Bioengineered 3D tissue model of intestine epithelium with oxygen gradients to sustain human gut microbiome

Abstract

The human gut microbiome is crucial to host physiology and health. Therefore, stable in vitro coculture of primary human intestinal cells with a microbiome community is essential for understanding intestinal disease progression and revealing novel therapeutic targets. Here, we present a three-dimensional (3D) scaffold system to regenerate an in vitro human intestinal epithelium that recapitulates many functional characteristics of the in vivo small intestine. The epithelium, derived from human intestinal enteroids, contains mature intestinal epithelial cell types and possesses selectively permeable barrier functions. Importantly, by properly positioning the scaffolds cultured under normal atmospheric conditions, two physiologically relevant oxygen gradients, a proximal-to-distal oxygen gradient along the gastrointestinal (GI) tract and a radial oxygen gradient across the epithelium, were distinguished in the tissues when the lumens were faced up and down in cultures, respectively. Furthermore, the presence of the low oxygen gradients supported the coculture of intestinal epithelial cells along with a complex living commensal gut microbiome (including obligate anaerobes) to simulate temporal microbiome dynamics in the native human gut. This unique silk scaffold platform may enable the exploration of microbiota-related mechanisms of disease pathogenesis and host-pathogen dynamics in infectious diseases including the potential to explore the human microbiome-gut-brain axis and potential novel microbiome-based therapeutics.

Graphical Abstract

Scaffold system is developed to engineer in vitro human intestinal epithelium that recapitulates many functional characteristics of the in vivo small intestine. By properly positioning the scaffolds cultured under normal atmospheric conditions, physiologically relevant oxygen profiles are generated which support host-gut microbiome cocultures and simulate temporal microbiome dynamics.

Introduction

Over 90% of food digestion and nutrient absorption occurs in the small intestine, which is comprised of three distinct segments: duodenum, jejunum, and ileum. These segments are all lined with a layer of columnar epithelial cells that are bound by junctional complexes (i.e., tight junctions (TJs), adherens junctions, and desmosomes) and play roles in mucus production, secretion of antimicrobial peptides and molecules, and harboring trillions of diverse commensal microbiota under a very low luminal oxygen tension (1). The intestine has been known to coevolve with its resident microbial community such that the intestinal mucosa (mainly formed by epithelial cells and mucus) physically serves as a barrier separating the host from its resident microbes and external environmental stimuli, while simultaneously permitting the selective passage and exposure of the favorable microbial contents and metabolites to the host, thus maintaining gut immune homeostasis (2). During the last decade, scientific studies have increasingly implicated the gut microbiome as an essential element involved in the host defense against invading pathogens and the production of metabolites to modulate the local and systematic immunity and many other host physiological processes (3). Microbial imbalance, dysbiosis, of the gut affects many human diseases ranging from gut-localized disorders such as inflammatory bowel disease (IBD) (4) and irritable bowel syndrome (5, 6), to immune-mediated diseases and metabolic conditions like diabetes and obesity (7), cancers (8), and even neurological diseases such as Parkinson’s and Alzheimer's (9-11) and neuropsychiatric disorders such as autism and dementia (12). Links between the gut microbiome and diseases have driven the gut microbiome to be therapeutic targets. For example, fecal microbiota transplantation (FMT) is utilized to eliminate recurrent Clostridium difficile infections (13, 14) and alleviate Ulcerative Colitis (UC) symptoms in patients (15, 16).

To date, most of our current understanding of the mechanisms regulating complex host-microbe interactions has been generated in murine models, particularly germ-free mice (17, 18), because of the inaccessibility of the human digestive tract. However, due to the metabolic, anatomical, physiological, and biochemical differences in the gastrointestinal (GI) tracts of humans and mice, a large portion of human-originated microorganisms fail to colonize or fully function in mouse intestinal mucosa (19). Therefore, the fundamental mechanisms by which the microbiome modulates mammalian cellular functions to influence health or observed diseases remain elusive (4). In addition, microbiome research findings obtained in mice do not always reliably translate to humans (20, 21). These discrepancies between the two systems most often result in disparate outcomes in clinical testing that have hampered the discovery and development of new microbiome-targeted therapies. To tackle this problem, tissue engineering approaches and biomaterial fabrication techniques have been proposed for the design of reliable in vitro intestine models with the aim of providing a simplified version of healthy or pathological human intestinal niches, in particular models that sustain human-microbial co-culture. Such models have been used for investigating the interplay between the human host and its microbiota to uncover the causes of disease onset and progression that often cannot be studied in human beings or animals (22, 23).

As most gut bacteria are obligate or facultative anaerobes that can only survive in anaerobic or very low-oxygen conditions, a major obstacle in developing an in vitro host-microbiome co-culture system is that while the majority of (>90%) obligate anaerobes die quickly after exposure to an oxygen enriched atmosphere (21% O₂) (1), human cells require oxygen to satisfy their metabolic needs (24). Therefore, in the field of GI tissue engineering, considerable research effort has been conducted in maintaining low oxygen levels on the apical intestinal epithelial surface to allow the survival of anaerobes while maintaining oxygen on the basal surface to ensure sufficient tissue oxygenation and survival. For instance, several co-culture systems of intestinal epithelium and anaerobic gut microbiota using transwell inserts have been developed (25, 26). These systems typically consist of a hypoxic apical chamber and a normoxic basal chamber by perfusing normally oxygenated and deoxygenated media in segregated compartments to maintain the co-culture environment. This type of system, while able to support the co-culture of human intestinal epithelial Caco-2 cells with some obligate anaerobes, does not maintain stability over time without continuous gas purge or degassing of the media or the placement of the device into an anaerobic chamber. Another system, the “Human-oxygen-Bacteria anaerobic” (HoxBan), was also established to co-culture an obligate anaerobic gut bacterium and Caco-2 cells (27). This system is based on the use of solid agar for the colonization of anaerobic Faecalibacterium prausnitzii, overlaid with liquid medium exposed to air for Caco-2 cells. While this system has some advantages, the lack of ventilation in the tubing causes a short life span of the co-culture (<36 hours). Over the past decade, microfluidic technology and organ-on-a-chip devices, such as HuMix model (human–microbial cross talk) (28) and gut-on-chip systems (29-31), have been established to create an oxygen gradient for sustaining the growth of epithelial cells and bacteria and investigating different aspects of the host-pathogen interactions in the human intestine. While these systems may generate oxygen gradients, they rely on in-depth technical knowledge and equipment to manufacture and maintain the stable flow systems. Additionally, existing microbe-intestinal systems remain severely limited due to the inability to use primary human intestinal cells and the lack of 3D architecture mimicking the in vivo intestine, both of which have been considered critical for increasing biological relevance for in vitro modeling (23). It is, therefore, critical to develop a simple tissue culture system that can be utilized to model a physiologically relevant human intestine with support of microbial co-culture.

To accommodate this need, our lab previously built a 3D tubular sponge scaffold system that formed a physiologically relevant representation of the luminal microenvironment for the residence of the intestinal epithelium derived from intestinal cell lines (32) or organoids (33). These 3D systems were able to sustain infections of facultative (32, 33) and obligate (34) anaerobes, and parasites (35). In the present study, these tubular scaffold designs were further simplified for higher throughput studies by developing half-scaffold systems (with half luminal size of the tubular scaffolds) for facile scale-up of the tissues for human gut microbiome studies. Using these 3D half scaffolds, we describe how functional human primary intestinal systems were established that regenerate the two steep oxygen gradients (luminal-across the lumen and radial-across the epithelium) existing in human intestine, which subsequently supported stable cocultures of human patient-derived intestinal epithelium with aerobic and anaerobic human gut microbiota.

Results

1. Design and fabrication of half scaffolds

We have previously established a tubular silk protein sponge scaffold system with a 3D compartmentalized structure that mimics physiologically relevant geometry of the native intestine. In the scaffolds, the central hollow lumen enabled the reconstruction of a functional 3D intestinal epithelial interface in vitro (32, 33, 36), while the bulk space enabled support for other intestinal cells including subepithelial cells (human intestinal myofibroblasts) (32, 33, 37), neural stem cells (38), macrophages (37) and smooth muscle cells (39). For the present study, the goal was to create a higher throughput 3D system with a simplified scaling by developing half-scaffold systems (Fig. 1). 3D printing (Fig. 1A-1,2) was adapted to print resin molds (Fig. 1A-3) for the generation of reverse polydimethylsiloxane (PDMS) molds (Fig. 1A-4,5). Subsequently, 4% silk solutions prepared from silkworm cocoons as described in previous studies (32), were poured into the reverse PDMS molds (Fig. 1A-6) and allowed to freeze and dry to obtain the sponge scaffolds (Fig. 1A-7). The initial printing design of the resin molds with smooth lumens resulted in final scaffolds with smooth lumens (luminal size: 2 mm diameters, 8 mm lengths) (Fig. 1-7). Scanning electron microscopy (SEM) imaging of different positions of these half scaffolds showed the formation of homogenous membranes with micropores ranging in diameter from approximately 2-5μm (Fig. 1C) throughout the luminal surface of the scaffolds (Fig. 1B) and a porous spongy silk bulk matrix directly underneath (Fig. 1D-E).

Figure 1.

3D half scaffold systems for intestinal tissue engineering were fabricated using silk fibroin. (A) Schematics of the fabrication process for building silk-based porous half scaffolds for 3D human intestine engineering. Silk-based scaffolds are constructed by 1) design 3D model using 3D CAD software, 2-3) 3D printing positive resin molds, 4) applying liquid PDMS to the positive 3D printed resin molds, 5) curing and obtaining the PDMS molds with heat exposure, (6) applying aqueous silk solutions to PDMS molds, and 7) freezing and lyophilizing. (B-E) SEM images of different views of the 3D half silk scaffolds, including top view of the whole scaffolds (B, scale bar: 1mm), the micropore structure on the luminal surface (C, scale bar: 5μm), the cutaway of the interface between the film and sponge (D, scale bar: 200μm), and the interconnected pores in the spongy bulk of the scaffolds (E, scale bar: 200μm).

2. In vitro engineering of 3D human intestine tissues with enteroid-derived nontransformed epithelial cells

The new 3D silk half scaffold system was used to build intestinal tissue constructs in vitro with human intestinal enteroids and myofibroblasts (Fig. 2A-C). Human intestinal enteroids were used for this study as these cells are derived from human intestinal tissues and have the ability to differentiate into heterogeneous cell populations that recapitulate the intricate pattern and functionality of the original epithelial tissue (40). InMyoFibs were selected to be epithelial supporting cells based on previous work showing that sub-epithelial myofibroblasts promote epithelial growth and barrier functions (32, 41). To initiate the co-cultures, a single cell suspension of human intestinal enteroids was generated by enzymatic digestion and seeded via micro-pipetting directly onto the luminal surface, permitting cell attachment and growth. Human intestinal myofibroblasts (InMyoFibs) were introduced in the porous bulk space to support epithelial cell growth and maintenance. Subsequently, the 3D tissues were maintained in complete Wnt3a, R-spondin 3 and Noggin (WRN) media (growth medium) for one week to trigger adequate cell proliferation and the formation of a confluent epithelial layer within the tissue lumen. The tissue constructs were then switched to differentiation medium to enable the epithelial differentiation for up to 2 weeks. For experimental controls, enteroid-derived epithelial cells cultured in the upper chamber of 2D transwell inserts were used in conjunction with InMyoFibs in the lower chamber (Fig. 2B).

Figure 2.

Enteroid derived epithelial cells formed a confluent and functional monolayer in 3D silk half scaffolds. (A-C) The illustration of cell culture of human intestinal enteroids and primary myofibroblasts (A), the cell seeding strategies for Transwells (B) and 3D half scaffolds (C). (D) Representative confocal z-stack of the DAPI staining on the luminal surfaces of the scaffolds (Scale bar: 200μm). (E-H) Confocal immunofluorescence images of the epithelia for SI (E), Muc-2 (F), Lysozyme (G), ChgA (H), and ZO-1 (I). Scale bars: 20μm. (J-K) SEM images on the apical surface of the epithelium revealed brush borders with tightly packed microvilli (Scale bars: 10μm and 5μm).

To determine whether the 3D half scaffolds provided a suitable environment for enteroid-derived cells to adhere, grow, differentiate, and form functional epithelium, undifferentiated or differentiated samples were collected and processed for immunostaining or SEM. 3D reconstruction of confocal z-stacks of DAPI staining on the luminal surfaces of the scaffolds revealed confluent intestinal epithelial monolayers formed in the lumens after a week of culture (Fig. 2D) suggesting the lumens with curvature supported cell adhesion and proliferation. Three days post-differentiation, immunofluorescence showed that enteroid-derived epithelial cells in the scaffold lumens differentiated into a heterogeneous population of cells expressing differentiation markers characteristic of native intestinal epithelium (42), including sucrase-isomaltase (SI) (a marker of enterocytes, Fig 2E), mucin 2 (Muc2) (Goblet cells, Fig 2F), lysozyme (Paneth cells, Fig 2G), Chromogranin A (ChgA) (enteroendocrine cells, Fig 2H), and Zonula occludens (ZO)-1 (tight junctions at the periphery of the cells, Fig 2I). Furthermore, SEM images of the differentiated monolayers identified tightly packed microvilli in a brush border arrangement on the apical surface of the polarized epithelial cells (Fig 2J, K), a typical ultrastructure feature of mature enterocytes with absorptive functions (43). These findings indicated that the 3D half scaffold lumen provided a suitable culture environment for stem cell-derived enteroids to attach and differentiate.

3. 3D half scaffolds promote epithelial cell differentiation and functions when compared to the 2D Transwell insert controls

Increasing evidence suggests that culture dimensionality influences the behavior of stem cells, including proliferation, differentiation, permeability, and transport functions (44). In this context, we characterized cell differentiation and function of enteroid-derived epithelial cells within the 3D half-scaffolds and 2D tranwells (Fig 2A-C). The timeline for cell seeding, proliferation, differentiation and sample collection are summarized in Fig 3A.

Figure 3.

3D architecture and multicellular co-culture system significantly improved the overall differentiation and function level of the epithelial cells. (A) Schematic diagram showing the experimental timeline for cell seeding, medium switching and sample collecting. (B-I) Gene expression levels of intestinal epithelial biomarkers, including SI (B), Lysozyme (C), Muc-2 (D), ChgA (E), and Lgr-5 (F), and cell-cell junction related genes, including ZO-1 (G), Occludin-1 (H) and Claudin-4 (I), were evaluated by quantitative reverse transcription-polymerase chain reaction (qRT-PCR). (J) Dextran-FITC permeability assay in epithelia in 3D half scaffolds and transwells. (K) Glucose uptake assay in epithelia in half scaffolds and transwells. Data is presented as mean ± SEM, n = 3 in each group, * p ≤ 0.05, ** p ≤ 0.01, *** p ≤ 0.001.

mRNA analysis was utilized to quantify gene expression of a panel of known intestinal differentiation markers over time in the 2D and 3D cultures (Fig 3B-I). In both systems, significant upregulation of transcripts for four major intestinal epithelial markers (SI, Lysozyme, Muc-2, and ChgA) were detected after 3 days of differentiation (Fig. 3B-E), consistent with the immunofluorescence results, while the relative mRNA expression level of Lgr-5, an intestinal stem cell marker, declined over time after differentiation (Fig. 4F). This demonstrated that the differentiation medium efficiently triggered LGR5-positive stem cells to differentiate into the major epithelial cell types, a requisite step for the development of a functional human small intestinal epithelium (45). When compared to Transwells, the half scaffolds sustained enteroid-derived epithelial monolayers longer, demonstrated by the decrease in RNA levels of all marker genes after day 7 in Transwells, versus the delayed decrease between days 9-11 in the half scaffolds (Fig. 3).

Figure 4.

3D bioengineered intestinal tissues generated low luminal oxygen levels. (A-C) Scaffolds cultured in a “up” position (A) generated microaerobic conditions in the lumen (pO2 between 5% and 7%, B) stable for up to 48 hours (C). (D-F) Scaffolds cultured in a “flipped” position (D) generated anaerobic conditions (pO2 ~1-2%, E) stable for up to 48 hours (F). (G-H) Confocal immunofluorescence staining for HIF1α) on scaffolds “up” (G) and “flipped” (H) for up to 48 hours. Scale bars: 50μm. White arrows: silk autofluorescence; yellow arrows: HIFa positive cells. (I) Quantification of HIF1a positive cells of scaffolds “up” vs “flipped”. Data is presented as mean ± SEM, n = 3 in each group, *** p ≤ 0.001.

Tight junctions (TJs) play an essential role in maintaining intestinal barrier functions and controlling paracellular transport pathways in the epithelium (46). TJs are assembled by TJ-associated proteins including members of the occludin, claudin, and zonula occluden (ZO proteins) families (47). To assess whether 3D scaffolds with curvature increased barrier functions compared to 2D Transwells, we next evaluated the gene expression level of tight junction associated proteins (ZO-1, Occludin-1 and Claudin-4) (Fig 3G-I). After differentiation, Occludin-1 gene expression patterns of the epithelium on 3D tissue constructs did not show significant differences compared that of the epithelium on Transwells, suggesting that there are no significant effects of 3D culture on the expression of the Occludin-1 gene (Fig 3H). However, significantly higher gene expression of ZO-1 (from day 5 to day 9, p<0.01, n=3) and Claudin-4 (from day 7 to day 9, p<0.05, n=3) in the 3D scaffolds were observed over time (Fig 3G and I), implying that half scaffold curvature may improve tight junction formation of an enteroid-derived epithelium by enhancing transcripts of ZO-1 and Claudin 4.

The functional intestinal barrier is an active semipermeable system that restricts the entry of harmful substances like pathogens and toxins, but simultaneously permits the selective uptake of beneficial nutrients like glucose and electrolytes from the luminal contents (48). Therefore, it is important to characterize intestinal permeability and transport function of enteroid-derived epithelia cultured in different platforms. We investigated the robustness of the 3D and 2D intestinal models regarding paracellular transport with fluorescein isothiocyanate (FITC)–Dextran 4 kDa permeability assay (Fig. 3I) (49). In the 3D sponge scaffolds, Dextran diffusion across the luminal epithelium could be entrapped in the silk sponge bulk space, thus affecting the accuracy of Dextran fluorescence signals that transited across the epithelial layer when reading how much Dextran was in the basal side of the epithelium. Therefore, we quantified the amount of Dextran retained on the apical side of the epithelium by measuring Dextran retained in the upper chambers of the Transwells of the scaffold lumens after incubation. In this setting, a higher concentration of Dextran in the upper chamber suggests lower Dextran permeability across the epithelium barrier, reflecting better epithelial integrity. Media from the chambers of the upper chambers of the Transwells of the scaffold lumens were collected for the fluorescence intensity readings to determine the epithelial permeability. While both tissue models showed a similar tendency of decreased permeability (increased FITC fluorescence signal in the apical side of the cells, Fig 3I) over time during differentiation, there was no significant difference in epithelial permeability between the two systems. This data conflicted with the TJ protein gene expression results, which suggested that the paracellular permeability of the epithelium in 3D scaffolds would be lower than that of the epithelium in 2D Transwells. The discrepancy may be a consequence of the lack of epithelial cell coverage on the inner surface of the two ends of the half scaffolds (Fig 2C) offsetting increased epithelial barrier function in the 3D environment. Taking this factor into consideration, paracellular epithelial permeability in 3D scaffolds is likely lower than it is in the Transwells based on the protein expression results. To further investigate the impact of 3D architecture on epithelial transport, we compared glucose transport in both the 2D and 3D intestinal tissue models using glucose uptake assay (Fig 3J). Three days post-differentiation, the glucose uptake levels of the epithelia in both 2D and 3D significantly increased. However, the epithelia on the 3D scaffolds demonstrated a significantly higher uptake level, with the most significant difference on days 5 and 7 (p<0.01, n=3). Glucose absorption across the small intestinal epithelium plays a pivotal role in human nutrition and initiates signal transduction and metabolic pathways to maintain metabolic homeostasis (50). It was noted that 3D human tissue cultures, such as liver (51) and adipose tissues (52), promoted glucose uptake and closely resembled human in vivo metabolic activity. One possible mechanism that explains this phenomenon is that 3D cell culture mimics the tissue environment more closely than 2D, propelling cell differentiation, leading to more pronounced cellular responses to glucose (53). Our findings further confirmed previous findings by showing that the 3D culture environment supported more efficient glucose uptake of intestinal epithelial monolayers than 2D transwells.

4. 3D half scaffolds autonomously generate low oxygen profiles mimicking oxygen gradients in the human intestine

The human intestine establishes two steep oxygen gradients: a proximal-to-distal oxygen gradient along the gastrointestinal tract from the stomach to the distal colon and a radial oxygen gradient from the intestinal smooth muscle wall to the mucus layer (54). In the present study, 3D luminal geometries for the half-scaffolds were exploited to achieve low oxygen gradients by positioning the scaffolds with the lumens facing up (scaffolds “up”, Fig 4A) or down (scaffolds “flipped”, Fig 4B) in culture. Briefly, after cell seeding, the constructs were cultured in growth medium for a week to obtain confluent intestinal epithelial monolayers and then switched to differentiation medium for 5 days to induce intestinal epithelial differentiation. An oxygen probe was used to measure oxygen levels in the lumens under different culture positions. We discovered that when cultured with the lumen facing up, microaerobic conditions (pO₂ between 5% and 7%) were achieved and stabilized in the scaffold lumens approximately half an hour after culture initiation (Fig 4B); while when cultured with lumens facing down (flipped), anaerobic conditions (pO₂ ~1-2%) were achieved (Fig 4E). These autonomous low oxygen profiles were sustained for at least 48 hours in all cultures (Fig. 1 C, F). Of note, further oxygen measurements taken in “flipped” scaffolds uncovered a depth-graded oxygen profile from culture medium (~14.95%) to scaffold bulk (~9.87%) and to the lumen (~1-2%) (Supplementary Fig 1). To confirm this hypoxia in the 3D half scaffold lumens, immunofluorescence staining for hypoxia inducible factor-1α (HIF1α), an intrinsic marker for tissue hypoxia (55) was performed on tissues under different culture positions (Fig 4G, H). Samples were fixed and stained at 2 hours, 24 hours, and 48 hours post positioning “up” or “flipped”. HIF1α protein was barely detectable in the scaffolds “up” over time (Fig 4G). In contrast, in the “flipped” scaffolds, expression of HIF1α protein was induced after 2 hours in culture and dramatically increased by 24 hours (Fig 4H and I), validating the hypoxic environment established in the lumens of the “flipped” scaffolds. Since the hypoxia gene responds optimally when exposed to oxygen concentrations below 5% (56), the absence of HIF1a fluorescence in scaffolds (up) and its presence in scaffolds (flipped) might further distinguished between the microaerobic conditions of the scaffolds (up) and the anaerobic conditions of the scaffolds (flipped) (Fig 4H, I). We also monitored myofibroblast cell survival with the different oxygen levels as a result of scaffold positioning by using AlamarBlue as a cell viability assay along with live staining (Supplementary Fig 4). The cell viability of myofibroblasts was not significantly affected by the different oxygen levels in the scaffold bulk.

5. Hypoxic atmosphere generated in 3D tissues supports co-culture of the intestinal epithelium with viable human microbiome

Key features of a healthy human intestine not only include a low oxygen profile throughout the lumen but also include a resident commensal microbiome (1). Therefore, we explored whether the tissue culture systems could be used to coculture gut microbiota isolated from fresh human stools. To establish stable co-cultures of obligate anaerobic human gut microbiota with the human intestinal epithelial cells in our 3D half scaffold systems, gut microbiota (Openbiome, Cambridge, MA) was introduced into the 3D tissue lumens at day 5 post differentiation of the intestinal epithelium (positioned “up” and “flipped”) at MOIs (multiplicity of infection) of 1 and 5 (ratio of bacteria/cells) to optimize the co-cultures (Fig 5A). The reason for selecting day 5 for the introduction of the microbiome is that the epithelial cells have established barrier integrity at this time point as indicated by gene expression of tight junction proteins exhibited peak expression at day 5 post differentiation (Fig 3F-I) and FITC-Dextran showed the least permeability (Fig 3J). We established multiple readouts for the infection process to confirm a stable coculture of host-microbiome. These readouts included: (i) SEM analysis of bacterial attachment and growth (Fig 5), (ii) SEM analysis of the ultrastructure of apical microvilli of the host cells (Fig 5), (iii) immunofluorescence and quantification for epithelial tight junction integrity (Fig 6), (iv) 16S rRNA gene sequencing for microbiome diversity (Fig 7).

Figure 5.

Interactions between 3D bioengineered intestinal tissues and human gut microbiome. (A) The illustration of the coculture process. (B-G) SEM images on the apical surface of the non-infected control intestinal epithelia in scaffolds “up” (B-D) and “flipped” (E-G) revealed brush borders with tightly packed microvilli (Scale bars: 10 μm). (H-N) SEM images on the apical surface of the infected intestinal epithelia with MOI 1 in scaffolds “up” (H-J) and “flipped” (K-M) revealed colonized and expanded the number of a live population of microbiome without disrupting the microvilli (Scale bars: 5 μm). (N-S) SEM images on the apical surface of the infected intestinal epithelia with MOI 5 in scaffolds “up” (N-P) and “flipped” (O-S) revealed colonized and expanded the number of a live population of microbiome with different extent of damages of microvilli and monolayer integrity (Scale bars: 2 μm).

Figure 6.

The effects of microbiome on epithelial tight junctions. (A-F) Confocal images of tight junction ZO-1 protein on non-infected control intestinal epithelia in scaffolds “up” (A-C) and “flipped” (D-F). (G-L) Confocal images of tight junction ZO-1 protein on the infected intestinal epithelia with MOI 1 in scaffolds “up” (G-I) and “flipped” (J-L). (M-R) Confocal images of tight junction ZO-1 protein on the infected intestinal epithelia with MOI 1 in scaffolds “up” (M-O) and “flipped” (P-R). Scale bars: 10μm. (S-T) The intact tight junction index was quantified in scaffolds “up” and “flipped” cocultured with microbiome at MOI 1 and MOI 5.

Figure 7.

3D bioengineered intestinal tissue supported the coculture of human microbiome. (A) Simpson diversity index showed the different diversity levels of microbiome identified from different samples and coculture conditions. (B) The relative abundance of top 20 bacterial genera different samples and coculture conditions. (C) Heatmap of log10 relative abundances of the species across all samples displayed bacterial level changes of the top 20 most abundant genera over time. Except for the intact human stool sample (intact, n=1), data from all groups represent each of the 3 replicates.

We first performed SEM on the epithelia to visualize host-microbiome interactions during the coculture in scaffolds in both “up” and “flipped” positions. The SEM analysis revealed that the microbiome adhered to and proliferated at the surface of the host epithelial tissues 2 hours after infection in both systems (“up” and “flipped”) at both MOI 1 and 5 (Fig. 5H-S). After 24 and 48 hours of co-culture at MOI 1, the microbiota had colonized and expanded in both scaffolds “up” and “flipped” without disrupting the host cells. This was evidenced by that fact the apical brush border carrying microvilli, an essential component of the epithelial barrier, in the coculture group with MOI 1 remained highly organized and densely packed (Fig 5H-M), comparable to that of the control group (Fig 5 B-G). In contrast, in coculture groups with MOI 5, microbiota overgrew after 24 hours, destroying the integrity of the microvilli lining and leading to an increase in dead cells in both scaffolds “up” (Fig 5O, P) and “flipped” (Fig 5R, S) cultures.

Tight junction expression is a major indicator of the integrity and permeability of the intestinal epithelium, and reflects general cell viability. To further assess the responses of the epithelial cells to microbiome, we examined the expression of the tight junction protein ZO-1 by confocal immunofluorescence in the epithelia in scaffolds “up” and “flipped” treated with different MOIs (1 and 5) at different time points (2, 24, and 48 hours). The ratio of the sum of smooth/straight lines within the junctions identified by ZO-1 staining to cell number was defined as “the intact tight junction (TJ) index” and employed to quantify the degree of integrity of intestine epithelium as previously reported (57). Similar to the control group with no microbiota (Fig 6A-F), the 48-hour microbiota cocultures with MOI 1 showed a continuous pattern of ZO-1 distribution along the cell periphery, marking cell-cell junctions with a classic chicken wire pattern (Fig 6G-L). However, when treated with microbiota at MOI 5 (Fig. 6M-R), the epithelial ZO-1 tight junctions in both “up” and “flipped” culture systems appeared slightly impaired after 24 hours of infection (Fig 6N and Q) and severely disrupted after 48 hours (Fig 6O and R).

Based on these observations, coculture systems with MOI 1 were chosen for the following study of microbiome diversity. Bacterial diversity and relative abundance in the original fecal samples and over 48 hours in coculture (n=3 for each condition) was assessed using 16S rRNA sequencing of the microbiome from “up” and “flipped” tissue systems, and 92 species and 69 genera were identified (Fig. 7). The number of phyla and genera in the coculture systems were lower than what was observed in human stool samples, as expected (Fig 7A), yet a diverse range of bacteria was observed, along with an expansion in the number of some species in the scaffolds over time compared to the starting inoculum. For example, Escherichia coli, Shigella and Streptococcus were present at higher relative abundance in both “up” and “flipped” culture systems over time compared to the original stool samples, suggesting some gut microbial species may be favored under conditions that more closely mimic regions of the living intestine tissue models compared to the isolated stool sample. Furthermore, the relative abundance of several obligate anaerobes, such as Megasphaera, Bacteroides thetaiotaomicron, Lachnoclostridium, and Anaerostipes, were higher in the scaffolds (flipped) than in the scaffolds (up), confirming the hypoxia in the mucosa required to maintain obligate anaerobes from the gut microbiota. In the study, infected Transwells were set as control. However, at both MOI 1 and 5 cocultures, bacteria underwent progressive proliferation and killed the epithelial monolayers on the inserts. The cocultures in Transwells did not pass 24 hours of infection (Supplementary Fig 2 and 3).

Discussion

In this study, a scalable and reproducible technique for fabricating 3D silk scaffolds with defined marco- and micro-architectures for intestinal tissue engineering was presented. The fabrication process was relatively simple and involved the use of a high-resolution 3D printer and a biocompatible 3D printing resin (Fig 1A-2) that enabled the printing of resin based “half scaffold” master molds, which were cuboids with a smooth semi-cylinder lumen (Fig 1A-3). The master resin molds were then adopted for casting the reverse PDMS molds as substrates for the fabrication of the 3D half silk scaffolds (Fig 1A4-7). Conventional ways to make the master molds for PDMS casting include standard microfabrication techniques (i.e., silicon etching or SU8 lithography) (58), micromilling on aluminum (59) and polymethylmethacrylate (PMMA) sheets (60). These methods are costly, time consuming and less user-friendly. The 3D printing technique allowed us to rapidly produce multiple (up to 49 each batch) highly customizable resin molds that could be used for casting PDMS negatives. The 3D printed master molds were also re-usable. In addition, the molding and scaffolding procedures employed here accurately preserved macro and microstructures in each step without special processes by the direct peeling of the cured PDMS molds or the crosslinked silk scaffolds from their casting molds. The 3D sponge scaffolds were prepared with a simple freeze-drying method using silk, the main protein of silkworm cocoons (32). Silk proteins have excellent properties for biomedical applications such as biocompatibility, biodegradability, non-toxicity, adsorption properties, and the ability to be proteolytically degraded and resorbed in vivo, and robust mechanics (61). All the fabrication steps can be conducted in any laboratory space without expensive facilities. Considering these advantages, this replica molding strategy combining 3D printing of a master mold with PDMS reverse molds and silk scaffolding can potentially be employed to create functional scaffolds capable of imitating the complex geometry of any type of tissue and organ.

With respect to intestinal tissue models, polarized Caco-2 epithelial cells grown on flattened Transwell inserts have become a standard approach in the field to study intestinal epithelial interactions with pathogens as well as other stimuli causing inflammation and toxicity (62). However Caco2 cells, originally derived from a colon carcinoma, and the planar cell culture inserts, have limited physiological structure and function (63). With advances in stem cell technology and regenerative medicine, intestinal organoids, including enteroids from small intestines, have become promising candidates for epithelial tissue modeling, as the intestinal stem cell-derived organoids have the potential to give rise to all types of mature intestinal epithelial cells, making them well-suited for studying host-microbe interactions (40). In recent years, the 3D scaffold environment, including surface patterns and matrix dimensions and configurations, has significantly promoted stem cell growth and differentiation and the formation of 3D structures. To improve the formation and function of intestinal epithelium, we previously constructed a 3D silk scaffold system with a hollow channel that provided more physiologically relevant characteristics of the intestinal epithelium microenvironment for the residence of the epithelial cells derived from intestinal cell lines (32, 36) or organoids (33). In the present study, these designs were further simplified for higher throughput studies, by developing half-scaffold systems (Fig. 1 A, B), for facile scale-up of human gut microbiome studies. Half scaffolds possess multiple advantages to support scale up: i) the general size and dimensions of the half scaffold represents half the size of the tubular scaffolds, which significantly reduces cell numbers needed for initial seeding, scaffold seeding duration, and culture medium volume used per culture compared to the tubular scaffolds, ii) as the half scaffolds have open lumens, the defects on the lumen surfaces could be identified by visual inspection allowing for improved quality control, and iii) direct imaging of the surface was now available due to the open lumen for access.

In the study, we determined whether the 3D scaffolds with a semi-luminal surface were sufficient to support the differentiation of stem cell-derived enteroids into the various epithelial cell subtypes to establish a functional intestinal epithelium. In the half-scaffolds, human intestinal enteroids were seeded on the semi-luminal surface to generate epithelial monolayers with the support of human intestinal myofibroblasts in the bulk (stromal layer) sponge. The resultant epithelial tissues in the 3D half scaffolds consisted of multiple differentiated cell types (enterocytes, Goblet cells, enteroendocrine cells, Paneth cells) and undifferentiated stem cells (LGR-5 positive cells), as found in human native intestine. Importantly, when compared to the same cells seeded in Transwells, the 3D scaffolds enhanced epithelial cell differentiation and improved the selectively permeable barrier function of the epithelium (Fig 3). These results suggested that, compared to the 2D Transwell inserts, the 3D half scaffolds enhanced epithelial cell differentiation and functions. The luminal curvature may play a role in influencing cell behavior as previous studies have demonstrated curved surfaces contributed to apical-basal polarization that allows ZO-1 expression and therefore increased intestinal epithelial barrier functions (64, 65). However, intestinal epithelial cell behavior on a planar version of the scaffold or on silk scaffolds with different degrees of luminal curvature should be compared to confirm the impact of the curvature on intestinal epithelial cells in this system. While many studies have attempted to study the effects of curvature on cell development, it remains challenging to engineer precisely curved and smooth surfaces with curvature radii at a micrometer range, such that they can be sensed by cells (66). The radius of curvature is important, as too large a radius can prevent the effects of curvature from having the effect, whereas too small a radius can result in large radial forces which can lead to monolayer detachment (44). Since the master molds used in our scaffolding process were 3D printed, different degrees of luminal curvature and the crypt-villus structures could be designed and manufactured. In this respect, our scaffolds with potentially tunable surface curvature, along with different mechanics and structure (crystallinity) can be further exploited to address question of surface curvature radius (including mechanics, crystallinity) correlates with epithelial morphogenesis, how the cells in tissues sense and respond to local curvature, and how these features ultimately affect cell behavior.

In the human intestine lumen, the apical side of the epithelium is exposed to a relatively low-pO₂ environment with a proximal-to-distal oxygen gradient (pO₂ of 3 ~ 7% in the mid-stomach, 2 ~ 4% in the mid-duodenum, ~1% in the mid-small intestine and <0.4% in the distal colon), while its basal compartment is oxygenated by a radial oxygen gradient across the epithelium (pO₂ of 7-10% across the small intestine wall, around 3% at the villus tip, 2% in the small intestinal lumen, and ~0.4% in the lumen of ascending and sigmoid colon) via passive subepithelial oxygen diffusion (54, 67). The low oxygen profiles in the intestinal lumen establish favorable environments for the residence of aerobic and anaerobic commensal microbes that modulate host physiological functions (19). Thus, although the recreation of the in vivo oxygen gradient has raised challenges for researchers, it should still be considered in intestinal tissue engineering. We previously demonstrated the use of the 3D luminal geometry with a tubular silk scaffold system to attain low oxygen conditions in the lumen without exposing the epithelial cells to a low oxygen cultivation atmosphere (32, 33). The creation of the oxygen gradients was due to the superior rate of oxygen consumption of the active epithelial cells lining in the 3D luminal surface compared to the diffusion rate. Following this mechanism, in the present study, we took advantage of the 3D geometry of the half lumen in the scaffolds by orientating the 3D half scaffold tissues “up” and “flipped” in culture and monitored in real-time the oxygenation in the tissue lumen. We successfully approached stable low oxygen gradients covering microaerobic conditions (pO₂ between 5% and 7%, scaffolds “up”, the oxygen level in mid-stomach) and anaerobic conditions (pO₂ ~1-2%, scaffolds “flipped”, the oxygen level in intestines) that mimic the in vivo proximal-to-distal oxygen gradient (Fig 4), and an oxygen gradient in the “flipped” scaffolds starting from culture medium (pO₂ ~14.95%) to scaffold bulk (pO₂ ~9.87%, oxygen level in the intestinal wall) moving to the lumen (pO₂ ~1-2%, the oxygen level in intestines) which resembles the in vivo radial oxygen gradient (Supplementary Fig 1). Importantly, the radial oxygen gradient in the “flipped” scaffolds recapitulated the anaerobic-aerobic interface between scaffold bulk and lumen where the epithelium resides (Supplementary Fig 1B). The interface permits the gut microbiome to survive in luminal hypoxia and the epithelial cells to be oxygenated by diffusion in the bulk silk scaffold matrix. This biomimetic oxygen gradient is analogous to the anatomical configuration of the microenvironments of native intestine, in which the microbiome lives in the anaerobic lumen and the epithelium continuously receives oxygen supplied by the underlying capillary vessels (68). Interestingly, the oxygen levels detected in the “up” scaffolds ranged from 5% to 7% which was lower than the oxygen tension (10-13%) that could be maintained in the confluent epithelial monolayers on Transwells (32). This could be explained by various phenomena. For example, compared to 2D substrates, 3D epithelial cells residing in the 3D curved environments more closely resemble the in vivo cell environments and more actively consume oxygen, since increased epithelial cell viability and proliferation has been observed in concave areas in growing epithelia, such as mammary (69) and renal epithelia (70). Another explanation could be that compared to 2D substrates, 3D curvature topography increases surface area to host a higher density of cells for oxygen consumption. It is worth mentioning that in the present study we were able to establish intestinal oxygen gradients simply by reorienting scaffolds in culture without the need for specialized equipment or consumption of large quantities of expensive media. Geometric control over tissue oxygen tensions could not be accomplished in any planar tissue culture systems including Transwells.

Besides the low level of luminal oxygen, the healthy intestine features a large gut microbial community of facultative and obligate anaerobes (3). Knowing that low oxygen gradients could be achieved in our 3D tissue systems, we determined if these hypoxic culture conditions enabled the co-culture of the intestinal epithelium with viable human microbiome. The enteroid-derived epithelium in both scaffolds “up” and “flipped” could be cocultured with human microbiome up to 48 hours (Fig 5) without compromising epithelial barrier function (Fig 6) at a proper MOI, suggesting that the aerobic-anaerobic interface formed in the scaffolds was sufficient to sustain host-microbiome interactions. When analyzing the bacterial diversity in these cocultures, the number of phyla and genera in the 3D coculture systems were lower than that of the intact human stool sample (Fig 7A), which was expected. This decrease in phyla and genera has multiple explanations: i) less fastidious members of the community are more likely to succeed in a competitive environment and out-compete the slower growing organisms (71), ii) the medium used in the cocultures was optimized for the intestinal organoids and may have been missing components required for survival or contains components preventing the growth of some organisms (72), and iii) partitioned oxygen concentrations in the scaffolds “up” and “flipped” were preferred by different microbes and therefore narrowed the microbiome diversity in the cocultures (73).

Another interesting observation was the upsurge of Escherichia coli/Shigella and Streptococcus in all cocultures and times (Fig 7B and C). This likely occurred as these groups of bacteria are facultative anaerobes which are capable of expanding aerobically or microaerobically (in the scaffolds “up”) or even anaerobically using fermentation (scaffolds “flipped”) (74). In addition, the interaction between the bacteria and the host epithelium may contribute to their predominance, as some species of E. coli grow more rapidly within host tissues than as isolated E. coli (75). E.coli has a doubling time of 15-20 minutes under favorable conditions (76), so these bacteria may grow faster than other groups identified in the cocultures, such as Bacteroids, which have a doubling time of 40-60 minutes in optimal conditions (77). Although some serotypes of E. coli and Streptococcus are pathogenic, most members of these groups inhabit the healthy human gut and form part of the commensal microbiota. They play a beneficial role in preventing the colonization of the intestine by pathogens and reducing inflammation in patients with IBD (78). Furthermore, since E. coli/Shigella and Streptococcus are known pioneer colonizers of the GI tract (18), their presence possibly functions to initiate and sustain the hypoxic environment during gut development by metabolically depleting the gut luminal oxygen, which allows obligate anaerobes to thrive and grow. Our 3D half scaffolds “flipped” maintained oxygen levels as low as ~1%, yet at this level multiple obligate anaerobes, such as Megasphaera, Bacteroides thetaiotaomicron, Lachnoclostridium, and Anaerostipes increased in abundance over time, suggesting strict anaerobic conditions in the lumens. The initial anaerobic conditions in the gut lumen are created by aerobic and facultative anaerobic microbes, that consume oxygen and thus provide an optimal niche for obligate anaerobic microbes (18, 73). Following this mechanism, the scenario in the “flipped” scaffolds could be similar, in that the initial colonization with some facultative anaerobes (e.g., Escherichia coli/Shigella and Streptococcus) further decreased oxygen concentration in the hypoxic lumen, creating strict anaerobic conditions favoring the growth of the obligate anaerobes. In this context, microbiome colonization in the 3D intestinal tissues simulates the temporal microbiome dynamics along the native human gut.

Conclusions

The human intestinal epithelium is a complex tissue composed of multiple cell types covering the innermost surface of the 3D intestinal lumen. It functions as a protective barrier, while simultaneously absorbing nutrients and maintaining an extremely low luminal oxygen tension to harbors trillions of commensal microbiomes. Although numerous efforts have been made in the scientific community, these unique features of human intestinal epithelium have made it particularly challenging to recreate in vitro. In this study, we demonstrated a relatively easy-to-use 3D half-tubular scaffold system that recapitulates the 3D human intestine structure for the in vitro recreation of human intestinal tissue. Further, we were able to establish two physiologically relevant oxygen gradients: a proximal-to-distal oxygen gradient along the GI tract and a radial oxygen gradient across the epithelium. The presence of the low oxygen profiles enabled the coculture of intestinal epithelium with facultative and obligate anaerobic human microbiome communities. By properly positioning the 3D tissues cultured under normal atmospheric conditions combined with oxygen consumption of the epithelial cells in the scaffold lumens, the oxygen gradients were autonomously generated in less than 30 minutes. By avoiding complex perfusion systems of liquid and gas in our 3D tissue system, the system was more user-friendly for users. One limitation of the work is that we only monitored the oxygen profiles and the microbiome co-culture in the system for 48 hours. This was designed as a proof-of-concept study to demonstrate that geometric control over tissue oxygen tensions could be accomplished by properly positioning the scaffolds in culture to support the co-culture of the human microbiome. In the future, the interplay between host and microbiome under different oxygen profiles can be investigated in prolonged culture. Fecal microbiota does not represent the full natural diversity of the gastrointestinal microbiota. In the future, niche-specific (e.g., duodenum, ileum, cecum, colon, rectum) gut microbiome compositions should be used to evaluate the diversity of interactions of the cells from different intestinal regions with their specific microbial communities. The promising features of the tissue system present here will provide the basis for the studies concerning gut barrier function, host-microbiome interactions and gastrointestinal diseases.