Adipose-on-a-chip: a dynamic microphysiological in vitro model of the human adipose for immune-metabolic analysis in type II diabetes

Yunxiao Liu a, Patthara Kongsuphol a, Su Yin Chiam a, Qing Xin Zhang a, Sajay Bhuvanendran Nair Gourikutty a, Shilpi Saha b, Subhra Kumar Biswas b and Qasem Ramadan *a
aInstitute of Microelectronic, A* STAR (Agency for Science, Technology and Research), 2, Fusionopolis Way, #08-02, Innovis Tower, 138635 Singapore. E-mail: alramadanq@ime.a-star.edu.sg; Tel: +(65) 67705786
bSingapore Immunology Network, A*STAR, 8A Biomedical Grove, Immunos, 138648 Singapore

Received 9th May 2018 , Accepted 8th December 2018

First published on 12th December 2018


Infiltration of immune cells into adipose tissue is associated with chronic low-grade inflammation in obese individuals. To better understand the crosstalk between immune cells and adipocytes, in vivo-like in vitro models are required. Conventionally transwell culture plates are used for studying the adipocyte–immune cell interaction; however, the static culture nature of this approach falls short of closely recapitulating the physiological environment. Here we present a compartmentalized microfluidic co-culture system which provides a constant-rate of nutrient supply as well as waste removal, resembling the microvascular networks of the in vivo environment. Human adipocytes and U937 cells were co-cultured in close proximity in an enclosed system. The porous barrier between the adjacent compartments comprises an array of microchannels, which enables paracrine interaction between cells in adjacent compartments and improved perfusion-based long term cell feeding. Human pre-adipocytes were fully differentiated into adipocytes on the chip and remained viable for several weeks. Upon co-culturing with immune cells, adipocytes showed a tendency to develop insulin resistance. The immune-metabolic correlation has been studied by monitoring adiponectin and IL-6 expression, as well as glucose uptake upon treatment with insulin. Our microfluidic system can be potentially used to develop physiologically relevant adipose tissue models to study obesity-associated diseases such as insulin resistance and type 2 diabetes and therefore, facilitate drug development to treat these diseases.


Introduction

Obesity, a condition characterized by excess accumulation of adipose tissue, has become a worldwide health concern due to its linkage with various diseases, including insulin resistance, type 2 diabetes, cardiovascular diseases and cancers.1–3 It is generally accepted that obesity is correlated with chronic low-grade inflammation, which is responsible for the onset of chronic diseases.1,4,5 Increased risk of developing colon cancer, as well as a higher incidence of recurrence, has been indicated in obese individuals.6

Adipose tissue is composed of various cell types, including adipocytes, pre-adipocytes, endothelial cells, fibroblasts and different types of immune cells.7 Adipocytes, the main component of adipose tissue, not only function as lipid storage compartments but also produce various endocrine proteins to regulate the whole-body metabolism.8 These proteins, including leptin, adiponectin, and pro-inflammatory mediators (e.g. interleukin (IL)-6, monocyte chemoattractant protein 1 (MCP-1) and tumour necrosis factor (TNF)-α), are collectively referred to as adipokines.9–12 Accumulating evidence indicates that adipokines have a key role in obesity-linked metabolic dysfunction.1

Macrophages are the most abundant immune cell type in adipose tissue. In lean adipose tissue macrophages make up around 5% of the total cell population, whereas the ratio can increase to 50% in obese adipose tissue.13 A key function of macrophages is phagocytosis to remove dead cells. In inflamed adipose tissue, macrophages are selectively localized around dead adipocytes forming a so-called crown-like structure.7 The underlying mechanism of macrophage infiltration is not fully understood. However, it is generally believed that following the onset of obesity, the expansion of adipose tissue leads to dysregulation of adipokine secretion, e.g. upregulated MCP-1 and downregulated adiponectin production,14,15 which consequently induces peripheral monocyte infiltration into the site.15 These monocytes then differentiate into macrophages in the inflamed adipose tissue.16 It has been speculated that the crosstalk between macrophages and adipocytes aggravates the chronic inflammation in obese adipose tissue.17,18

To understand the interaction between immune cells and adipocytes, adipocytes and monocytes/macrophages have been co-cultured in vitro either via direct co-culture (e.g. layer macrophages on top of adipocytes)13 or indirect co-culture in transwell tissue culture plates.19–21 These studies suggested that the crosstalk between adipocytes and immune cells occurred through both paracrine factors and direct cellular contact. However, these systems fall short of closely mimicking the cellular micro-environment due to their static nature and therefore, provide only limited physiological relevance. Microfluidic-based cell culture, on the other hand, provides dynamic culture conditions with constant supply of nutrients as well as removal of waste, which closely resembles the in vivo microvascular networks.22

The validity of an in vitro model is dependent on how well it reproduces the key physiological characteristics of the in vivo system. Organs-on-a-chip technology has great potential to mimic many physiological properties of in vivo organ systems. Inside the body, cells are organized either in direct contact or in close proximity to other cell types in a tightly controlled architecture that regulates tissue function. Homotypic and heterotypic cellular interactions are essential for tissue development, repair, and homeostasis.23 Microfabrication and microfluidics technologies provided tools to create advanced cell culture systems which can be employed to mimic a specific tissue structure and create in vivo-like cellular microenvironments. Simple co-culture systems permit physical direct contact between heterotypic cells24,25 to study the paracrine signalling and response to soluble signalling factors.26,27 However, it is difficult to determine the relative contributions of each cell type to any observed effects of co-culture, particularly, when the effect of paracrine signalling is of interest. To allow better differentiation of the heterotypic cellular signalling physical porous barriers between cell types would permit the exchange of soluble factors, while preventing cell–cell direct contact. Therefore, heterogeneous cell culture can be designed to emulate the structure of a specific tissue/organ within a porous compartmentalized microfluidic structure that enables cells to interact through paracrine/endocrine chemical signals like in vivo.

Microfabricated porous barriers were employed to recapitulate the blood capillary endothelium in various microphysiological in vitro models.28–30 In general, these devices are fabricated in PDMS with the cell culture channels and nutrient feeding channels being positioned parallel to each other, therefore, the dynamic fluidic interaction and mass transfer or chemical diffusion between the cells and nutrients may not recapitulate the real physiological conditions. In addition, despite the simple fabrication process of PDMS that is mainly used in fabrication of cell culture microdevices, major issues such as absorbing small organic compounds, including the precious analytes, and the inability to be mass fabricated may hinder the use of this material in the organ-on-a-chip technology.

Here we present a silicon (Si)-based microfluidic chip which allows realization of organotypic co-culture of adipocytes and monocytes in close proximity in an architecture that simulates the relevant microenvironment in the human body. The organotypic culture is ensured by the cell–cell interfacing structure characterized by porous barriers within the compartmentalized cell culture system with each cell type enclosed within a dedicated compartment. The development process of the microfabricated system which facilitates high throughput chip fabrication and experimentation with minimum chip–chip or experimental variation is discussed. The chip has been characterized and tested by hosting the human adipose in vitro model which was previously established and tested.31Fig. 1a shows the overall chip structure, the fluidic setup and the biological model within it.


image file: c8lc00481a-f1.tif
Fig. 1 (a) Conceptual schematic drawing of the biological model of the adipocyte–immune cell system within a microfluidic cell co-culture device. (b) Fabrication process of the Si chip. (c) Schematic drawing of the chip and the fluidic cartridge with a detailed view of the porous barriers between the three fluidic compartments. (d) An optical image of the chip and fluidic setup with an SEM detailed image of the porous barriers.

Materials and methods

Materials and reagents

Sylgard® 184 elastomer base and curing agent were obtained from Dow Corning. Poly-L-lysine solution (0.1% w/v, in H2O) was purchased from Sigma-Aldrich. Human recombinant insulin (#12585014) was purchased from Life technologies. Human interleukin-6 (IL-6) and adiponectin ELISA reagents, including LEAF™ purified anti-human IL-6 antibody (#501110), human IL-6 recombinant peptides (#570808), biotin anti-human IL-6 antibody (#501110), purified anti-human adiponectin antibody (#528701), human adiponectin protein (#558409), biotin anti-human adiponectin antibody (#528803), HRP streptavidin (#405210), and TMB solution (#421501), were all obtained from Biolegend (USA). Other chemicals were obtained from Sigma-Aldrich unless mentioned otherwise. Krebs-Ringer-Phosphate-HEPES (KRPH) buffer was prepared with 20 mM HEPES, 5 mM KH2PO4, 1 mM MgSO4, 1 mM CaCl2, 136 mM NaCl, and 4.7 mM KCl. The pH value was adjusted to 7.4.

Chip fabrication

The chip was fabricated in a Si substrate using standard lithography and etching techniques (Fig. 1b). A bilayer of silicon oxide and silicon nitride was deposited on the Si substrate with a thickness of 300 Å and 1500 Å (Fig. 1bi), respectively. Using the deep reactive ion etching (DRIE) technique, shallow trenches with a width, length and depth of 5 μm, 50 μm and 5 μm, respectively, were created to form the microchannel array of the porous barriers between the fluidic compartments (Fig. 1bii). Then, wide trenches, with a depth of 200 μm, were also etched using DRIE to form the fluidic compartments as well as the fluidic inlets/outlets (Fig. 1biii). The fluidic inlets and outlets were opened by further backside wet etching such that the fluidic port will be accessed through the back side of the chip (Fig. 1biv). Then the patterned Si wafer was bonded to a glass substrate using anodic bonding to form a fully sealed compartmentalized microfluidic structure with porous barriers. Finally, the Si wafer was diced to individual chips with dimensions of 40 mm × 15 mm. The chip comprises three concentric planar fluidic compartments (A, B and C) which are connected through microchannel arrays realized on top of the separating walls of the fluidic compartments (Fig. 1c and d). The distance between the microchannel array and the surface of the cell culture compartment is 195 μm, therefore, the media flow through these channels ensures mild shear stress onto the cell membranes. This will be shown in the simulation section below.

Perfusion setup

A fluidic connector was fabricated in poly(methyl methacrylate) (PMMA) to accommodate the chip during the cell culture. L-shaped fluidic channels with a diameter of 1 mm were created inside the PMMA block to which the fluidic inlets/outlets of the chip were aligned (Fig. 1c). A PDMS gasket was placed between the chip and the connector surface, and the chip was pressed against the connector using a PMMA frame and screwed to ensure good connection and prevent liquid leakage. The fluidic connector also comprises a vertical holder to hold the media reservoir (Fig. 1d). The chip was connected to 10 mL syringes through PEEK fittings (1/32 in., M-645X, ThermoFisher) and tubing. The syringes were mounted onto a set of high precession syringe pumps (neMESYS, Cetoni, Germany) which were controlled using QmixELEMENTS software (Cetoni, Germany). Finally, the perfusion setup was completed by placing the chip (within the holder) and the media reservoir inside a CO2 incubator and the syringe pumps were placed outside the incubator. The syringe pumps were run in withdrawing mode such that the cell culture media perfused through the chip and the media reservoir were maintained under the same culture conditions inside the CO2 incubator. An optical image of the chip and an SEM image of the inter-compartment porous barriers are shown in the inset (Fig. 1d).

Finite element analysis

Finite element analysis (FEA) was carried out using COMSOL Multiphysics software. A three dimensional model of the chip whose geometry imitated the layout of the chip device was created and the fluidic boundary conditions, materials and physics were applied. The velocity field due to the continuous liquid flow inside the compartments was determined by using the fluid flow module. A laminar flow interface was used to compute the velocity of the fluid by solving the Navier–Stokes equations. Four flow conditions corresponding to the input/output ports were simulated to investigate the flow dynamics through the compartments by selectively applying the fluid flow through individual inlets (Fig. 2a). The velocity field and shear stress profiles were calculated at a fluid (water) flow rate of 8 nL s−1.
image file: c8lc00481a-f2.tif
Fig. 2 FEA of the flow field and shear stresses through the fluidic compartments under different flow conditions. (a) Schematic drawing of the chip indicating the various fluidic ports and the lines along which the shear stress is calculated. (b) The perfusion flow profile across the compartments A and B. (c) Enlarged view focusing on the porous barrier between A and B. (d–g) Velocity profile across the three compartments under different flow injection conditions. (h–k) Calculated shear stress along the lines indicated in a, corresponding to the injection conditions in the middle row.

Adipocyte culture

Prior to cell inoculation, the fluidic compartments were filled with 70% ethanol for at least 4 hours for sterilization. Then the chip was washed with sterilized deionized (DI) water and 100 μL of poly-L-lysine solution (0.1 mg/mL, in H2O) was loaded into the chip and incubated overnight at 37 °C. The chip was then washed with sterilized DI water and phosphate buffered saline (PBS), followed by chip conditioning with pre-adipocyte growth medium (#811-500, Cell Applications). Cryopreserved human pre-adipocytes (#802s-05a) were obtained from Cell Applications, Inc. (USA). Thawed pre-adipocytes were sub-cultured in pre-adipocyte growth medium and passaged until passage 4. For cell seeding into the chip, 50 μL of pre-adipocyte suspension (cell density >1.5 × 106 cells per mL, cell viability >85%) was inoculated into compartment B. Cells were maintained under static culture conditions (without perfusion) for 2 hours to allow cell attachment. Then perfusion was started at a flow rate of 8 nL s−1. The medium was perfused through compartment B and diffused to the other two compartments through the microchannel array. Pre-adipocytes grew to confluence after ∼2–3 days. To induce cell differentiation, the pre-adipocyte growth medium was replaced with adipocyte differentiation medium (#811D-250, Cell Applications). After 12–14 days of differentiation, adipocyte maintenance medium (#811 M-250) was applied for at least 2 days before cell characterization or co-culture with immune cells. During cell culture, cell morphological images were acquired with an upright microscope (Olympus, BX3-CBH, Japan).

Lipid droplet staining

After differentiation, on-chip cultured adipocytes were first fixed with 4% paraformaldehyde (PFA). Oil Red O solution was prepared comprising 40% DI water and 60% Oil Red O stock solution (0.5% in isopropanol, Sigma-Aldrich). Adipocytes were stained with Oil Red O solution for 30 min at room temperature and then washed with DI water. Bright-field images were taken using a microscope. To quantify the differentiation rate which is proportional to the number and size of lipid droplets, the total surface covered with lipid droplets was analysed using cellSens software.

Adipocyte/U937 cell co-culture

U937 mononuclear cells (American Type Culture Collection) were cultured in a 5% CO2 atmosphere in DMEM supplemented with 10% fetal bovine serum and 1% penicillin–streptomycin antibiotics. Cells within passage 20 were used. After adipocyte differentiation, U937 cells were collected and re-suspended in adipocyte maintenance medium. 50 μL of U937 cell suspension was inoculated into compartment A. The adipocyte maintenance medium was perfused at a rate of 8 nL s−1 through compartment B for 2–3 days. The cell morphological change was monitored using a microscope. After co-culture for 2 days, the medium was collected from the perfusion outlet with a lower flow rate of 3 nL s−1 in order to increase the concentration of proteins. The collected medium was stored at −20 °C for further ELISA testing. After co-culture for 3 days, U937 cells were flushed out of the chip with PBS, and the cell number and viability were measured using trypan blue dye with a Luna™ cell counter (Invitrogen, USA). Cell viability is defined as the percentage of viable cells in the whole cell population.

Immune assay

∼100 μL of the supernatant was collected from each chip/experiment for ELISA testing. A lower flow rate of 3 nL s−1 was applied during medium collection for protein thickening purpose. The adiponectin and IL-6 concentrations in the medium were measured using the ELISA assay following the manufacturer's instructions. Briefly, antibody-coated 96 well microplates (Nunc MaxiSorp, #434797) were first prepared via primary antibody incubation. The supernatant solution was then incubated in the antibody-coated microplates. After washing, biotin-labelled detection antibodies were applied, followed by streptavidin–HRP. Finally, a TMB substrate was used for colour development, and the absorbance was measured using an EnSpire multimode plate reader (Perkin Elmer, USA).

Glucose uptake assay

A glucose uptake assay kit, which uses 2-deoxy-2-[(7-nitro-2,1,3-benzoxadiazol-4-yl)amino]-D-glucose (2-NBDG) as a fluorescence-labelled deoxyglucose analog probe, was purchased from Cayman Chemicals (USA). Adipocytes were starved with adipocyte starvation medium (#811S-250, Cell Applications) overnight and then washed with KRPH buffer. Cells were treated with sterilized KRPH buffer with 2% BSA for 1 hour. 50 μL of 2-NBDG (200 μg mL−1) in KRPH buffer (with 2% BSA) was then loaded into the adipocyte culture compartment (B). After 30 min of incubation, chips were washed with PBS, and fluorescence images were taken using an upright microscope with a fluorescein filter (excitation/emission of 485/535 nm). For insulin stimulation, 50 μL of 2-NBDG (200 μg mL−1) together with insulin (8 μg mL−1) in KRPH buffer with 2% BSA was injected into the adipocyte culture compartment. To quantify glucose uptake, cells within the chip were frozen at −80 °C and then were thawed at room temperature. 100 μL of lysis buffer (50 mM Tris, 150 mM NaCl, 1% Triton-X100, pH 8) was injected into compartment B and incubated for 10 min in the dark. The lysis buffer was then collected, and the fluorescence intensity was measured using the plate reader. The lysis buffer was also diluted 100 times in PBS, and the protein concentrations in the lysis buffer were determined by adding a Bradford reagent (Sigma-Aldrich #B6919) and measuring the absorbance at 595 nm. The fluorescence intensity was then normalized to the total protein concentration and expressed as fluorescence intensity (a.u.) per μg mL−1 of protein.

Results and discussion

The microfluidic chip

To mimic the dynamic physiological microenvironment of adipose tissue we have developed a Si-based microfluidic chip which has three main features: concentric cell culture compartments (compartments A, B and C), microchannel arrays between compartments, and media channels accessible to individual compartments (Fig. 1a and c). The microchannel arrays act as a perfusion barrier which mimics the in vivo endothelial barrier. Similar barrier systems have been previously reported,28–30 whose cell loading and perfusion channels were fabricated in PDMS with straight parallel channel structures. Our current device was designed as three concentric compartments that were separated by capillarized thin walls, with the capillary opening positioned at the top of the barrier wall (Fig. 1). This structure significantly improves the cell–cell, cell–ECM and cell–stimuli interactions during the long term culture and reduces the effect of possible high shear stress on the cell membrane and direct lateral flow close to the cell layer surface which may induce cell detachment. It should be noted that our current device was designed as a platform for expanding the number of cell types emulating the key structure and functions of tissue. While only two cell types have been demonstrated in the current study, crucial biological components (e.g. pancreatic beta cells) will be integrated into the model in the near future. Therefore, three different cell types will be hosted in the compartments A, B and C. In addition, the feasibility of in situ capture and detection of the released cytokines using functionalized magnetic beads is investigated, where beads are loaded into the central compartment (C) for dynamic capture of cytokines (Fig. S1). Positioning the functionalized magnetic beads, which can be constantly agitated and trapped using an external magnetic source, offers efficient quantification and higher sensitivity and allows time-resolved analyte detection.

Flow profile in the chip

The flow profile in the three-compartment chip was analysed with FEA. The perfusion profile (flow streamlines) was simulated at different flow injection configurations by injecting the liquid (water) into a selected fluidic inlet or combination of more than one inlet (Fig. 2a). The corresponding shear stresses on the cell membranes were calculated along the YY′ line across compartment B where adipocytes were cultured. Fig. 2b and c show the flow velocity profile across compartments A and B through the porous barrier. The latter ensures laminar flow perfusion and hence fluidic connection between the compartments. Fig. 2(d–g) show uniform velocity profiles across the three compartments thanks to the perfusion channel arrays, while Fig. 2(h–k) show the calculated shear stresses along the YY′ line indicated in Fig. 2a. It was found that different flow injection conditions result in different shear stresses in cell culture compartment B. When the liquid is injected through IB, the resultant average shear stress is ∼3 × 10−3 dynes per cm2 (Fig. 2k) and it is reduced to half of its value when the liquid is injected into the three inlets (Fig. 2h). All the resultant shear stresses in all configurations are within the range of in vivo physiological interstitial shear stress (≤0.1 dynes per cm2).32,33

Pre-adipocyte culture and differentiation on the chip

Primary adipocytes are difficult to handle in terms of harvesting and culturing due to their fragile and buoyant nature. Therefore, pre-adipocytes were used and differentiated into adipocytes in the microfluidic system. Pre-adipocytes are adipogenic precursor cells with the ability to attach to the substrate surface, proliferate and differentiate into adipocytes. To promote cell attachment, the microfluidic chip was coated with poly-L-lysine overnight. On average, ∼1.5 × 106 cells per mL were inoculated into the chip in most experiments. The glass layer permits clear observation of cells inside the chip under an upright microscope. Cells were initially kept under static conditions for at least 2 h to allow cell attachment. Most of the cells were attached to the chip surface 2 h after inoculation. Then the cell culture medium was perfused through compartment B where the pre-adipocytes were, at a flow rate of 8 nL s−1. Pre-adipocytes grew to confluence within 2 days (Fig. 3a). Fig. 3b shows an overview of the chip with pre-adipocytes cultured for 2 days. Some cells were observed to migrate to compartment A. After confluence, the pre-adipocyte growth medium was replaced with the differentiation medium. Small droplets gradually appeared (Fig. S2a), which were confirmed to be lipid droplets by Oil Red O staining (Fig. 4a). Development of lipid droplets, lipogenesis, is the primary indicator for the differentiation of pre-adipocytes to mature adipocytes.34 Oil Red O, a dye that stains lipids, has been used to analyse the differentiation of pre-adipocytes to adipocytes for more than 40 years.35–37 The lipid droplet size and number rapidly increased during differentiation. The total surface area, within the adipocyte compartment, occupied with lipids was calculated with cellSens software as shown in Fig. 4b. Some cells were observed to detach during the differentiation which could be attributed to the excess accumulation of lipids rendering cells buoyant and fragile. Fig. S2b and c show overview images of the chip with pre-adipocytes differentiated for 3 days (a total culture of 5 days) and 10 days (a total culture of 12 days), respectively. With constant nutrient supply and waste removal, adipocytes could be cultured on the chip for several weeks with high cell viability (Fig. S3).
image file: c8lc00481a-f3.tif
Fig. 3 (a) Optical images of the pre-adipocyte culture at different time intervals. (b) An overview image of the chip with pre-adipocytes cultured for 2 days.

image file: c8lc00481a-f4.tif
Fig. 4 (a) Bright-field images of cells stained with Oil Red O after different days of culture (days in growth medium + days in differentiation medium + days in maintenance medium). (b) The rate of change of the total lipid coverage area with the time of culture. Data were obtained from Oil Red O stained images using cellSens software. Lipid area refers to the total area of lipid droplets in one frame shown in (a). (c) Adiponectin and IL-6 production from adipocytes cultured in adipocyte maintenance medium for 3 and 8 days. Supernatants were collected from 3 chips at day 3 and day 8, and each bar represents the mean ± standard deviation (SD). A paired two-tailed Student's t test was performed between different culture days. P < 0.05 was considered significant.

A few days after cell differentiation, the adipocyte differentiation medium was replaced with the adipocyte maintenance medium. A gradual increase in the size of the lipid droplets was observed (Fig. 4b and S4). It is well known that the size of adipocytes dramatically varies with different physiological conditions. For example, the size of adipocytes in healthy humans can increase as high as three folds during energy intake.38 The increase could be even higher in obese individuals. Several studies also found that the increase in adipocyte size leads to the impairment of adipocyte metabolic function which in consequence increases the risk of type 2 diabetes development.39,40 For instance, an inverse relationship has been suggested between the adipocyte size and the level of serum adiponectin,40,41 which is a potent anti-inflammatory hormone, and obesity may be correlated with the development of insulin resistance.42–44

Our perfusion setup permitted supernatant sampling at different time intervals without cell disturbance, hence, measuring the rate of release of various adipokines from the cell culture. Adiponectin and IL-6 were selected and monitored through the experiments. Adiponectin secretion did not change significantly (P = 0.50) with culture days in the maintenance medium (Fig. 4c). However, there was a decreasing trend, which may imply an adverse correlation between the adiponectin secretion level and the adipocyte size.

IL-6, a pro-inflammatory cytokine, is abundantly expressed by adipose tissue. Similar to adiponectin, secretion of IL-6 demonstrated a strong relationship with obesity and insulin resistance.11 However, the role of IL-6 in inflammation is controversial, i.e. it can have a pro- or anti-inflammatory effect.13,45 IL-6 has also been indicated as a potential marker of adipocyte differentiation. For instance, Vicennati et al. reported that IL-6 secretion by adipocytes (in vitro) gradually increased in a time-dependent manner during differentiation.46 Our results showed that the level of IL-6 expression did not change much during the culture period in maintenance medium (Fig. 4c).

Insulin-stimulated glucose uptake in adipocytes is an important process to control the whole-body energy homeostasis.47 We measured the basal and insulin-stimulated glucose uptake in cultured adipocytes using 2-NBDG. Fig. 5a shows the fluorescence images that indicate the uptake of glucose into adipocytes. Due to cell aggregation, the fluorescence signals appear to be strong which makes the uptake difficult to interpret and quantify. Therefore, to quantify the glucose uptake by the adipocytes, cells were lysed in the chip and the fluorescence intensity of the lysis buffer was measured using a microplate reader. As shown in Fig. 5b, insulin-treated adipocytes showed a significantly higher (2.6-fold, P = 0.007) glucose uptake value as compared to the basal glucose uptake.


image file: c8lc00481a-f5.tif
Fig. 5 (a) Bright field (left column) and fluorescence (right column) images of adipocytes taking 2-NBDG without/with insulin stimulation. (b) Glucose uptake measured in adipocyte lysate. The fluorescence intensity of 2-NBDG was measured and normalized to the total protein concentration of the cell lysate solution. The result for each condition was obtained from three independent chips. Each bar represents the mean ± standard deviation (SD). An unpaired two-tailed Student's t test was performed between different culture days. P < 0.05 was considered significant.

Adipocyte/U937 cell co-culture

A key feature of our microfluidic chip is its ability to co-culture, in close proximity, two/three types of cells that are physically separated but fluidically/chemically connected through microchannel arrays. Adipocytes and U937 cells were co-cultured on the chip (Fig. 6a) to emulate the structure of the inflamed human adipose tissue. The U937 cell line was used as a model of the human monocytes, which responds to various stimuli and adopts the morphology and characteristics of macrophages.48–50 For example, with the treatment of phorbol 12-myristate 13-acetate (PMA), U937 cells transformed from non-adherent, round cells to adherent elongated “macrophage-like” cells with phagocytic ability.51,52 U937 cells were inoculated into the chip (compartment A), within which adipocytes were cultured and differentiated in compartment B. U937 cells formed visible clusters after 1 day of co-culture. After 3 days of co-culture, spindle-shaped cells, which were attached to the chip surface, were observed in compartment A (marked by the arrows in Fig. 6a), a morphology that was similar to PMA-stimulated U937 cells (Fig. S5). This may indicate the possible differentiation of U937 cells into “macrophage-like” cells under co-culture conditions. It was observed that the U937 cell viability decreased with the initial cell seeding density (Fig. 6c). Due to the small size of the fluidic compartment, the cell density as well as the media flow rate must be optimized to maintain sufficient nutrient supply and high cell viability.
image file: c8lc00481a-f6.tif
Fig. 6 (a) Adipocyte/U937 cell co-culture on the chip from seeding, day 1 and day 3. The arrows indicate the attached macrophages. (b) An overview image of the cell co-culture with low magnification. (c) U937 cell viability versus cell numbers after 3 days of culture on the chip. The initial cell viability was 75%. (d) Adiponectin and IL-6 secretion in adipocyte mono-culture and 2 days after U937 co-culture. Media were collected from 3 chips, and each bar represents the mean ± standard deviation (SD). A paired two-tailed Student's t test was performed between the mono-culture and co-culture. P < 0.05 was considered significant.

Adipocyte/U937 cell co-culture did not change the production of both adiponectin and IL-6 significantly as compared with the monoculture (Fig. 6d). However, there was a decreasing trend of production for both proteins. A significant decrease of adiponectin and increase of IL-6 have been reported for the co-culture of adipocytes and mature macrophages.19,20,53 The decreasing trend of adiponectin secretion in our work agrees with previous studies. IL-6 production with a decreasing trend upon addition of U937 cells could be attributed to the limited activation of U937 cells. These cells may not phenotypically and functionally resemble the adipose-associated macrophages. It has been reported that macrophages are the main source of IL-6 production when they are in direct contact with adipocytes,13 whereas U937 cells secrete only minute amounts of IL-6 without stimulation.54 The human adipocyte/macrophage co-culture model will be investigated in future studies.

Glucose uptake in adipocytes was also measured after U937 cell co-culture. Fig. 7a shows the bright field and fluorescence images of adipocytes after treatment with 2-NBDG. Insulin-stimulated adipocytes showed a moderate increase of glucose uptake compared to the basal uptake value (1.4-fold, P = 0.21) and compared to adipocytes in the mono-culture (2.6-fold, P = 0.007) (Fig. 7b) indicating the tendency towards insulin resistance in adipocytes when the cells were co-cultured with U937 cells.


image file: c8lc00481a-f7.tif
Fig. 7 (a) Bright field (left column) and fluorescence (right column) images of adipocytes co-cultured with U937 cells after 2-NBDG treatment with/without insulin stimulation. (b) Glucose uptake of co-cultured adipocytes with/without insulin treatment. Adipocytes were lysed and the fluorescence intensity was measured and normalized to the total protein concentration of the cell lysis. The result for each condition was obtained from three independent chips. Each bar represents the mean ± standard deviation (SD). An unpaired two-tailed Student's t test was performed between different culture days. P < 0.05 was considered significant.

It is well understood that adiponectin is an insulin sensitizer.1 The autocrine effect has been observed for adiponectin on adipocytes, e.g. adiponectin increased the insulin's ability to stimulate glucose uptake in adipocytes.55 In contrast, IL-6 was found to decrease insulin-stimulated glucose transport in 3T3-L1 adipocytes.56 In our current model, slight reduction of both adiponectin and IL-6 was observed in the co-culture. To further understand the role of immune cells in insulin resistivity as well as adipokine expression, various adipokines need to be investigated and measured. For instance, TNF-α, whose expression was increased in a 3T3-L1 mouse adipocyte and RAW264 macrophage co-culture model,18 is known to impair insulin receptor signalling and hence contributes to insulin resistance.57

In summary, we have demonstrated the co-culture of adipocytes and U937 cells, as well as characterization in terms of adipokine expression and glucose uptake in a Si-based microfluidic chip. The microfluidic system enables continuous medium collection for cytokine monitoring. To build a more physiologically-relevant human adipose tissue model, peripheral blood mononuclear cells (PBMCs) as a better immune cell source are currently being investigated.

Conclusions

An in vitro microphysiological model of the human diabetic adipose has been demonstrated in a Si-based microfabricated compartmentalized cell culture chip. This model comprises co-culture of adipocytes and immune cells in an enclosed system that provides an in vivo-like environment, i.e. continuous nutrient supply and cell–cell interaction between the two cell types. The microfabricated system allows various on-chip/off-chip characterizations, such as microscopy imaging and the ELISA. Human pre-adipocytes were cultured, differentiated and maintained with high viability in the chip for several weeks. Adipocytes showed a tendency to develop insulin resistance when co-cultured with U937 cells. Our in vitro model would allow study of the immune-metabolic profiling in the human adipose tissue related to insulin resistance and type 2 diabetes. Finally, the compartmentalized microfluidic system provides a versatile tool for in vitro modelling of the complex tissue structure, e.g. including pancreatic β cells in the adjacent compartments as well as utilizing functionalized magnetic beads for in situ capture and quantification of the released adipokines which will be demonstrated in our future work.

Conflicts of interest

There are no conflicts to declare.

Acknowledgements

This work was funded by the Science and Engineering Research Council of Agency for Science, Technology and Research (A*STAR), Singapore under JCO Grant #1431AFG123.

References

  1. N. Ouchi, J. L. Parker, J. J. Lugus and K. Walsh, Nat. Rev. Immunol., 2011, 11, 85–97 CrossRef CAS PubMed .
  2. P. Poirier, T. D. Giles, G. A. Bray, Y. Hong, J. S. Stern, F. Xavier Pi-Sunyer and R. H. Eckel, Circulation, 2006, 113, 898–918 CrossRef PubMed .
  3. E. E. Calle and R. Kaaks, Nat. Rev. Cancer, 2004, 4, 579–591 CrossRef CAS PubMed .
  4. G. S. Hotamisligil, Nature, 2006, 444, 860–867 CrossRef CAS PubMed .
  5. S. E. Shoelson, J. Lee and A. B. Goldfine, J. Clin. Invest., 2006, 116, 1793–1801 CrossRef CAS PubMed .
  6. K. A. Gendall, S. Raniga, R. Kennedy and F. Frizelle, Dis. Colon Rectum, 2007, 50, 2223–2238 CrossRef PubMed .
  7. C. A. Curat, A. Miranville, C. Sengenes, M. Diehl, C. Tonus, R. Busse and A. Bouloumie, Diabetes, 2004, 53, 1285–1292 CrossRef CAS PubMed .
  8. J. M. Stephens, PLoS Biol., 2012, 10, e1001436 CrossRef CAS PubMed .
  9. Y. Zhang, R. Procenca, M. Maffei, M. Barone, L. Leopold and J. M. Friedman, Nature, 1994, 372, 425–432 CrossRef CAS PubMed .
  10. E. Hu, P. Liang and B. M. Spiegelman, J. Biol. Chem., 1996, 271, 10697–10703 CrossRef CAS PubMed .
  11. P. A. Kern, S. Ranganathan, C. Li, L. Wood and G. Ranganathan, Am. J. Physiol., 2001, 280, E745–E751 CAS .
  12. G. S. Hotamisligil, N. S. Shargill and B. M. Spiegelman, Science, 1993, 259, 87–91 CrossRef CAS PubMed .
  13. A. K. Sarvari, Q.-M. Doan-Xuan, Z. Bacso, I. Csomos, Z. Balajthy and L. Fesus, Cell Death Dis. Nerv. Syst., 2015, 6, e1613 CrossRef CAS PubMed .
  14. P. Sartipy and D. J. Loskutoff, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 7265–7270 CrossRef CAS PubMed .
  15. J. Kawano and R. Arora, J. CardioMetab. Syndr., 2009, 4, 44–49 CrossRef PubMed .
  16. C. N. Lumeng, J. L. Bodzin and A. R. Saltiel, J. Clin. Invest., 2007, 117, 175–184 CrossRef CAS PubMed .
  17. J. Couturier, S. G. Patel, D. Iyer, A. Balasubramanyam and D. E. Lewis, Obesity, 2012, 20, 253–262 CrossRef CAS PubMed .
  18. T. Suganami, J. Nishida and Y. Ogawa, Arterioscler. Thromb. Vasc. Biol., 2005, 25, 2062–2068 CrossRef CAS PubMed .
  19. C. N. Lumeng, S. M. Deyoung and A. R. Saltiel, Am. J. Physiol., 2007, 292, E166–E174 CAS .
  20. L. Xie, M. T. Ortega, S. Mora and S. K. Chapes, Clin. Vaccine Immunol., 2010, 17, 651–659 CrossRef CAS PubMed .
  21. C. F. Nitta and R. A. Orlando, PLoS One, 2013, 8, e77306 CrossRef CAS PubMed .
  22. S. Kim, H. Lee, M. Chung and N. L. Jeon, Lab Chip, 2013, 13, 1489–1500 RSC .
  23. D. R. Bogdanowicz and H. H. Lu, Multifunction Co-culture Model for Evaluating Cell–Cell Interactions, Methods Mol. Biol., 2014, 1202, 29–36 CrossRef PubMed .
  24. B. L. Proffen, C. M. Haslauer, C. E. Harris and M. M. Murray, Connect. Tissue Res., 2013, 54, 14–21 CrossRef CAS PubMed .
  25. I. E. Wang, J. Shan, R. Choi, S. Oh, C. K. Kepler, F. H. Chen and H. H. Lu, J. Orthop. Res., 2007, 25, 1609–1620 CrossRef CAS PubMed .
  26. Y. Zhang, C. Chai, X. S. Jiang, S. H. Teoh and K. W. Leong, Tissue Eng., 2006, 12, 2161–2170 CrossRef PubMed .
  27. H. H. Lu and I. E. Wang, Multi-scale co-culture models for orthopaedic interface tissue engineering, Biomedical nanostructures, Wiley, 1st edn, 2007 Search PubMed .
  28. C. Y. Chan, V. N. Goral, M. E. DeRosa, T. J. Huang and P. K. Yuen, Biomicrofluidics, 2014, 8, 046505 CrossRef PubMed .
  29. A. Mathur, P. Loskill, K. Shao, N. Huebsch, S. Hong, S. G. Marcus, N. Marks, M. Mandegar, B. R. Conklin, L. P. Lee and K. E. Healy, Sci. Rep., 2015, 5, 8883 CrossRef CAS PubMed .
  30. S. Kim, H. Lee, M. Chung and N. L. Jeon, Lab Chip, 2013, 13, 1489–1500 RSC .
  31. P. Kongsuphol, S. Saha, Y. Liu, S. K. Biswas and Q. Ramadan, In vitro adipocyte-PBMC based micro-physiological model of the inflamed human adipose tissue for Immune-metabolic analysis in type II diabetes-Part I: Development of the biological model, Sci. Rep. Search PubMed  submitted.
  32. K. M. Kim, Y. J. Choi, J. Hwang, A. R. Kim, H. J. Cho, E. S. Hwang, J. Y. Park, S. Lee and J. Hong, PLoS One, 2014, 9, e92427 CrossRef PubMed .
  33. Y. Liu, P. Kongsuphol, S. B. N. Gourikutty and Q. Ramadan, Biomed. Microdevices, 2017, 19, 18 CrossRef PubMed .
  34. C. E. Lowe, S. O'Rahilly and J. J. Rochford, J. Cell Sci., 2011, 14, 2681–2686 CrossRef PubMed .
  35. H. Green and O. Kehinde, Cell, 1974, 1, 113 CrossRef CAS .
  36. H. Green and O. Kehinde, Cell, 1975, 5, 19 CrossRef CAS .
  37. N. A. Kraus, F. Ehebauer, B. Zapp, B. Rudolphi, B. J. Kraus and D. Kraus, Adipocyte, 2016, 5, 351 CrossRef CAS PubMed .
  38. C. Farnier, S. Krief, M. Blache, F. Diot-Dupuy, G. Mory, P. Ferre and R. Bazin, Int. J. Obes., 2003, 27, 1178–1186 CrossRef CAS PubMed .
  39. T. Skurk, C. Alberti-Huber, C. Herder and H. Hauner, J. Clin. Endocrinol. Metab., 2007, 92, 1023–1033 CrossRef CAS PubMed .
  40. L. K. Meyer, T. P. Ciaraldi, R. R. Henry, A. C. Wittgrove and S. A. Phillips, Adipocyte, 2013, 2, 217–226 CrossRef CAS PubMed .
  41. R. Drolet, C. Belanger, M. Fortier, C. Huot, J. Mailloux, D. Legare and A. Tchernof, Obesity, 2009, 17, 424–430 CrossRef CAS PubMed .
  42. S. Li, H. J. Shin, E. L. Ding and R. M. van Dam, JAMA, 2009, 302, 179–188 CrossRef CAS PubMed .
  43. M. M. Smith and C. T. Minson, J. Physiol., 2012, 590, 1787–1801 CrossRef CAS PubMed .
  44. A. M. Komai, C. Brännmark, S. Musovic and C. S. Olofsson, J. Physiol., 2014, 592, 5169–5186 CrossRef CAS PubMed .
  45. J. Mauer, B. Chaurasia, J. Goldau, M. C. Vogt, J. Ruud, K. D. Nguyen, S. Theurich, A. C. Hausen, J. Schmitz, H. S. Bronneke, E. Estevez, T. L. Allen, A. Mesaros, L. Partridge, M. A. Febbraio, A. Chawla, F. T. Wunderlich and J. C. Bruning, Nat. Immunol., 2014, 15, 423–430 CrossRef CAS PubMed .
  46. V. Vicennati, A. Vottero, C. Friedman and D. A. Papanicolaou, Int. J. Obes., 2002, 26, 905–911 CrossRef CAS PubMed .
  47. D. Leto and A. R. Saltiel, Nat. Rev. Mol. Cell Biol., 2012, 13, 383–396 CrossRef CAS PubMed .
  48. M. Y. Liu and M. C. Wu, Exp. Hematol., 1992, 20, 974–979 CAS .
  49. H. Hosoya and T. Marunouchi, Cell Struct. Funct., 1992, 17, 263–269 CrossRef CAS .
  50. L. Minafra, G. Di Cara, N. N. Albanese and P. Cancemi, Leuk. Res., 2011, 35, 226–236 CrossRef CAS PubMed .
  51. K. Sintiprungrat, N. Singhto, S. Sinchaikul, S. Chen and V. Thongboonkerd, J. Proteomics, 2010, 73, 602–618 CrossRef CAS PubMed .
  52. T. Matsusaki, K. Kawanabe, K. Ise, T. Nakayama, J. Toguchida and T. Nakamura, J. Arthroplasty, 2007, 22, 960–965 CrossRef PubMed .
  53. L. F. Liu, C. M. Colleen, L. L. Tolentino, O. Choi, J. Morton, H. Rivas, S. W. Cushman, E. G. Engleman and T. McLaughlin, PLoS One, 2017, 12, E0170728 CrossRef PubMed .
  54. J. Jin, K. P. Sundararaj, D. J. Samuvel, X. Zhang, Y. Li, M. F. Lopes-Virella and Y. Huang, Clin. Immunol., 2012, 143, 188–199 CrossRef CAS PubMed .
  55. Y. Fu, N. Luo, R. L. Klein and W. T. Garvey, J. Lipid Res., 2005, 46, 1369–1379 CrossRef CAS PubMed .
  56. V. Rotter, I. Nagaev and U. Smith, J. Biol. Chem., 2003, 278, 45777–45784 CrossRef CAS PubMed .
  57. C. Qi and P. H. Pekala, Proc. Soc. Exp. Biol. Med., 2000, 223, 128–135 CrossRef CAS .

Footnote

Electronic supplementary information (ESI) available. See DOI: 10.1039/c8lc00481a

This journal is © The Royal Society of Chemistry 2019