Gut microbe–skin axis on a chip for reproducing the inflammatory crosstalk

Abstract

The gut–skin axis has emerged as a crucial mediator of skin diseases, with mounting evidence highlighting the influence of gut microbiota on skin health. However, investigating these mechanisms has been hindered by the lack of experimental systems that enable direct study of gut microbiota–skin interactions. Here, we present the gut microbe–skin chip (GMS chip), a novel microfluidic platform designed to model microbiome–gut–skin axis interactions. The GMS chip allows the coculture of intestinal epithelial cells (Caco-2), human epidermal keratinocytes (HEKa), and gut microbes with fluidic connection mimicking the blood flow. We validated that the gut compartment, with a self-sustaining oxygen gradient, enabled coculturing gut bacteria such as Escherichia coli (E. coli) and Lactobacillus rhamnosus GG (LGG), and the skin cells properly differentiated in the chip in the presence of fluid flow. Disruption of intestinal epithelial integrity by dextran sodium sulfate (DSS) combined with lipopolysaccharides (LPS) selectively decreased skin cell viability while sparing gut cells. Notably, pretreatment with LGG showed a protective effect against the skin cell damage by enhancing the intestinal barrier function. The GMS chip effectively recapitulates the influence of gut microbiota on skin health, representing a pivotal step forward in studying gut–skin axis mechanisms and the role of the gut microbiome in skin diseases.

---

Introduction

Recent studies have highlighted the crucial roles of gut microbiota in intestinal homeostasis. Commensal bacteria contribute to host health by producing vitamins and extracting additional energy from undigested dietary sources, such as fibers.^1,2 Short-chain fatty acids (SCFAs), major bacterial metabolites, alleviate inflammation and reinforce tight junctions in the intestinal epithelium.² Additionally, commensal gut bacteria produce antimicrobial peptides (AMPs) and bacteriocins that suppress pathogenic expansion, protecting host tissues from infection and inflammation.¹

Gut dysbiosis, a state of microbial imbalance induced by factors such as antibiotics, unhealthy diet, or ingestion of pathogenic bacteria, can impair the intestinal barrier and potentially lead to leaky gut syndrome.³ Consequently, harmful molecules, including bacterial endotoxins, may infiltrate the systemic circulation and affect other organs, for example, the skin. Clinically, it has been known that gastrointestinal disorders often coincide with skin conditions such as acne, atopic dermatitis, and psoriasis.⁴ Research indicates a higher prevalence of gut dysbiosis in patients with skin diseases compared to healthy individuals, supporting the concept of a gut–skin axis.⁵ The growing evidence for a gut–skin axis supports the important roles of gut microbiota in modulating skin health. For instance, phenol and para-cresol, metabolites produced by certain gut pathogens, have been shown to reduce keratin 10 expression in epidermal keratinocytes, thereby compromising the skin barrier function.⁶ Conversely, oral administration of probiotics alleviated atopic symptoms by improving the barrier function and reducing the need for steroid treatments.⁷ These studies underscore the substantial impact of gut bacteria on skin conditions and highlight the importance of understanding gut–skin interactions for improved disease prediction and treatment strategies. Clinical studies and big data analyses significantly expanded our understanding of the complex relationship between the gut and skin. However, unraveling the mechanisms of their interactions remains challenging due to their complexity. Consequently, the specific role of the gut microbes in the gut–skin axis remains relatively underexplored.

Microphysiological systems (MPS), also known as organs-on-chips, which aim to replicate human physiology in vitro, offer more physiologically accurate models compared to conventional 2D cultures by incorporating 3D tissue structure, fluid flow and/or mechanical movements. In addition, MPS can be valuable tools for building a multi-organ model through fluidic connections between different organ models.^8,9 Previous studies have demonstrated successful cocultures of four and eight-organ models, affirming the potential of MPS in drug ADME profiling and toxicity assessments.^10,11 We previously reported gut–blood brain barrier (BBB) axis chip,¹² gut–liver chip,¹³ and gut–skin chip systems¹⁴ and demonstrated the influences of gut bacteria-derived molecules on exosome infiltration through the BBB and lipid accumulation in the hepatocytes. However, these systems lack live bacterial culturing, presenting only limited influences of gut bacteria on the host cells.

To address this limitation, we present the gut microbe–skin chip (GMS chip), a simple, pumpless microphysiological system that facilitates the coculture of the gut microbes, intestinal epithelial cells (IECs), and skin cells. The GMS chip utilizes two inserts: one mimicking gut tissue and the other mimicking skin tissue. These individually mountable inserts allow separate differentiation of cells in each organ model before integrating for coculture in the chip. These inserts are fluidically connected in a microfluidic chip with a gravity-driven flow (Fig. 1). This fluidic connection allows the exchange of molecules, such as lipopolysaccharide (LPS), between the two organ compartments in the chip. This design enables assessment of how gut bacteria cocultured in the luminal side of the gut compartment influence the skin compartment. Within the GMS chip, we generated the oxygen gradient mimicking that of the in vivo gut, which is one of the key physiological features of the gut. In the human intestine, the lumen is depleted of oxygen while the intestinal tissue is oxygenated through blood vessels (Fig. 1B).^15–17 To reproduce this oxygen environment of the gut, several intestinal model systems with physiological oxygen gradients have been developed, including microfluidic chips such as HuMix, intestine-on-a-chip, anoxic–oxic interface-on-a-chip (AOI), and a gut microbiome physiome platform GuMI.^18–21 These systems effectively reproduce the physiological oxygen environments of the in vivo gut. However, they require continuous perfusion of anoxic medium using a pump, making them difficult to use and be integrated into a multi-organ system. We introduced a cell culture insert that forms a self-sustaining oxygen gradient into the gut compartment.²² The device is composed of a cell culture insert for cell culture and a companion plug and rubber cap, which blocks oxygen influx to the luminal side of IECs. Then, the cellular oxygen consumption readily generates an anaerobic condition in the luminal compartment without perfusion of anoxic medium. The gut model in the self-sustaining oxygen gradient insert and a reconstructed human epidermis (RHE) model containing primary keratinocytes in a microfluidic chip were combined to build the GMS chip (Fig. 1C and D). We verified the oxygen gradient formation and gut bacterial coculture feasibility using Escherichia coli and Lactobacillus rhamnosus GG (LGG) in the gut compartment and confirmed the viability and proper differentiation of the RHE skin model when integrated into the chip. Furthermore, we modeled intestinal inflammation by treating the gut with dextran sulfate sodium (DSS) and LPS and assessed its influences on the IECs and RHE. Finally, we investigated whether LGG, a well-known probiotic strain, can alleviate DSS/LPS-induced inflammation. This novel gut microbe–skin chip provides a unique platform for investigating the complex interactions of the gut–skin axis, offering new insights into the role of gut microbiota in skin health and potential therapeutic strategies for related disorders.

Fig. 1 Configuration of the gut microbes–skin axis chip (GMS chip). (A) The schematic diagram of the entire chip. (B) The schematic diagram of the intestinal oxygen gradient in vivo. (C) The assembly (left) and the cross-sectional (right) diagram of the gut compartment. (D) Annotated exploded view of the GMS chip. (E) The photographic image of the gravity-driven flow machine. (F) The photographic image of the GMS chip.

Methods

Fabrication of the GMS chip with an oxygen-gradient generating cell culture insert

Our GMS chip consists of three cell chambers and two reservoirs (Fig. 1A) for multi-organ modeling. Each cell chamber can hold a cell culture insert for 24 well plates. The channels fluidically connect the cell chambers and the reservoirs. Using our GMS chip, up to three different tissue cultures can be integrated into the chip while fluidic channels connect the cultures, enabling chemical signaling.

The gut model was prepared in the cell culture insert designed to generate a physiological oxygen gradient (Fig. 1B and C).²² The inserts and plugs were fabricated from polycarbonate (PC). The cell growth area of the insert was designed to be the same as commercially available cell culture inserts for 24 well plates. The hole in the plug was used to discharge the air inside the insert when the plug was installed. Once the plug was installed, the hole was sealed with a rubber cap (6448K117, McMaster-Carr, USA). For oxygen measurement of the apical medium, a needle-type oxygen probe was installed through the rubber cap. For the cell culture, a PET membrane with 0.4 μm pores (Corning 3450, USA) was attached to the insert using epoxy adhesive (EpoxySet, EB-107LP-1, USA). The insert was autoclaved prior to use and reused after removing the used membrane. For the skin model, Transwell inserts with a 0.4 μm pore PC membrane (Corning 3413, USA) were used with a PDMS ring (outer diameter 12 mm, inner diameter 10 mm) to tightly fit the insert into the chip (Fig. 1D). Fluid flow in the basal side was induced by a gravity-driven flow machine (Fig. 1E) operated with a tilting angle of 15° and an angle change interval of 500 s.¹³Fig. 1F shows the picture of the assembled GMS chip.

Cell culture and bacterial coculture

For the gut compartment, Caco-2 cells with passage numbers between 35 and 50 were cultured in high-glucose DMEM (Gibco, 11995, USA) supplemented with 10% fetal bovine serum (FBS, Welgene, S001, Korea) and penicillin–streptomycin (Sigma, P4333, USA). Caco-2 cells were seeded at a density of 3 × 10⁵ cells per cm² in a cell culture insert and cultured in a 24-well plate for 14 days prior to bacterial coculture. One day before the bacterial coculture, the cell culture medium was replaced with the antibiotic-free culture medium after washing twice with PBS.

Escherichia coli (BL21(DE3); Enzynomic, Korea) was grown in Luria-Bertani (LB) medium containing tryptone 1% (Formedium, TRP03, UK), sodium chloride 1% (Glentham, 070DGR,UK), and yeast extract 0.5% (Formedium, YEA03, UK). Lactobacillus rhamnosus GG (LGG, ATCC (American Type Culture Collections, 53103, USA) was purchased from Korean Collection for Type Cultures (KCTC) and cultured in De Man–Rogosa–Sharpe (MRS) broth (KisanBio, MB-M1025, Korea). For overnight cultures, a single colony was cultured in a shaking incubator for 17 h. For bacterial coculture, the apical medium of the gut compartment was replaced with an antibiotic-free fresh medium, the plug was installed, and then bacteria at a specified density (10⁴, 10⁵, and 10⁶ CFU mL⁻¹ for the screening and 10⁵ cells per mL for GMS chip experiments) were inoculated through the hole in the plug using a syringe (Hamilton, 80300, USA). The gut cells in the inserts with the bacterial inoculum were then cocultured for 24 h in 24 well plates or moved to the GMS chip. To determine total viable bacterial counts, planktonic and adherent bacteria were recovered separately and used for enumeration by counting colony-forming units (CFUs) on agar plates. Planktonic bacteria were collected from the apical medium and two subsequent washes. Adherent bacteria (and Caco-2 cells) were detached from the porous membrane support by trypsin–EDTA treatment for 5 min at 37 °C, followed by centrifugation at 13 500 rpm for 5 min to isolate the cell pellet.

Human primary epidermal keratinocytes (HEKa, ATCC, PCS-200-011, USA) were routinely cultured in a T25 flask using KGM gold medium (Lonza, 192060, Switzerland). The cells with passage numbers 2–3 were used for the experiments. To generate the RHE model, cell culture inserts were coated with 100 μL of 2 mg mL⁻¹ dopamine hydrochloride (Sigma, H8502, USA) in 10 mM Tris buffer (pH 8.5, Chembio, CBT H-9085, Korea) for 2 h at room temperature, followed by deionized water rinsing, to minimize shrinkage. Subsequently, 100 μL of 8.3 × 10⁵ cells per mL were seeded onto each dopamine-coated cell culture insert. One day after seeding, the medium in the inserts was removed, and then the cells were cultured in an air–liquid interface to differentiate with medium exchange every day. The compositions of the culture media for the skin model and the cocultures are listed in Table S1.†

Immunofluorescence and hematoxylin and eosin (H&E) staining

To visualize the tight junction protein ZO-1, Caco-2 cells were fixed with 4% paraformaldehyde for 30 minutes at room temperature. Then, the samples were blocked with 1% BSA in PBS for 1 h at room temperature, and then incubated with 5 μg mL⁻¹ of primary antibody for ZO-1 (Invitrogen, 61-7300, USA) diluted in 1% BSA in PBS at 4 °C for 16 h. After thorough washing with 1% BSA, the secondary antibody conjugated with Alexa Fluor 488 (Invitrogen, A-1108) diluted in 1% BSA in PBS at a 1 : 1000 ratio was incubated with the samples for 1 h at room temperature. A confocal fluorescence microscope (Zeiss, LSM 700, Germany) was used to detect the fluorescence with a 488 nm laser for excitation.

An immunohistochemistry assay was performed to observe the cross-section of the RHE model. The RHE samples were fixed with 4% formaldehyde for at least 2 h, embedded in the paraffin, sectioned into 4 μm thick slides by microtoming, and then placed on a silane-coated glass slide. To remove the paraffin, the samples were dipped into xylene, pure ethanol, 90% ethanol, 70% ethanol, and finally, deionized water for 5 minutes each, and then rinsed with water to remove ethanol completely. H&E staining was performed on the samples placed in a PAP-pen-drawn circle on a glass slide by adding Mayer's hematoxylin for 5 minutes, rinsing with water, incubating with a bluing reagent for 15 minutes, rinsing with water, and immersing in ethanol for 1 minute, and finally treating with eosin for 30 s.

To label differentiation markers on sectioned samples, paraffin was removed from them as described above, and the samples were microwaved in TE buffer (Welgene, ML014, Korea) for 20 minutes to retrieve the antigen. Then, the samples were blocked with 1% BSA in PBS for 1 hour at room temperature. Primary antibodies for keratin 5 (Abcam, ab64081, UK), involucrin (Invitrogen, PA5-32454, USA), filaggrin (Abcam, ab221155, UK), and Ki67 (Abcam, ab16667, UK) diluted in 1% BSA in PBS (at 1 : 100, 1 : 100, 1 : 1000, and 1 : 100 respectively) were incubated with the samples for 16 h at 4 °C. Then the primary antibody solutions were removed, the samples were rinsed, and then the secondary antibody conjugated with Alexa fluor 488 (Abcam, A-1108, UK) diluted in 1% BSA in PBS at 1 : 1000 ratio was incubated with the samples for 1 hour at room temperature. After rinsing with PBS, the samples were incubated with 300 nM of DAPI (Invitrogen, D1306, USA) diluted in PBS to stain DNA. A Faramount mounting medium (TMbio, S3025, Korea) was used to mount the stained samples. Protein and DNA visualization was performed using a Nikon Ts2R-FL fluorescence microscope with two filter sets: one for proteins (excitation 470/40 nm, emission 534/55 nm) and another for DNA (excitation 390/38 nm, emission 475/90 nm).

Measurement and mathematical simulation of the oxygen concentration gradient

To measure the oxygen level inside the cell culture inserts with the oxygen gradient, an oxygen sensor (FireSting-GO2, Pyroscience) with a retractable oxygen probe (TROXR430, Pyroscience) was used. The retractable oxygen probe was installed in the cell culture insert by puncturing the EPDM rubber cap placed in the hole of the plug to measure the luminal oxygen level for 20 hours. The basal oxygen level was measured for 20 hours by placing the retractable oxygen probe vertically between the insert and the well plate.

The oxygen levels in the cell culture insert and the well plate (with the plug in place) were predicted with COMSOL Multiphysics (COMSOL Inc., MA, USA). In this model, oxygen diffuses from the atmosphere to the basal culture medium and to the cells through the porous membrane. The atmospheric oxygen is set to 17.2%, the oxygen level in a CO₂ incubator,²³ while the oxygen influx to the luminal side was set to zero. The viscosity and density of the culture medium were assumed to be similar to those for water (0.89 mPa s and 0.997 g mL⁻¹, respectively, at 25 °C). The diffusion coefficient of oxygen in the culture medium was assumed to be identical to that in water (2.95 × 10⁻⁹ m² s⁻¹ at 36.5 °C). The oxygen consumption rate (OCR, mol s⁻¹) in the medium was calculated by solving the differential equation of Michaelis–Menten kinetics:

Here, OCR_max is the maximum oxygen consumption rate and was set to 3.3 × 10⁻¹⁷ mol s⁻¹, the oxygen consumption rate of mitochondria with sufficient oxygen supply.²⁴C_O2,min, the minimum oxygen concentration a cell reaches, was set to 0.006 mol m⁻³, the lowest oxygen level measured from the measurements. 0.6 × 10⁻¹³ mol per cell was selected for K_m from the range reported in the literature (1.0 ± 0.4 × 10⁻¹³ mol per cell).²² Since the open area of the PET porous membrane is small (around 0.5%), the oxygen diffusion coefficient in PET, 7.0 × 10⁻¹³ m² s⁻¹, was used for oxygen diffusion through the 10 μm thick porous membrane.²⁵

Cell viability assay

The viability of the intestinal epithelial cells was assessed by staining the cells with calcein AM (Invitrogen, C3100MP, USA) and ethidium homodimer-1 (EthD-1, Invitrogen, E1169, USA). The cells were washed with PBS, and then 4 μM of calcein AM and EthD-1 in PBS were incubated for 30 minutes. The samples were imaged using a fluorescence microscope (Nikon, Ts2R-FL, calcein AM; excitation 470/40 nm, emission 534/55 nm, EthD-1; excitation 525/50 nm, emission 597/58 nm), and the cell viability was obtained from the fluorescence images using ImageJ.²⁶ When the cells needed to be detached from the porous membrane to enumerate the adherent microbes, trypan blue staining was used. Viability was calculated by dividing the number of live cells (calcein AM positive or trypan blue negative cells) by the number of total cells.

The viability of skin cells was estimated using the MTT assay kit (Abcam, ab211091, UK) following the manufacturer's protocol. The inserts with skin cells were transferred to a 24-well plate, and 100 μL and 600 μL of MTT reagent mixed with DMEM at a 1 : 1 ratio were added to the inserts and the well plates, respectively, and incubated for 2 hours at room temperature. After removing the MTT reagent, MTT solvent was added (400 μL and 1.5 mL for the inserts and the well plates, respectively) and incubated in a rocking shaker at 50 rpm for 3 hours at room temperature to dissolve formazan. Then, 100 μL of the solution from each sample was transferred to a 96-well plate, and optical density at 570 nm was measured using a spectrophotometer (Varian, 9102883) with untreated wells as the control.

Measurements of transepithelial electrical resistance (TEER) and permeability

To evaluate the intestinal epithelial barrier function, TEER was measured using a volt–ohm meter (Millipore, ERS-2) with a chopstick electrode in HBSS (Sigma, H6648, USA). The TEER values of cells were calculated by subtracting the TEER values of cell-free inserts from the TEER readings of the inserts containing the cells.

The permeability of the intestinal epithelium was estimated using 70 kDa FITC–dextran (Invitrogen, D1823, USA). For the inflammation model, cells were subjected to DSS (Sigma, 31404-5G-F) and LPS (Sigma, L2630, USA) treatment (1% or 1.5% and 100 μg mL⁻¹, respectively) for 20 hours, washed with PBS twice, and then 25 mg mL⁻¹ of FITC–dextran in PBS was added to the luminal side. The fluorescence of the basal medium was measured with a microplate reader (Thermo Fisher, Multiskan Go, 51119200, USA) with an excitation wavelength of 475 nm and an emission wavelength of 525 nm. The concentration of FITC–dextran was estimated using a calibration curve. The permeability was calculated following the equation:

Here, C₀ is the initial concentration of the FITC–dextran added to the luminal side, A is the transport area, and dQ/dt is the flux of the FITC–dextran.

Results

Oxygen gradient formation and cell viability in the gut model

To verify whether a physiological oxygen gradient can be generated in the gut model with the custom-made cell culture insert and the companion plug, the oxygen levels in the culture medium were simulated using COMSOL and experimentally measured with an oxygen probe. In the COMSOL simulation model, the luminal oxygen concentration dropped rapidly within the first hour to about 2% (Fig. 2A) and remained low for over 20 hours (Fig. 2B), while the oxygen concentration on the basal side remained above 15% for over 20 hours, clearly indicating the oxygen gradient can be generated. The oxygen measurements also confirmed experimentally that the oxygen gradient with an anaerobic luminal environment (p_O2 = 4.4 mmHg, 0.006 mol m⁻³, 0.6% oxygen) and oxygen-rich basal compartment (p_O2 = 117 mmHg, 0.161 mol m⁻³, 15.6% oxygen) was indeed generated in the gut model (Fig. 2B). The cell viability assessed by live/dead assay showed that the cell viability under the oxygen gradient was high (above 95%), similar to that under the normoxic condition (Fig. 2C and D), suggesting that the oxygenation through the basal side only is sufficient to maintain the high viability of gut epithelial (Caco-2) cells. In summary, we demonstrated that the oxygen gradient is successfully generated in the gut model without losing cell viability.

Fig. 2 Viability of Caco-2 cells under the oxygen gradient. (A) The simulated and (B) the measured oxygen concentrations of the culture medium in the cell culture inserts with plugs installed. The symbols and lines indicate the measured and simulated oxygen concentrations, respectively. (C) The representative images of viability assay on Caco-2 cells and (D) cell viability quantified from the cell viability assay. Calcein AM and EthD-1 were used for staining live and dead cells, respectively. n = 3, an unpaired t-test was used for statistical analysis.

Gut bacterial coculture in the gut model under the oxygen gradient

The feasibility of gut bacterial coculture with the intestinal epithelium under the oxygen gradient was evaluated. E. coli, one of the most frequently used model bacteria, at varying densities (10⁴, 10⁵, 10⁶ CFU mL⁻¹) was inoculated into the luminal side of the gut model with the oxygen gradient. The viability of the intestinal epithelium after 24 h of coculture exceeded 80% at low bacterial density (10⁴ CFU mL⁻¹). However, increasing bacterial density resulted in a marked decrease in cell viability (Fig. 3A). Both planktonic (in the supernatant) and adherent bacteria increased after 24 h in all bacterial densities, but the ratio of adherent to planktonic bacteria increased as bacterial inoculation density increased from 1 : 100 at 10⁴ CFU mL⁻¹ to 1 : 10 at 10⁶ CFU mL⁻¹ (Fig. 3B). When E. coli is cocultured without the oxygen gradient plug, i.e., under the aerobic condition, the viability of the intestinal epithelium was lower (51% at 10⁵ CFU mL⁻¹ seeding density, Fig. S1A†). Fewer bacteria were adherent in the aerobic condition than in the oxygen gradient, while the total number of bacteria in the cocultures was similar under the two conditions (Fig. S1B†). LGG, one of the most studied probiotic lactic acid bacteria, was also inoculated at the same varying density. Caco-2 cells exhibited over 80% viability at up to 10⁵ CFU mL⁻¹ of LGG (Fig. 3C), indicating higher tolerance of Caco-2 to LGG than E. coli. Interestingly, the number of adherent bacteria increased as inoculation density increased (Fig. 3D), while planktonic bacterial counts were at the plateau around 10⁸ CFU mL⁻¹ (Fig. 3D), which is comparable to the bacterial density of the human colon.²⁷ In contrast to the case of E. coli, the viability of Caco-2 cells and the LGG bacterial growth and adherence were similar under both the aerobic condition and the oxygen gradient (Fig. S1†). In summary, we demonstrate that our gut model enables gut bacterial coculture under an oxygen gradient with optimized bacterial inoculation density.

Fig. 3 Cell viability and bacterial growth in gut bacterial cocultures in the gut model under the oxygen gradient. (A and C) The cell viability and (B and D) bacterial growth after 24 h coculture of Caco-2 cells with E. coli (A and B) and LGG (C and D), respectively. n = 3.

Behaviors of the RHE model in the chip with fluid flow

We investigated whether a fluid flow affects the behaviors of the RHE model. The viability, assessed by MTT assay, of the keratinocytes cultured in the chip with the gravity-driven flow was statistically not different from that in a static culture in a well plate (Fig. 4A). We also verified whether the RHE models properly differentiated in the chip with the fluid flow and compared with those in a static culture. The RHE models, both in the (static) well plate and the chip with the fluid flow, exhibited multi-layers of the cells labeled with different differentiation markers (Fig. 4B). Proliferation marker protein Ki-67 positive cells were located at the bottom of the samples under both conditions, indicating normal cell proliferation from the basal layer. Also, keratin 5 (K5), involucrin (Inv), and filaggrin (Fil), which are a basal layer, a cornified layer, and a granular layer marker, respectively, were properly expressed in the RHE cultured in the chip and a well plate. In summary, we demonstrated that the RHE in the chip with the fluid flow exhibited high viability and normal differentiation patterns.

Fig. 4 Characterization of the RHE in well plates and the chip. (A) Relative cell viability of RHE models in well plates and chips. Viability of the RHE in well plates set to 100%. n = 3, an unpaired t-test with Welch's correction was used for statistical analysis. “ns” indicates “not significant”. (B) Representative cross-sectional images of RHE models in well plates and chips. HES: H&E staining. From the second row to the bottom, blue indicates nuclei, green indicates antibodies detecting target proteins listed on left side of the image. Scale bars = 100 μm.

Effect of LGG on the viability of the GMS chip treated with DSS/LPS

Then, we established a GMS chip by fluidically connecting the RHE and gut model (Fig. 5A). Cell viability remained high (>90%) in both compartments after 24 hours of coculture, as determined by calcein AM/Ethd-1 staining for IECs and MTT assay for keratinocytes (Fig. 5B), indicating that the fluidic connection does not compromise cell viability in either compartment. Furthermore, inoculating LGG in the gut compartment did not affect the viability of Caco-2 or HEKa cells (Fig. 5B), validating the GMS chip as an effective model for studying the gut–skin axis. To investigate the gut–skin interaction and potential probiotic effects of LGG, we exposed LGG to the luminal compartment of the gut model for 4 hours, allowing bacterial binding to Caco-2 cells. Subsequently, we treated the gut model with dextran sodium sulfate (DSS), a known inducer of inflammatory colitis^28,29 in the presence of LPS, the endotoxin from commensal Gram-negative bacteria, for 20 h (Fig. 5A). LPS treatment, either alone or in combination with LGG, did not influence Caco-2 cell viability (Fig. S2†). Similarly, 1% or 1.5% DSS treatment with LPS had minimal effect on Caco-2 cell viability (Fig. 5C). However, DSS/LPS treatment significantly decreased keratinocyte viability (Fig. 5D with DSS/LPS vs.Fig. 5B without DSS/LPS), suggesting that compromised gut barrier integrity may adversely affect skin health. Notably, LGG mitigated the DSS/LPS-induced decrease in keratinocyte viability (Fig. 5D), indicating a potential role for gut microbiota in maintaining skin homeostasis. These findings demonstrate that our GMS chip model effectively simulates the gut–skin axis in vitro, providing a valuable platform for investigating the intricate interactions between the gut microbiome and skin health.

Fig. 5 Cell viability of the gut and skin compartments with fluidic connection in the GMS chip. (A) The schematic of the experimental setup and the procedure. (B) Cell viability of the gut and skin cells in the GMS chip with and without LGG exposure. n = 3, unpaired t-tests were used for the statistical analyses. (C and D) Viability of Caco-2 cells (C) and HEKa cells (D) in DSS/LPS treatment in the absence and presence of LGG. n = 3, unpaired t-tests were used for the statistical analyses. *p < 0.05.

The protective role of LGG in gut barrier integrity

We hypothesized that the decreased viability of HEKa cells in the GMS chip was due to the permeation of LPS leaking through the intestinal epithelium disrupted by DSS despite the high Caco-2 cell viability. First, we examined barrier integrity using TEER measurements. While 1% DSS treatment did not significantly alter the TEER values, 1.5% DSS treatment decreased the TEER values by almost 50% (Fig. 6A). Notably, pre-treatment with LGG increased the TEER values in non-treated controls and the samples treated with 1.5% DSS. This was corroborated by permeability assay with FITC–dextran (70 kDa) with a similar molecular weight of LPS (Fig. 6B). DSS/LPS treatment significantly increased the permeability coefficient of the intestinal epithelium to 3.3 × 10⁻⁶ cm s⁻¹ and 6.4 × 10⁻⁶ cm s⁻¹ in 1% and 1.5% DSS-treated samples, respectively, compared to 0.46 × 10⁻⁶ cm s⁻¹ in non-treated controls. LGG markedly decreased the permeability coefficient of DSS-treated samples, 0.8 × 10⁻⁶ cm s⁻¹ and 2.5 × 10⁻⁶ cm s⁻¹ at 1% and 1.5% DSS treatments, respectively (Fig. 6B). Additionally, we assessed tight junction integrity through immunofluorescence of the tight junction protein ZO-1. Non-treated control and LGG-treated samples exhibited intact tight junctions. In contrast, DSS/LPS-treated samples exhibited marked disruption of tight junctions, which was attenuated by the pre-treatment with LGG (Fig. 6C). This observation aligns with previous reports on the probiotic effect of LGG to alleviate DSS-induced colitis.³⁰ These results collectively suggest that LGG can enhance the intestinal epithelial barrier function and inhibit the DSS/LPS-induced impairment of the intestinal epithelium. This protective effect may explain the observed mitigation of HEKa cell viability reduction in the presence of LGG, highlighting the potential role of probiotics in maintaining gut barrier integrity and, consequently, skin health in the context of the gut–skin axis.

Fig. 6 The influences of LPS/DSS treatment and LGG exposure on the intestinal barrier function. Changes in the TEER (A) and permeability coefficient (B) by DSS/LPS treatment in the presence and absence of LGG. n = 3, unpaired t-tests were used for the statistical analyses. (C) Representative confocal fluorescence images of the tight junction protein ZO-1 in the Caco-2 cells treated with DSS/LPS and/or LGG. The scale bar = 50 μm. *p < 0.05 and **p < 0.005.

Discussion

We developed a novel microphysiological system that effectively models the interactions between gut microbes, intestinal epithelium, and the skin. Our findings demonstrate the feasibility of creating a physiologically relevant in vitro model that integrates key aspects of the gut–skin axis, including oxygen gradients, bacterial coculture, and the impact of gut barrier disruption on skin health.

First, we evaluated whether the individual organ models, gut and skin compartments, mimic the key aspects of the in vivo organs. Our custom-made cell culture inserts and companion plugs used for the gut compartment generated a stable oxygen gradient with the anaerobic luminal environment (p_O2 = 4.4 mmHg) and oxygen-rich basal compartment (p_O2 = 117 mmHg), mimicking the oxygen gradient found in the intestine in vivo.³¹ This oxygen gradient was maintained for over 20 hours without compromising the viability of Caco-2 cells, demonstrating the robustness of our model. Our device does not require any sophisticated fabrication or manipulation to achieve the stable oxygen gradient, which contribute to the robustness and reproducibility of the system. We further demonstrated the successful coculture of intestinal epithelial cells with commensal (E. coli) and probiotic (LGG) bacteria under physiological oxygen gradients. Our results showed that bacterial inoculation density is important in maintaining epithelial cell viability, probably due to the absence of fluid flow washing out the excessive bacteria, with higher tolerance observed for probiotic LGG than E. coli. Increasing bacterial inoculation density is correlated with the ratio of adherent to planktonic bacteria rather than the total viable bacteria, except for the low inoculation density (10⁴ CFU mL⁻¹) of LGG. In addition, we confirmed that the skin model exhibits high viability and proper differentiation patterns under the fluid flow in the chip as well as in the static well plate.

Our results demonstrated that the fluidic connection between the two compartments did not compromise cell viability, allowing for the investigation of the systemic effects of gut perturbations on skin health or vice versa. Importantly, we showed that disruption of the intestinal epithelium by DSS/LPS treatment in the gut compartment led to decreased keratinocyte viability in the skin compartment, providing strong evidence of the systemic effects of gut barrier disruption on skin health in vitro. LPS binds to toll-like receptor (TLR) 4 and CD14, and induces secretion of pro-inflammatory cytokines such as TNF-α.^32,33 TNF-α is a pleiotropic cytokine that can promote cell survival or death by binding to TNF-α receptor 1 (TNFR1), which activates MAPK and NF-κB signaling pathways.³⁴ TNF-α- induced cell death occurs when protein complexes called cell death checkpoints form to activate caspases.³⁴ In pathogenic infection, recruits of immune cells by epithelial cell death can eliminate the pathogen, therefore, cell death can be an effective host defense mechanism against pathogenic infection.³⁴ Healthy human colon epithelia, as well as Caco-2 cells, express very low levels of TLR4, mostly concentrated in the crypt epithelium (stem cells) and mononuclear cells in the lamina propria.^35–37 This explains the non-responsiveness of Caco-2 cells to LPS treatment in this study. In contrast, human keratinocytes constitutively express TLR4 and CD14.^38,39 Therefore, it is likely that disruption of the intestinal epithelium by DSS leads to LPS infiltration to the skin compartment and induces pro-inflammatory responses such as reduced cell viability. Our result aligns with skin manifestations in patients with inflammatory bowel diseases where gut barrier functions are impaired.⁴⁰

LGG coculture mitigated the decrease in TEER values, reduced epithelial permeability, and preserved tight junction integrity in DSS-treated samples, which led to improved skin cell viability. Oral supplements of Lactobacilli have been shown to improve various skin conditions,^41,42 by suppressing inflammation, modulating immune systems, or enhancing gut barrier functions.^42,43 The LGG conditioned medium, but not denatured conditioned medium, LGG DNA, or LGG cell wall, protected the human intestinal epithelial enteroids from interferon-γ induced tight junction protein expressions, suggesting that LGG may secrete proteins that improve gut barrier functions.⁴⁴ Our results support the protective effect of LGG on the skin cells by enhancing gut barrier functions. This also indicates that our model has great potential for investigating the role of gut microbiota and evaluating probiotic interventions in dermatologic health.

Our study demonstrates that the GMS chip can simulate and investigate the gut–skin axis in a simple and controlled way. However, while the Caco-2 cell line allows for reproducibility, it may not fully capture the complexity of primary human intestinal tissues. Additionally, the current model does not incorporate immune components, which play a crucial role in both gut and skin homeostasis. Future iterations of the model could address these limitations by incorporating primary cells and immune components to further enhance its physiological relevance.

Conclusions

We demonstrated that the GMS chip is a useful experimental platform to recapitulate the gut–skin axis in the presence of gut microbiota. Using the GMS chip with the fluidic connection between the gut and the skin compartments, we show that disruption of intestinal barrier integrity significantly influences the skin cells. Pretreatment with the probiotic strain LGG conferred protective effects on the skin cells by enhancing the intestinal barrier function. The GMS chip system provides a novel platform for investigating the intricate relationships between gut microbiota, intestinal barrier function, and skin health. By enabling the study of these interactions in a controlled, physiologically relevant environment, this model can accelerate our understanding of the gut–skin axis and facilitate the development of innovative therapies for various gut and skin disorders.

Data availability

The data for this article are available upon reasonable request from the corresponding author.

Author contributions

J. H. S. conceptualized and administrated the project. B. K. and J. S. carried out the experiments and data analyses. J. H. S., R. K., and J. I. W. supervised the project. B. K. and R. K. wrote the original draft. R. K. and J. H. S. revised and edited the draft. J. H. S., R. K., and J. I. W. secured funding.

Conflicts of interest

The authors have no conflicts of interest to declare.

Acknowledgements

This work is supported by the National Research Foundation of Korea (NRF) (RS-2022-NR071880) funded by the Ministry of Science and ICT (MSIT) of the government of the Republic of Korea. This work was supported in part by Basic Research Lab (2022R1A4A2000748), Bio & Medical Technology Development Program (2022M3A9B6018217) by National Research Foundation of Korea, Alchemist Project of the Korea Evaluation Institute of Industrial Technology (KEIT 20018560, NTIS 2410005252) by the Ministry of Trade, Industry & Energy (MOTIE), Republic of Korea, Korea Institute of Marine Science & Technology Promotion (KIMST) funded by the Ministry of Oceans and Fisheries, Korea (RS-2024-00402200). This work was supported by Hongik University Research Fund and 2024 Hongik University Innovation Support Program Fund.

References

1. J. G. LeBlanc, C. Milani, G. S. de Giori, F. Sesma, D. van Sinderen and M. Ventura, Curr. Opin. Biotechnol., 2013, 24, 160–168.
2. G. den Besten, K. van Eunen, A. K. Groen, K. Venema, D.-J. Reijngoud and B. M. Bakker, J. Lipid Res., 2013, 54, 2325–2340.
3. Y. Kinashi and K. Hase, Front. Immunol., 2021, 12, 673708.
4. I. Salem, A. Ramser, N. Isham and M. A. Ghannoum, Front. Microbiol., 2018, 9, 1459.
5. A. Malaviya, K. V. Krishna, S. Malviya and T. Nimisha Das, in Probiotic Research in Therapeutics, Probiotics and Gut Skin Axis–Inside Out and Outside In, ed. K. Beri, P. K. Deol and S. K. Sandhu, Springer Singapore, Singapore, 2022, vol. 3, pp. 1–26,  DOI:10.1007/978-981-16-5628-6_1.
6. R. Iizuka, K. Kawakami and K. Chiba, Microb. Ecol. Health Dis., 2009, 21, 221–227.
7. V. Navarro-López, A. Ramírez-Boscá, D. Ramón-Vidal, B. Ruzafa-Costas, S. Genovés-Martínez, E. Chenoll-Cuadros, M. Carrión-Gutiérrez, J. Horga de la Parte, D. Prieto-Merino and F. M. Codoñer-Cortés, JAMA Dermatol., 2018, 154, 37–43.
8. J. H. Sung, Expert Opin. Drug Metab. Toxicol., 2021, 17, 969–986.
9. R. Kim and J. H. Sung, Adv. Healthcare Mater., 2024, 13, 2302777.
10. K. Ronaldson-Bouchard, D. Teles, K. Yeager, D. N. Tavakol, Y. Zhao, A. Chramiec, S. Tagore, M. Summers, S. Stylianos, M. Tamargo, B. M. Lee, S. P. Halligan, E. H. Abaci, Z. Guo, J. Jacków, A. Pappalardo, J. Shih, R. K. Soni, S. Sonar, C. German, A. M. Christiano, A. Califano, K. K. Hirschi, C. S. Chen, A. Przekwas and G. Vunjak-Novakovic, Nat. Biomed. Eng., 2022, 6, 351–371.
11. R. Novak, M. Ingram, S. Marquez, D. Das, A. Delahanty, A. Herland, B. M. Maoz, S. S. F. Jeanty, M. R. Somayaji, M. Burt, E. Calamari, A. Chalkiadaki, A. Cho, Y. Choe, D. B. Chou, M. Cronce, S. Dauth, T. Divic, J. Fernandez-Alcon, T. Ferrante, J. Ferrier, E. A. FitzGerald, R. Fleming, S. Jalili-Firoozinezhad, T. Grevesse, J. A. Goss, T. Hamkins-Indik, O. Henry, C. Hinojosa, T. Huffstater, K.-J. Jang, V. Kujala, L. Leng, R. Mannix, Y. Milton, J. Nawroth, B. A. Nestor, C. F. Ng, B. O'Connor, T.-E. Park, H. Sanchez, J. Sliz, A. Sontheimer-Phelps, B. Swenor, G. Thompson, G. J. Touloumes, Z. Tranchemontagne, N. Wen, M. Yadid, A. Bahinski, G. A. Hamilton, D. Levner, O. Levy, A. Przekwas, R. Prantil-Baun, K. K. Parker and D. E. Ingber, Nat. Biomed. Eng., 2020, 4, 407–420.
12. M.-H. Kim, D. Kim and J. H. Sung, J. Ind. Eng. Chem., 2021, 101, 126–134.
13. J.-w. Jeon, S. H. Lee, D. Kim and J. H. Sung, Biotechnol. Prog., 2021, 37, e3121.
14. H. R. Lee and J. H. Sung, Biotechnol. Bioeng., 2022, 119, 2590–2601.
15. N. E. Zeitouni, S. Chotikatum, M. von Köckritz-Blickwede and H. Y. Naim, Mol. Cell. Pediatr., 2016, 3, 14.
16. P. B. Eckburg, E. M. Bik, C. N. Bernstein, E. Purdom, L. Dethlefsen, M. Sargent, S. R. Gill, K. E. Nelson and D. A. Relman, Science, 2005, 308, 1635–1638.
17. Z. Lu and J. A. Imlay, Nat. Rev. Microbiol., 2021, 19, 774–785.
18. P. Shah, J. V. Fritz, E. Glaab, M. S. Desai, K. Greenhalgh, A. Frachet, M. Niegowska, M. Estes, C. Jäger, C. Seguin-Devaux, F. Zenhausern and P. Wilmes, Nat. Commun., 2016, 7, 11535.
19. S. Jalili-Firoozinezhad, F. S. Gazzaniga, E. L. Calamari, D. M. Camacho, C. W. Fadel, A. Bein, B. Swenor, B. Nestor, M. J. Cronce, A. Tovaglieri, O. Levy, K. E. Gregory, D. T. Breault, J. M. S. Cabral, D. L. Kasper, R. Novak and D. E. Ingber, Nat. Biomed. Eng., 2019, 3, 520–531.
20. W. Shin, A. Wu, M. W. Massidda, C. Foster, N. Thomas, D.-W. Lee, H. Koh, Y. Ju, J. Kim and H. J. Kim, Front. Bioeng. Biotechnol., 2019, 7, 13.
21. J. Zhang, Y.-J. Huang, J. Y. Yoon, J. Kemmitt, C. Wright, K. Schneider, P. Sphabmixay, V. Hernandez-Gordillo, S. J. Holcomb, B. Bhushan, G. Rohatgi, K. Benton, D. Carpenter, J. C. Kester, G. Eng, D. T. Breault, O. Yilmaz, M. Taketani, C. A. Voigt, R. L. Carrier, D. L. Trumper and L. G. Griffith, Med, 2021, 2, 74–98.e79.
22. R. Kim, P. J. Attayek, Y. Wang, K. L. Furtado, R. Tamayo, C. E. Sims and N. L. Allbritton, Biofabrication, 2020, 12, 015006.
23. L. A. Zenewicz, Front. Immunol., 2017, 8, 324.
24. L. M. JanssenDuijghuijsen, S. Grefte, V. C. J. de Boer, L. Zeper, D. A. M. van Dartel, I. van der Stelt, M. Bekkenkamp-Grovenstein, K. van Norren, H. J. Wichers and J. Keijer, Front. Physiol., 2017, 8, 794.
25. A. Muniandy, P. Benyathiar, F. Ozadali and D. K. Mishra, J. Phys.: Conf. Ser., 2023, 2444, 012014.
26. C. A. Schneider, W. S. Rasband and K. W. Eliceiri, Nat. Methods, 2012, 9, 671–675.
27. R. Sender, S. Fuchs and R. Milo, PLoS Biol., 2016, 14, e1002533.
28. C. Yang and D. Merlin, Inflamm. Bowel Dis., 2024, 30, 844–853.
29. Y. Araki, H. Sugihara and T. Hattori, Oncol. Rep., 2006, 16, 1357–1362.
30. X. Qiu, X. Han, X. Zhang, L. A. Teng, M. K. Sriwastva, L. Zhen, Z. Li, M. Liu, Y. Ren and S. Wang, J. Gastroenterol. Hepatol., 2023, 38, 1158–1169.
31. G. Malatesha, N. K. Singh, A. Bharija, B. Rehani and A. Goel, Emerg. Med. J., 2007, 24, 569.
32. Y.-C. Lu, W.-C. Yeh and P. S. Ohashi, Cytokine+, 2008, 42, 145–151.
33. A. Ciesielska, M. Matyjek and K. Kwiatkowska, Cell. Mol. Life Sci., 2021, 78, 1233–1261.
34. G. van Loo and M. J. M. Bertrand, Nat. Rev. Immunol., 2023, 23, 289–303.
35. E. Cario and D. K. Podolsky, Infect. Immun., 2000, 68, 7010–7017.
36. E. Furrie, S. Macfarlane, G. Thomson, G. T. Macfarlane and T. T. Microbiology, Gut Biology Group and B. Tumour, Immunology, 2005, 115, 565–574.
37. U. Böcker, O. Yezerskyy, P. Feick, T. Manigold, A. Panja, U. Kalina, F. Herweck, S. Rossol and M. V. Singer, Int. J. Colorectal Dis., 2003, 18, 25–32.
38. A. Pivarcsi, L. Bodai, B. Réthi, A. Kenderessy-Szabó, A. Koreck, M. Széll, Z. Beer, Z. Bata-Csörgoő, M. Magócsi, É. Rajnavölgyi, A. Dobozy and L. Kemény, Int. Immunol., 2003, 15, 721–730.
39. P. I. Song, N. Neparidze, C. A. Armstrong, J. C. Ansel, Y.-M. Park, T. Abraham, B. Harten and A. Zivony, J. Invest. Dermatol., 2002, 119, 424–432.
40. S. Larsen, K. Bendtzen and O. H. Nielsen, Ann. Med., 2010, 42, 97–114.
41. L. Drago, M. Toscano, E. De Vecchi, S. Piconi and E. Iemoli, J. Clin. Gastroenterol., 2012, 46, S56–S63.
42. A. Xie, A. Chen, Y. Chen, Z. Luo, S. Jiang, D. Chen and R. Yu, Front. Cell. Infect. Microbiol., 2023, 13, 1137275.
43. W. P. Bowe and A. C. Logan, Gut Pathog., 2011, 3, 1.
44. X. Han, A. Lee, S. Huang, J. Gao, J. R. Spence and C. Owyang, Gut Microbes, 2019, 10, 59–76.

---

Footnote

† Electronic supplementary information (ESI) available. See DOI: https://doi.org/10.1039/d4lc01010h

---