Inflammatory bowel disease

The prevalence of inflammatory bowel disease (IBD) in North America ranges from 0.3% to 0.5%, and incidence is increasing starkly, especially in younger age groups [1]. IBD is divided into two main categories: ulcerative colitis (UC) and Crohn’s disease (CD). The disease can be debilitating; 32.9% of patients with Crohn’s disease eventually become disabled by the disease and 80% require at least one surgery in their lifetime, while between 5 and 15% of patients with UC require a colectomy to remove all or part of the colon [2, 3]. Despite the availability of multiple drug options such as amino salicylates, glucocorticoid (GC), immunosuppressive (such as azathioprine methotrexate), and TNF-α monoclonal antibodies, a large proportion of patients either do not respond or lose response to therapy [4]. Therefore, there is an urgent need to understand the molecular mechanisms underlying the pathogenesis of IBD to develop new therapeutic approaches to treat this debilitating disease.

Non-immune cell inflammatory responses present an opportunity for new and localized therapeutics

Since 1997, researchers have recognized that “epithelial cells play a critical role in the inflammatory response” [5]. Epithelial cells manufacture and secrete cytokines and chemokines to regulate inflammation and signal recruitment of immune cells, and they synthesize growth factors with autocrine actions to facilitate their own repair [5]. It has been suggested that epithelial cells are an important source of pro-inflammatory cytokine signaling and may be used to control intestinal homeostasis to some extent [6]. Their influence over inflammatory signaling makes them particularly interesting targets for alleviating symptoms of IBD, and pausing the iconic feed-forward cycle of inflammation characteristic of the disease.

Intestinal epithelial cells (IECs) change their inflammation-related expression profiles in the context of IBD. For example, Toll-like receptors (TLR), critical innate immune system pathway regulators, are known to exhibit altered expression profiles in IBD patients. Specifically, TLR4 is overexpressed, and receptors shift from the basolateral side of cells to the apical side. In contrast, TLR2 and TLR5 are expressed at significantly lower levels in IBD [6]. Colonic epithelial cells isolated from active IBD patients secrete the neutrophil-attracting cytokine IL-8, which then exacerbates the harmful feed-forward cycle of inflammation observed [7, 8]. IL-21R has also been observed to be upregulated in IECs of IBD patients; this cytokine leads to increased CCL20 synthesis, which subsequently attracts T cells and dendritic cells [9, 10]. The IL-22 inflammatory cascade has also been implicated in IBD, wherein IL-22 is increased, leading to the dysregulation of intestinal epithelial cell (IEC) proliferation and migration, and stimulating IEC secretion of IL-1, TNF-α, IL-6, and IL-8 [11].

Epithelial cells may also play a major role in adaptive immune response. These cells express the antigen-presenting major histocompatibility complex II (MHCII) mostly on the villi, but in IBD, MHCII is instead upregulated in the small intestinal crypts, a change mostly driven by IFN-γ [12, 13]. In vitro studies have shown that suppressor T cells can be signalled to proliferate by antigen-presenting epithelial cells, contributing to the development of oral tolerance [14].

It is also clear that T cell originating cytokines IFN-γ and TNF-α can control epithelial cell properties by multiple mechanisms including disrupting barrier function (i.e. increasing gut permeability) [15, 16]. Disruption of barrier function is one of the hallmarks of IBD; poor barrier function can lead to antigen leakage into the lamina propria, creating inflammatory responses of massive proportions. While the direct cause of IBD remains unknown, growing evidence suggests it may be caused by a disruption in the finely controlled equilibrium between host response to antigens in the intestinal epithelium [17]. Therefore, there is an urgent need for the development of models that reflect this disruption in equilibrium.

In vitro models of non-immune cell inflammatory response

In vitro models enable the investigation of the precise function of candidate molecules to unravel cellular mechanisms and identify novel specific targets [5]. Existing in vitro models have been indispensable in developing the current IBD therapies, however, they remain insufficient as they largely focus on very specific, rather than general, abnormalities and mechanisms [17].

The most common in vitro models of IBD involve treating immortalized cell lines (e.g. Caco2, HT29, T84) with high concentrations of dextran sodium sulfate (DSS) or lipopolysaccharide (LPS) [17]. Though often utilized in vivo, both DSS and LPS can be used in vitro to elicit many of the same phenotypes; treatment with DSS is directly toxic to colonic epithelium; it causes ulcers, loss of crypts, and infiltration of granulocytes, all of which are characteristic of IBD [18–20]. This model lends itself to the study of factors that maintain or reestablish barrier integrity, or factors that protect against toxicity. However, it should not be considered an accurate mimic of human IBD, as it assumes a specific mechanism of origin, massive epithelial damage, which is unlikely to be the cause of disease in patients [18]. LPS treatment, or alternatively infection with invasive E. coli or Salmonella, activates extreme innate immune responses to bacterial components as a way of simulating IBD. This model lends itself to the study of factors that prevent or inhibit infection, but should not be considered an accurate model as it assumes an infectious disease origin of IBD [18, 21, 22].

Cytokine-based, non-immune cell models of IBD do also exist, though they mainly focus on barrier integrity measurements in epithelial cells. Amasheh et. al treated primary rat colon cells with TNF-α and IFN-γ in Ussing chambers, and showed that changes in barrier function and mucosal structure did occur, as would be expected in IBD [23]. However, the use of cells from a different species limits the generalizability of these conclusions to humans. Others have shown that treating Caco-2 cells co-cultured with macrophages with IL-1β and IFNγ results in a decrease in transepithelial electrical resistance (TEER), which indicates a decrease in barrier integrity [24]. Dosh et. al report that Caco-2 and HT29-MTX co-cultures grown on hydrogel scaffolds and treated with IL-1β and TNF-α exhibit changes in tight junction protein expression (ALP and ZO-1), as well as an increase in apoptosis signaling (caspase 3), both of which may indicate a decrease in barrier function [25]. As discussed above, non-immune cells play a vital role in cytokine and chemokine signaling that is not limited to barrier function.

Despite the role epithelial cells play in the inflammatory response, there is currently a dearth of in vitro models for the study of factors that interrupt or reduce the non-immune cytokine response characteristic of IBD. Here, we present a gut-on-a-chip co-culture model that stimulates non-immune cells with T-cell originating pro-inflammatory cytokines to simulate the non-immune cytokine response.

Gut on a chip

Organ-on-a-chip models are robust platforms to explore pathophysiological mechanisms in multiple diseases. In this model category, multiple cell types are co-cultured and subjected to tissue microenvironmental forces such as dynamic flow and/or mechanoactuation [26]. In the case of the gut on a chip, the result is a more physiologically relevant system that recapitulates gut architecture (crypts and villi-like structures), gene expression, and infection dynamics [26–29]. Lastly, cells in this system can be cultured for up to 21 days, enabling the study of longer-term therapeutic effects [26].

Permeability

We then tested the apparent permeability of cell layers using the Cascade Blue fluorescent marker as a permeability tracer. Cascade Blue cannot be metabolized, allowing us to test the permeability of the cell layer by looking at the ratio of salt in the top vs. bottom channels. The marker was loaded into inlet reservoirs on Day 1, and quantity of marker found in the bottom channel outlet media at the indicated times was used to assess barrier function (see Methods). We found that 25 added to either the bottom or the top and bottom channel did not result in significant changes in barrier function. However, 100 of pro-inflammatory cytokines added to the bottom channel did increase apparent permeability significantly and consistently across all 3 days measured (Fig 3).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Caco and HUVEC change in permeability due to inflammatory challenge.

Apparent permeability after treatment with: (A) 25 of TNF-α and IFN-γ in the bottom channel (B) 25 of TNF-α and IFN-γ in the top and bottom channel (C) 100 of TNF-α and IFN-γ in the bottom channel. Cells were treated with pro-inflammatory cytokines on Day1, and concentrations are reported as log fold change over untreated controls.

https://doi.org/10.1371/journal.pone.0289314.g003

Model usage

The described model was developed using the Emulate organ-on-a-chip basic research kit (Emulate, MA).We report here the cytokine secretion and permeability characteristics of Caco2 epithelial and HUVEC endothelial cells treated with TNF-α and IFN-γ in different concentrations and in different orientations on a gut-on-a-chip. Caco2 cells were seeded in the top channel, and HUVEC cells in the bottom channel. As such, responsive cytokines/chemokines found in the bottom channel may be produced by HUVEC cells, and/or by basolateral secretion from Caco2 cells. Responsive cytokines/chemokines found in the top channel may be produced by apical secretion from Caco2 cells, or by diffusion of HUVEC-produced cytokines through a disrupted Caco2 cell barrier. Overall, we found that 25 of cytokines in the bottom channel produced CXCL-10 and IL-6 in both channels, but IL-8 only in the bottom channel; 25 added to both channels produced CXCL-10 and IL-8 in both channels, but IL-6 only in the bottom channel; and 100 in the bottom channel triggered production of CXCL-10 and IL-8 in both channels, but IL-6 only in the top channel.

100 in the bottom channel was the only treatment that also disrupted barrier function as is seen in inflammatory bowel disease. Further investigations are required to determine the molecular mechanisms of this observed loss in barrier function, those it was determined that increased permeability was not cell death-related.

In conclusion, we recommend the use of this system with 25 in the top and bottom channel for investigation of factors surrounding cytokine/chemokine production, as this treatment resulted in the most consistent secondary cytokine response in both channels. We recommend the use of 100 in the bottom channel for the investigation of factors surrounding barrier integrity of gastrointestinal cells, as this treatment was the only one observed to disrupt barrier function.

Spatial considerations

It is worthwhile to note the spatial considerations of the model presented. 25 of cytokine stimulus in the bottom channel produced IL-8 in the bottom but not the top channel. This is consistent with previous literature, which found that Caco-2 cells exhibit polarized IL-8 secretion, secreting apical IL-8 when treated apically, and both basal and apical IL-8 when treated basolaterally [30]. Given higher TNFα concentrations found in the intestinal lumen in IBD, it is relevant to explicitly differentiate between the luminal and circulating concentrations of IL-8 [31]. Additionally, contrary to expectation, IL-6 was not increased in the top channel when high concentrations of cytokine are added to the bottom channel (100 ), but was observed when only 25 was added to the bottom channel. This may be related to the dual role of IL-6 as a pro- and anti-inflammatory cytokine; in chronic inflammation, IL-6 is considered pro-inflammatory, while in acute inflammation, IL-6 is known to be protective against aberrant inflammation [32–34]. We hypothesize that epithelial cell-secreted IL-6, in comparison to that produced by macrophages and other immune cells, may only be produced and secreted apically under low pro-inflammatory stimulus. This is consistent with work by Cudicini et al. which finds that Sertoli cells, as subtype of epithelial cell, are found to predominantly secrete IL-6 apically [35]. CXCL-10 exhibited no differences in terms of vectorial secretion, and was also the most consistently secreted under all conditions. Further investigation culturing single cell-types within the dynamic chip environment (e.g. mechanical and flow forces) is necessary to fully characterize the vectorial secretion of cytokines as well to determine the effect of cell-cell interactions on inflammatory response.

Cytokines relevant to this model

In this work, we report on a model investigating the non-immune cell inflammatory response as it relates to IBD. We induced inflammation in epithelial and endothelial cells using major pro-inflammatory cytokines observed in IBD, such as TNF-α and IFN-γ, then measured the secretion of other cytokine/chemokines relevant to IBD, including CXCL-10, IL-6, and IL-18.

Elevated IFN-γ levels are characteristic in CD, and prevention of IFN-γ production can significantly alleviate disease symptoms [15].

TNF-α has a significant function in IBD pathogenesis by increasing IL-1β, IL-6, and IL-33 expression. The clinical severity of UC and CD correlates with TNF-α levels in the serum of IBD patients [36, 37]. Anti-TNF therapy is very effective at treating IBD symptoms, leading to mucosal healing, reduced hospitalizations and surgeries, and improved patient quality of life [38]. Anti-TNF agents are more effective in patients with a shorter disease history, and are often used earlier during IBD therapy [39].

On the other hand, IL-8 is a ubiquitous neutrophil attracting chemokine that has been proposed as a potential diagnostic or prognostic marker for several GI disease states [40]. It is well established that IL-8 production is increased in the tissue of IBD patients, especially UC patients, compared with healthy controls [41–44]. As a result, therapeutics that target IL-8 are currently under development. One such example is an IL-8 blocking antibody shown to significantly inhibit the recruitment of neutrophils and decrease markers of inflammation in patients with palmoplantar pustulosis (PPP), a rare chronic inflammation of the skin. This finding suggests the potential use of anti-IL-8 therapeutics in treating IBD [45].

IL-6 is a main proinflammatory cytokine that activates signal transducer and activator of transcription 3 (STAT3), and has a key role in the pathogenesis of UC and the carcinogenesis of colorectal cancers related to UC [46]. STAT3 expression is induced by TNF-α, and STAT3 levels are increased in the tissues of ulcerative colitis and Crohn’s disease patients [47–49]. Blocking IL-6 signal transduction led to clinical remission in 19% of patients in a small clinical trial, suggesting that decreasing IL-6 response may be a promising therapeutic avenue [50]

CXCL-10 is a chemokine characteristic of IBD that attracts Th1-polarized effector T cells to inflamed sites. There is a positive feedback loop in which IFN-γ induces the production of this chemokine, which then recruits T cells that produce IFN-γ [51]. CXCL-10 levels are increased in active IBD, both UC and CD, mainly localized to mucosal epithelial cells, and its expression is inducible by IFN-γ [52]. Treatment with anti-CXCL10 during colitis in mice decreased clinical and histological disease severity and lowered mononuclear and TH1 cell recruitment, suggesting its promise as a therapeutic target for IBD [53].

Advantages and limitations of the described system

The co-culture system presented here provides a framework through which small molecules, microbial products, or even live biologic therapeutics may be tested in vitro to investigate inflammatory action and mechanism of action in epithelial and endothelial cells. Different phenotypes can be created by simply changing the concentration, orientation, and type of inflammatory cytokine, and multiple output metrics (e.g. cytokine concentration, permeability characteristics) may be measured over multiple days. The system is simple, highly reproducible, and does not assume any original cause of inflammation. This flexibility enables the study of mechanisms relevant to multiple IBD-subtypes, and may be expanded to other applications, such as inhibiting cytokine storm phenomenon in other diseases like SarsCov2, where non-immune cell cytokine secretion play a critical role [54, 55]. We showed here the use of this model to report cytokine secretion of epithelial and endothelial cells as markers of response to inflammatory stimuli. However, the system could be adapted for the assessment of other relevant responses, such as changes in MHCII expression patterns, differentiation state of the epithelium, TLR expression, and gap junction protein expression. Further characterization of the model is necessary to develop a full platform on which to screen anti-inflammatory therapeutics or microbial products. Importantly, this gut-chip model reconstitutes elements of IBD, and presents conditions that better resemble the bowel as a dynamic barrier subject to mechanical forces in comparison to traditional 2D systems.

Cell lines and culture conditions

Human intestinal epithelial cell line Caco-2 cells were cultured in DMEM (high glucose, GlutaMAX) supplemented with 20% FBS and 1% penicillin-streptomycin. Human umbilical vein endothelial (HUVEC) cells were cultured in endothelial growth media (EGM2 bullet kit). Caco2 cells were used on passage 58, HUVEC cells were used on passage 6. Cells were grown in 10 cm plates to the appropriate density for the cell type being tested, 80–100%. Cell cultures were incubated in a humidified 5% CO₂ incubator at 37 degrees C.

Co-cultures of Caco-2 and HUVEC on gut-on-a-chip

Protocols provided by Emulate Inc were followed to co-culture Caco2 and HUVEC cells in the organ chip model. In greater detail:

Day -1:

To prepare chip for cultures, the microporous membrane was first activated using UV light (10 min, wash, then 5 min). The membrane was then coated with extracellular matrix (Matrigel and TypeI collagen), and allowed to incubate at 37C, 5% CO2 overnight.

Day 0:

HUVEC cells were incubated with trypsin-EDTA for 3 min. Detached cells were collected and centrifuged for 4 min at 400 g, and resuspended in fresh media to a concentration of 7x10⁶ cells / ml. Cells were then seeded onto the bottom channel of the chip, and chips were inverted and incubated at 37 degrees C for 2 hours to allow cells to attach to the bottom side of the microporous membrane.

Caco2 cells were incubated with trypsin-EDTA for 3 min. Detached cells were collected and centrifuged for 4 min at 400 g, and resuspended in fresh media to a concentration of 3x10⁶ cells / ml. Cells were then seeded onto the top channel of the chip, and chips were again incubated at 37 degrees C, 5% CO2 for 2 hours to allow cells to attach to the top side of the microporous membrane.

After adherence, a gravity wash, wherein cells were gently washed with fresh media, was performed to give cells sufficient access to media, and chips were incubated at 37 degrees C, 5% CO2 overnight.

Day 1:

The following day, Caco2 media (DMEM + 20% FBS) and HUVEC media (EGM-2) began a continuous flow through the top and bottom channel respectively at a rate of 30 μl/hr. This flow was continued for the remainder of the experiment.

Day 2:

Peristaltic-like force stretching the central membrane 2% of its width at a frequency of 0.15 Hz was applied. Membrane was stretched at this frequency continuously from this point onward.

Day 3:

Excess media was aspirated from the reservoirs and replaced with fresh media. The concentration of FBS in HUVEC media was changed to 0.5% FBS to discourage overgrowth. Peristaltic-like force was increased to deform the membrane 10% of its width.

Day 6:

Media in the top channel was aspirated and replaced with Caco2 media supplemented 10% FBS, to discourage overgrowth in long-term co-culture.

Day 7:

Media was aspirated and replaced. In the bottom channel, HUVEC media with 0.5% FBS was mixed with appropriate cytokine for the testing condition. In the top channel, Caco2 media with 10% FBS was mixed with appropriate cytokine treatment for the testing conditions and 100 of Cascade Blue fluorescent permeability tracer.

Day 8–10 (Day 2–4 in the results plots):

Media from inlet and outlet ports was collected and stored at -80C. Cytokine and fluorescent marker concentrations were later measured using appropriate assays.

Cytokine treatment

Three cytokine treatment conditions were tested: TNF-α and IFN-γ in the bottom channel, TNF-α and IFN-γ in both channels, and TNF-α and IFN-γ in the bottom channel. On Day 6, Caco2 and HUVEC media were prepared as normal, top channel cytokines were added to Caco2 media, and bottom channel cytokines were added to HUVEC media. Cell culture media was placed into appropriate inlet reservoirs. Effluent media was collected on days 7, 8, and 9 (days 1–3 after cytokine treatment, as expressed in Figs 2 and 3) to measure cytokine concentration and permeability.

Quantifying cytokine secretion

Effluent media was collected from each channel, and cytokines IL-8, IL-6, and chemokine CXCL10 were measured using an ELISA assay per manufacturer directions (R&D Systems DuoSet).

Permeability measurements

On Day 6, Cascade Blue fluorescent salt () was added to Caco2 media (top channel). Effluent media was collected and Cascade Blue was quantified using a plate reader (excitation: 380 nm; emission: 460 nm). The following equation was used to calculate apparent permeability: where P_app with the apparent permeability in ; SA is the surface area of sections of the channel that overlap (0.17 cm²); Q_R and Q_D are the fluid flow rates in the dosing and receiving channels (cm³/s); and C_R,0 and C_D,0 are the recovered concentrations in the dosing and receiving channels respectively, in any consistent units (units will cancel out). (https://emulatebio.com/wp-content/uploads/2021/06/EP187_v1.0_Barrier_Function_Analysis_Protocol.pdf).

Monocultures in 2D

Cells were grown in 10 cm dishes to 80–100% density. Media was removed and plates were rinsed with 10 ml DPBS or HBSS depending on the cell type. Cells were then incubated with 2 ml 0.05% trypsin for 3–5 minutes, and trypsin neutralized with 8 ml of DMEM/10% FBS. Cells were centrifuge at 400xg for 4 min and the pellet resuspended in 2 ml of the appropriate cell culture media. Cells were seeded in 96-well plates at a density of 40,000 cells/well for high density plates. Cells were initially plated in 100 μl media/well and incubated at 37 degrees C, 5% CO2 for 2–3 hours to allow attachment before the addition of treatments. Treatments were prepared in complete media via serial dilution at a concentration of 2x the final concentration and then added to the plates at a volume of 100 μl/well. Plates were then incubated at 37C, 5% CO2 for 48 hours. Media was extracted from each well individually and cytokine concentration was measured by ELISA assay as described above.