, indicating 24.84 ± 0.73 μm as compared to villi from non-treated LPS co-cultured intestinal models (37.04 ± 2.38 μm, ***p < 0.001) (Fig. 7B). Furthermore, the paracellular permeability was measured by TEER analysis. Administration of 15 μg/mL LPS resulted in a significant decrease in TEER (non-treated LPS and LPS was 59.83 ± 1.68 and 28.44 ± 1.96 Ω∙cm2, respectively, *p < 0.05, Fig. 7C). We further evaluated the protective effect of probiotics HY7715 probiotic on LPS-induced epithelial barrier dysfunction. As seen in Fig. 7C, when differentiated Caco-2 cell barriers were treated with 1 × 108 CFU/mL, the TEER analysis was significantly increased as compared to HY7715 probiotic-treated group after LPS treatment at day 8 (LPS and HY7715 probiotic was 28.44 ± 1.96 and 44.39 ± 1.25 Ω∙cm2, respectively, *p < 0.05). These results suggested that the epithelium layer differentiated from Caco-2 cells was damaged by LPS treatment in the co-cultured model on a gut-on-a-chip. Its similar trends were observed in other in vitro gut inflammation models [56] and inflammatory bowel disease (IBD) patients [57]. Additionally, the strain HY7715 probiotic suppressed LPS-induced decreases in TEER analysis on a co-cultured intestinal models. It suggests that strain HY7715 probiotic can play an important role in changes in intestinal cell permeability. This result is consistent with the finding that probiotic strains of bacteria, including Lactobacillus rhamnosus GG (LGG), have been reported to elevate intestinal epithelial integrity [58] in vitro and improve intestinal barrier function in a human [59]. The previous studies have reported that L. plantarum significantly reduced the production of inflammatory cytokines and gut permeability in an IBD pathology by regulating the LPS [60]. This result demonstrated that intestinal epithelial integrity significantly increased in the presence of HY7715 probiotic co-cultures. The presence of the probiotics clearly provides useful microenvironmental signals that enhance epithelial cell functions, which are necessary to maintain this dynamic interface.