Establishment of physiologically relevant oxygen gradients in microfluidic organ chips

In vitro models of human organs must accurately reconstitute oxygen concentrations and gradients that are observed in vivo to mimic gene expression, metabolism, and host-microbiome interactions. Here we describe a simple strategy to achieve physiologically relevant oxygen tension in a two-channel human small intestine-on-a-chip (Intestine Chip) lined with primary human duodenal epithelium and intestinal microvascular endothelium in parallel channels separated by a porous membrane while both channels are perfused with oxygenated medium. This strategy was developed using computer simulations that predicted lowering the oxygen permeability of poly-dimethlysiloxane (PDMS) chips in specified locations using a gas impermeable film will allow the cells to naturally decrease the oxygen concentration through aerobic respiration and reach steady-state oxygen levels < 36 mm Hg (< 5%) within the epithelial lumen. The approach was experimentally confirmed using chips with embedded oxygen sensors that maintained this stable oxygen gradient. Furthermore, Intestine Chips cultured with this approach supported formation of a villus epithelium interfaced with a continuous endothelium and maintained intestinal barrier integrity for 72 h. This strategy recapitulates in vivo functionality in an efficient, inexpensive, and scalable format that improves the robustness and translatability of Organ Chip technology for studies on microbiome as well as oxygen sensitivity.


Introduction
Many living organs, such as the intestine, experience steep oxygen gradients that play a central role in health and disease, and this is particularly relevant for healthy host-microbiome interactions. The low-oxygen environment in the intestinal lumen and presence of a range of oxygen concentrations allows communities of complex anaerobic and aerobic microbes to thrive; however, the underlying human tissue requires higher levels of oxygen to survive. Most of what we know about the role of microbiome in medicine is based on genomic or metagenomic studies because it is extremely difficult to co-culture commensal bacteria in direct contact with human cells. In fact, oxygen is required for nearly every metabolic process in the human body, but vascular transport, molecular diffusion, and metabolism produce complex oxygen gradients in different organs that enable stable co-existence of complex communities of commensal microbes with living tissues. Recently, microfluidic organ-on-a-chip (Organ Chip) culture devices have been used to establish oxygen gradients that enable stable co-culture of human intestinal epithelium with a complex gut microbiome for days in vitro. 1,2 However, these cultures required development of a specially fabricated hypoxia chamber and use of pumps and tubing that are not available in most laboratories, and this approach cannot be used with many Organ Chip operating systems. Thus, here we set out to explore whether a simpler approach can be developed to establish oxygen gradients within both lab fabricated and commercially available microfluidic Organ Chips in vitro.
The lumen or parenchyma of different organs operate within a characteristic oxygen concentration range. For example, the partial pressure of oxygen (pO2) ranges from 34-36 mm Hg (~5%) in the lumen of the small intestine 3 and 3-11 mm Hg (0.4-1.5%) in the lumen of the large intestine 4 to 34-54 mm Hg (4.5-7.1%) in the marrow compartment of bone. 5 In all of these organs, nearby arterial and venous vessels experience higher oxygen concentrations of 70-100 mm Hg (9.2-13%) and 40 mm Hg (5.3%), 6 respectively. Conversely, deviations from normal oxygen concentrations are characteristic of many disease pathologies, including diabetes, cancer, and ischemia. 7 Thus, maintaining oxygen homeostasis is critical to maintain viability of different tissues and cells, as well as the commensal microbiome, which each have their own distinct oxygen requirements. Given the importance of oxygen status in the human body, it is therefore critical that in vitro organ and tissue models be cultured at physiologically relevant oxygen concentrations to ensure they optimally mimic in vivo functions.
Microfluidic culture systems can be precisely engineered to generate oxygen gradients through careful control over oxygen diffusion, convection, consumption, and generation.
Perfused gas mixtures are commonly used to recreate physiologically relevant oxygen gradients in microfluidic culture systems. For example, we previously generated a hypoxia gradient in Intestine Chips, which contain two parallel microchannels separated by a porous membrane lined by intestinal epithelial and endothelial cells, by perfusing the lumen of the vascular channel with medium pre-exposed to a premixed gas containing 5% oxygen and culturing the chips in an anaerobic, nitrogen gas-filled atmosphere to which the epithelial channel had access via the gas permeable poly-dimethylsiloxane (PDMS) walls of the chip. 1,8,9 A hypoxia gradient was also established in a two-channel Gut Chip lined with Caco-2 intestinal epithelial cells by increasing the thickness of the PDMS block above the cell-lined microchannel and perfusing anoxic medium through it while flowing oxygenated medium through a parallel channel separated from the first by a porous membrane. 10 Continuous perfusion of anoxic and oxygenated media through adjacent culture chambers similarly has been used to generate oxygen gradients in the HuMiX Intestine Chip and GuMI platform. 2,11 Although these methods recreated physiologically relevant hypoxia gradients, they are resource-intensive and prone to failure because gas tanks must be continuously monitored and replaced, gas connections often leak, the systems take hours to equilibrate, and the oxygen levels are difficult to control throughout the duration of the study. Additionally, routine laboratory procedures that are required for the experiment, such as changing media or moving the cultures to be viewed on a microscope, equilibrate the chips to aerobic conditions within minutes after removal from flow, which introduces variation into the experiments and greatly complicates the interpretation of results.
Alternative approaches that use aerobic respiration of cells cultured inside in vitro microsystems have the advantage that they do not require use of perfused gas mixtures. For example, a steady-state oxygen gradient can be established along the fluidic stream of a perfusion bioreactor when cells upstream in the bioreactor deplete oxygen for the cells downstream. 12,13 However, this method generates an axial oxygen concentration gradient along the length of the fluidic channel, which does not permit analysis of key cell and tissue level functional parameters and biomarkers (e.g., epithelial barrier function, gene expression, cytokine release) because the cells along the fluidic stream are not maintained in a constant environment. Hypoxia gradients also have been established perpendicular to a cell layer in Transwell culture models by inserting an oxygen-impermeable plug over the apical side of the membrane and leaving the basal side open to atmospheric oxygen. 14,15 But, these studies are limited because Transwells do not experience dynamic fluid flow or recapitulate physiological cell differentiation and tissue functions with the fidelity observed in Organ Chips. [16][17][18] Here, we describe a simple strategy to achieve physiologically relevant oxygen tension in two-channel, microfluidic, Organ Chips by lowering the oxygen permeability in specific locations of the chip and allowing the cultured cells to further decrease the oxygen concentration through aerobic respiration. Our strategy generates a consistent oxygen gradient throughout the length of the chip when cultured in a conventional aerobic cell culture incubator and does not require anaerobic glove boxes or premixed gases. Furthermore, the oxygen levels remain stable when the chips are removed from the incubator. We used computational simulations of oxygen transport to determine that modifying commercially available PDMS Organ Chips by application of a polyvinylidene dichloride (PVDC) film along most of the surface of the chip allows the cells to naturally deplete oxygen in the microengineered lumen. We experimentally confirmed our approach using Intestine Chips with embedded oxygen sensors that quantify local oxygen concentrations. We verified the feasibility of this strategy for maintaining a hypoxic epithelial lumen by demonstrating no loss in intestinal barrier function over 72 h and maintenance of a viable villus epithelium interfaced with a continuous endothelium.

Simulating aerobic Intestine Chips
We previously described a two-channel, microfluidic, human Intestine Chip containing epithelial cells isolated from patient-derived organoids interfaced with human intestinal microvascular endothelium that recapitulates many anatomical and functional features of the in vivo small intestine (Fig. 1a-c). 16 To explore potential methods to generate a stable oxygen gradient in this microfluidic Organ Chip, we first carried out computational simulations of the oxygen distribution. The Organ Chip we used for modeling is a commercially available (Emulate Inc.) PDMS microfluidic culture device (37.2 mm long x 16.2 mm wide) that contains parallel fluidic channels (15.46 mm long x 1 mm wide) separated by a porous PDMS membrane (50 µm thick; 7 µm pores). A rendering of the PDMS Organ Chip was imported into the simulation and the inlet and outlet regions were removed to increase computational efficiency (Fig. 1b). The pores in the membrane were also removed because the diffusion coefficient of oxygen in medium and PDMS are approximately the same (3 × 10 −5 2 and 5 × 10 −5 2 , respectively) 10 , and removing the pores significantly decreased the geometric complexity of the model. The height of the epithelium was determined from morphometric analysis of crosssections of Intestine Chips lined with primary human intestinal epithelium isolated from patientderived organoids, as previously described. 1,19 The endothelium was incorporated into the fluidic domain of the basal channel and not modeled as a separate geometry because it is very thin (<10 µm) compared to the rest of the chip. The simulation of oxygen distribution incorporated convection, diffusion, and the rate of endothelial and epithelial oxygen consumption. Airsaturated medium was flowed through the apical and basal channel at 60 µL/h, and the temperature, atmospheric pressure, diffusivity of oxygen in medium, and diffusivity of oxygen in PDMS, were set constant. The steady-state simulation of the aerobic Intestine Chip reveals that epithelial oxygen consumption generates a sharp oxygen gradient in both the apical and basal channel (Fig. 1d).
The oxygen concentration rapidly increases across the height of the apical channel because the flux of oxygen through the PDMS exceeds the rate of epithelial oxygen consumption. To be relevant for the small intestine, the apical channel should sustain an oxygen concentration <35 mm Hg (<5%) above the epithelium. 3 Therefore, additional modifications are required in order to achieve a physiologically relevant oxygen gradient in the Intestine Chip.

Establishing Relevant Oxygen Gradients
As oxygen rapidly permeates through PDMS, we hypothesized that reducing the oxygen flux into the chip would allow the epithelium to naturally lower the oxygen concentration through aerobic respiration. We simulated coating the chip with a material that has low oxygen permeability (Pm), while leaving the bottom of the basal channel open to atmospheric oxygen to generate a hypoxia gradient between the fluidic channels and lower the oxygen concentration above the epithelium in the upper channel. Specifically, we modeled coating the top surface of the chip with polyvinylidene chloride (PVDC) film because it is a flexible and self-adhering polymer film that has one of the lowest oxygen permeabilities at 100% humidity and 37 °C (Fig.   2a). 20,21 Additionally, PVDC is transparent to allow for microscopy studies and is commercially manufactured as thin film (<0.025 mm) that is impenetrable to bacteria and mold, 22 making it well-suited for use with Organ Chips. The steady-state simulation shows that coating the chip with PVDC lowers the oxygen concentration and that the chip sustains PO2<35 mm Hg (<~5% O2) within a 250 µm-tall region above the epithelium (Fig. 2b). Higher oxygen concentration is  We then simulated coating the chip with a polyethylene terephthalate (PET) film because PET has similar material properties to PVDC that also make it compatible with Organ Chip culture, however, PET is slightly more oxygen-permeable than PVDC. The steady-state simulation of the PET-coated chip shows that the oxygen concentration above the epithelium more rapidly increases than in the PVDC coated chip, and that the overall oxygen concentration in the PET coated chip is higher (Suppl. Fig. S1). Therefore, the oxygen permeability coefficient of the coating material affects the oxygen concentration in the apical channel and the steepness of the gradient between the apical and basal channel. Materials with lower oxygen permeability coefficients will therefore establish a steeper gradient with lower oxygen concentrations in the apical channel. The oxygen concentration can be tuned by selecting materials with different oxygen permeabilities and the coating material must be carefully selected to meet the oxygen requirements of the in vitro model. Together, the simulations indicate that a PVDC coating is the optimal choice for recapitulating the oxygen microenvironment of the intestine.
We also considered the impact of applying cyclic strain on the oxygen gradient, which is made possible in the Organ Chips we used by application of cyclic suction to hollow chambers that are located on either side of the cell-containing central channels (Fig. 1b). However, application of cyclic strain also requires that the vacuum channels parallel to the fluidic channels are exposed to the aerobic incubator atmosphere. Continuously stretching the membrane at 10% strain and 0.15 Hz lowers the PO2 in the vacuum channels by approximately 28% (PO2~ 100 mm Hg, 13%). The PO2 under cyclic strain was applied to all sides of the vacuum channel and the oxygen concentration was simulated at steady state (Suppl. Fig. S2). The simulation shows that cyclic strain eliminates the hypoxia gradient and significantly increases the oxygen concentration because the rate of oxygen entry into the fluidic channels from the vacuum channel exceeds the rate of aerobic respiration. Thus, using this method, a hypoxia gradient can only be generated under static conditions in these chips, however, this limitation could be overcome in the future by dip-coating the vacuum channels in PVDC, which would reduce the rate ambient air enters into the fluidic channels. 23

Experimentally Validating the Oxygen Gradients
To experimentally validate the anaerobic method defined by the computational simulations, we fabricated Organ Chips with embedded oxygen sensors that read out the in-situ oxygen concentrations at four distinct locations along the fluidic channels. The sensors are located at the top and the bottom of the apical and basal channels, and at two locations along their length (Fig. 3a). These sensor chips were not fabricated with vacuum channels because they interfered with the placement of the sensor spots. However, as previously described, the epithelium still undergoes spontaneous differentiation and expresses many in vivo features of the human small intestine under continuous fluidic flow without cyclic strain. 17 We introduced human epithelium derived from intestinal organoids into the oxygen-sensing chips and cultured the chips for 19 days under continuous medium perfusion through both channels (60 µL/h) Fig. S3). The sensor chips were cultured with epithelium only because the endothelium has a minimal impact on the oxygen distribution in the Intestine Chip (Suppl. Fig. S4). We initiated anaerobic culture by wrapping the chips with PVDC film and introducing holes through this film along the bottom of the basal channel (Suppl. Fig. S6). The chips were placed in an aerobic cell culture incubator at 37°C and 5% CO2, and medium was flowed through both channels (60 µL/h) for 24 h. The oxygen concentration was measured from each sensor spot immediately after removal from flow (Fig. 3b). oxygen gradient that decreases from the top of the apical channel to the epithelium (Fig. 3c).
Plotting the simulated oxygen concentration at different distances away from the epithelium shows that the chip maintains oxygen < 35 mm Hg 0-250 µm above the epithelium, which is physiologically relevant for applications, such as bacterial co-culture and studies on oxygen sensitivity. This system establishes a hypoxia gradient across the epithelium and endothelium, which allows oxygenation of the endothelium while simultaneously providing an anaerobic environment above the epithelium for culture of both anaerobic and aerobic microbes.
This anaerobic strategy is inexpensive and uses laboratory equipment already inside tissue culture labs. Methods that use premixed gases require a large initial investment in capital equipment, demand more materials, and are difficult to scale up. These approaches have a large risk of experimental loss because they require continuously monitoring gas flows and connections, which can be prone to failure. Coating the Intestine Chip with a commercially produced film is a simple and scalable procedure that widens translatability of Organ Chip technology towards more oxygen-sensitive applications. Moreover, this approach is not limited to in vitro models of the intestine, as other systems such as the female genital tract, bone marrow, and colon experience in vivo oxygen gradients. Additionally, a variety of pathologies are associated with decreased oxygen levels or induced oxygen gradients, including inflammatory bowel syndrome 24 , cancer 25 , and reproductive diseases 26 , and thus, in vitro models that recapitulate the oxygen microenvironment should be able to more accurately represent diseased as well as healthy phenotypes.
Chips cultured with premixed gases experience wide fluctuations in oxygen concentration and take a long time to equilibrate back to steady-state. The ability to maintain steady oxygen levels outside of the incubator and without fluidic flow is important to carry out standard laboratory procedures (e.g., changing medium, microscopy imaging), and it enables experiments that demand extended time at room temperature. We removed the oxygen-sensing Intestine Chips from the incubator and measured a gradual increase in oxygen concentration over the course of 1 h (Suppl. Fig. S5a). The oxygen level increases because the rate of aerobic respiration in the intestine epithelium decreases by approximately 70% from 37°C to room temperature (22°C) 27 . We performed an additional computational simulation to predict the effect of removing the chips from flow and bringing them to room temperature, which incorporated this 70% reduction in aerobic respiration. The computational simulation predicted the experimental data with at least 90% accuracy at each time point, thus providing further validation of this simulation tool for assessing oxygen distributions on-chip (Suppl. Fig. S5a).
The oxygen concentration sharply increases from the epithelial layer towards the top of the apical channel because the cells consume oxygen at a lower rate at room temperature (Suppl. Our strategy for establishing a hypoxia gradient on-chip is more physiologically relevant than using premixed gases because aerobic respiration, diffusion, and convection, are the same mechanisms that generate hypoxia gradients in vivo. In the intestine, countercurrent blood flow in the villi and aerobic respiration by the epithelium and commensal microbiome together reduce PO2 in the lumen 31,32 . The flux of oxygen entering the lumen is offset by diffusive mixing, convective transport, and consumption of oxygen which contribute towards lowering the PO2.
Similarly, our anaerobic Intestine Chip maintains low and stable oxygen levels by balancing

Maintenance of Intestinal Barrier Integrity
We then exposed Intestine Chips containing human small intestine epithelium and underlying vascular endothelium to the hypoxia gradient. Differential interference contrast (DIC) and immunofluorescence microscopic analysis confirmed that Intestine Chips cultured under hypoxia for 72 h maintained the villus epithelium interfaced with a continuous endothelium (Fig.   4a, 4b). We also evaluated the effect of hypoxia on intestinal barrier function by measuring changes in the apparent permeability coefficient (Papp) and observed maintenance of barrier integrity for 72 h (Fig. 4c). Taken together, these data confirm that hypoxic Intestine Chips with primary human intestinal epithelium in contact with human endothelium maintain morphology and structural integrity.

Conclusion
We developed a simple strategy to achieve physiologically relevant oxygen tensions in human Intestine Chips by lowering the oxygen permeability in specific locations of the chip and allowing the cultured cells to decrease the oxygen concentration through aerobic respiration.
Our approach is physiologically relevant because it uses respiration, convection, and diffusion to establish hypoxia, which are the same processes that drive the formation of oxygen gradients in vivo. This strategy is also scalable, cost-efficient, and easy to implement into standard laboratory workflows because an oxygen gradient can be generated in a conventional aerobic incubator and the chips can be removed from the incubator without significantly disrupting the oxygen distribution. This simplified approach for generating a hypoxia gradient in microfluidic Organ Chips enables the development of in vitro models that require a controllable oxygen environment, which could accelerate the discovery of therapeutic strategies for a broader range of human pathologies. Tables S1 and S2, respectively.

Cell Culture
Human small intestine (duodenal) epithelium was obtained and isolated from de-identified

Anaerobic Chip Assembly and Culture
The

Oxygen Sensor Chip Fabrication
The chip consists of 5 layers: an apical block, apical gasket, porous membrane, basal gasket, and basal block (Fig. S7). The membrane was produced by casting PDMS preset polymer (10:1 ratio, Sylgard 184) onto a photolithographically prepared master that has an array of 50 µm-tall posts and cured at 60 °C for 1 h in a drying oven. The apical block was made by casting PDMS

Oxygen Measurements
The oxygen concentration was quantified in the oxygen sensing Intestine Chips using an optical fiber (SPFIB-BARE-CL2, PyroScience GmbH) connected to a FireSting-PRO oxygen meter (FSPRO-4, PyroScience GmbH) with an external temperature probe (Pt100, PyroScience GmbH). A manual background correction was performed on an Intestine Chip that did not contain oxygen sensor spots. A 2-point calibration was carried out with 0% and 100% oxygen standard solutions, where the 0% standard was obtained by degassing a 1 mg/mL slurry of oxygen-sensing nanoparticles (OXNANO, PyroScience GmbH) in water with argon. A background correction and 2-point calibration was performed before acquiring data from each time point and when switching between apical and basal measurements. (1)

Paracellular Permeability Measurements
where is the volume of the receiving channel (basal channel effluent), is the volume of the dosing channel (apical channel), is the concentration of tracer in the receiving channel, is the concentration of tracer in the dosing channel outlet (apical effluent), is the area of the membrane (cm 2 ), and is time (sec).

Morphological Studies
Intestine Chips were fixed with 4% paraformaldehyde (

Statistical Analysis
Tests for statistically significant differences between groups were performed using a twotailed t-test. Differences were considered significant when the P value was less than 0.05 (**, P<0.01; ***, P<0.001). All experimental results are expressed as means ± standard deviation (SD). Each experiment was conducted with a sample size of n=3 Intestine Chips per condition.