Inflammatory diseases are often characterised by excessive neutrophil infiltration from the blood stream to the site of inflammation, which damages healthy tissue and prevents resolution of inflammation. Development of anti-inflammatory drugs is hindered by lack of in vitro and in vivo models which accurately represent the disease microenvironment. In this study, we used the OrganoPlate to develop a humanized 3D in vitro inflammation-on-a-chip model to recapitulate neutrophil transmigration across the endothelium and subsequent migration through the extracellular matrix (ECM). Human umbilical vein endothelial cells formed confluent vessels against collagen I and geltrex mix, a mix of basement membrane extract and collagen I. TNF-α-stimulation of vessels upregulated inflammatory cytokine expression and promoted neutrophil transmigration. Intriguingly, major differences were found depending on the composition of the ECM. Neutrophils transmigrated in higher number and further in geltrex mix than collagen I, and did not require an
Inflammatory responses are required to successfully combat invasion by pathogens. However, excessive or unresolved inflammation can damage healthy tissue and result in chronic inflammatory conditions such as rheumatoid arthritis, ulcerative colitis, cardiomyopathies, cardiovascular disease, and asthma.
During an inflammatory response, neutrophils move from the blood to the site of inflammation by transmigrating across the endothelial barrier and through the basement membrane (BM) of the surrounding extracellular matrix (ECM). This multi-step process is enabled by inflammatory receptors and molecules expressed by the endothelial cells which capture the neutrophils and support their extravasation, alongside gradients of chemokines that guide the neutrophils through the ECM. The neutrophils respond by upregulating integrins that allow attachment to the endothelium and subsequently aid their movement through the BM
Neutrophil transmigration in vivo occurs in a 3D environment, and therefore typical 2D in vitro techniques cannot accurately recapitulate the process
Organ-on-a-chip is a new and fast-growing field that has the potential to cross the translation gap and provide humanized models for use in both basic biology research and drug discovery. These computer chip-sized devices provide cells with a more physiological 3D environment that supports cell differentiation, enabling organ-like functions. The chips commonly consist of channels coated with ECM proteins and lined with one or more cell types. Additional physical factors such as shear flow, air–liquid interfaces, and cyclical deformation of the chips can also be incorporated
To model neutrophil transmigration during inflammation, the key components are the endothelial barrier and the ECM encountered post-transmigration. Neutrophils most commonly extravasate through post-capillary venules, which increase in permeability during inflammation. To recapitulate this, inflammatory mediators such as TNF-α are used to stimulate the endothelial barrier and induce production of inflammatory molecules that promote transmigration. After crossing the endothelium, neutrophils then encounter the basement membrane, a meshwork of proteins consisting primarily of collagen type IV, laminin, heparan sulfate proteoglycans, and nidogen
Although some organ-on-a-chip models have incorporated an endothelial component, this is usually a 2D sheet of endothelial cells on a porous membrane
Here, we developed a novel model of neutrophil transmigration during inflammation, using the OrganoPlate system (Mimetas), a commercially available platform which has been used to model many organs already including the blood–brain barrier
In this study, we used this platform to explore the effects of different ECM composition on trans endothelial neutrophil migration. This work demonstrates the importance of considering choice of ECM when designing organ-on-a-chip models. The multi-well plate format of the Organoplate enabled testing of anti-inflammatory compounds, supporting a use for the model in screening potential drug candidates for diseases involving excessive neutrophil infiltration and inflammation.
The neutrophil transmigration model was constructed in a Mimetas OrganoPlate 3-lane. The 3-lane OrganoPlate consists of a 384-well plate containing 40 chips, each with three channels. The channels are separated by Phaseguides, a liquid pinning technology which enables vessels to be grown against ECM, without any physical separation by membranes Incorporating collagen type I into geltrex extracellular matrix (ECM) enables formation of a leak-tight endothelial vessel in the 3-lane OrganoPlate. (
Two different ECMs, geltrex and collagen I, were compared in the OrganoPlate. Geltrex consists of a mix of proteins found physiologically in the endothelial basement membrane. However, when HUVEC were grown against geltrex alone, cells differentiated into tubes on the ECM, preventing proper vessel formation (Fig.
To develop an inflammation model, vessels were grown in the top channel with ECM patterned into the middle channel. Stimulation of confluent vessels with TNF-α upregulated expression of ICAM-1 on the vessel surface (Fig. Tumour necrosis factor alpha (TNF-α) stimulation induces vessel expression of inflammatory molecules. (
Addition of freshly isolated human neutrophils into the inflamed vessels led to neutrophil transmigration out of the vessels and into the ECM (Fig. ECM composition affects characteristics of neutrophil transmigration across HUVEC vessels. (
Very few neutrophils transmigrated across unstimulated vessels. Neutrophil transmigration differed between geltrex mix and collagen I ECMs. Fewer neutrophils transmigrated into collagen I, and post-transmigration neutrophils appeared unable to migrate towards the fMLP, stalling within 150 µm of the vessel. In contrast, neutrophils easily migrated throughout geltrex mix, covering the whole width of the gel channel during the 1.5-h experiment (Fig.
Interestingly, when the chemoattractant was not present, neutrophils still transmigrated into geltrex mix, although the number transmigrated was slightly reduced. In contrast, fMLP was required for transmigration into collagen I, with transmigration without fMLP comparable to unstimulated levels. The distance of migration in gel was not affected by presence of chemoattractant, in either geltrex mix or collagen I (Fig.
To investigate the potential of our model for testing anti-inflammatory or immunomodulatory compounds, we investigated the effect of inhibiting neutrophil proteases. A cocktail of protease inhibitors was used; an elastase inhibitor (sivelestat), a cathepsin G inhibitor (Cathepsin G Inhibitor 1), and a dual elastase and cathepsin G inhibitor (human recombinant SLPI). These were first tested in HUVEC-free OrganoPlates, to isolate their effect on neutrophil migration into the ECM. In this experiment, the ECM was plated 1 day before neutrophil addition. fMLP was added directly before neutrophil perfusion.
Neutrophils were unable to migrate into collagen I at all without HUVEC present (Fig. Inhibition of neutrophil proteases. (
When HUVEC were incorporated into the model, number of neutrophils transmigrating into geltrex mix was inhibited by the protease inhibitor cocktail across (Fig.
In this study we present a new 3D in vitro model of neutrophil transendothelial migration and ECM infiltration. Confluent endothelial vessels were successfully cultured against two different types of ECM. The endothelial cells were stimulated to mimic an inflammatory state, leading to primary human neutrophil transmigration into the ECM. Significant differences were found between neutrophil transmigration into the different ECMs, including differences in response to pharmacological inhibition. This work highlights the ECM as an important modulator of cell phenotype, and ECM composition should therefore be considered carefully when designing complex in vitro models.
The Mimetas 3-lane OrganoPlate was chosen as the platform for the model, as it allows culture of cells under perfusion and against a 3D ECM without any physical separation by membranes or barriers
ECM composition is a major regulator of neutrophil infiltration and yet physiologically relevant ECM are often overlooked in favour of experimentally convenient substrates. Two types of ECM were tested in the model, collagen I and geltrex. Collagen I is well established as a 2D coating or 3D ECM for endothelial cells, as it robustly supports adherence and monolayer formation
The difference in endothelial phenotype between collagen I and BMEs is probably due to multiple factors, including differences in stiffness of the gels and protein composition. A previous study cultured HUVEC on polyacrylamide gels, and found that as stiffness of the gel decreased HUVEC tended to undergo tube formation, but maintained monolayer formation at a higher stiffness
Inflammation was induced in the vessels using TNF-α, a commonly used inflammatory stimulus, although other inflammatory stimuli could be readily compared in future studies. Immunofluorescence staining of surface ICAM-1, a key receptor for neutrophil binding, showed a significant upregulation upon stimulation. Analysis of vessel supernatants found induction of four inflammatory cytokines—soluble ICAM-1, CXCL1, CCL2 and IL-8. These cytokines have established roles in neutrophil transmigration
When neutrophils were added to stimulated vessels, differences in transmigration were observed depending on the ECM used. The first major difference seen was that many more neutrophils transmigrated into geltrex mix than collagen I, suggesting that vessels are more permissive to transmigration on geltrex mix. There is very little data directly comparing transmigration across endothelium on substrates of different protein composition, likely due to the propensity of HUVEC to undergo tube formation when cultured on BMEs. Studies have found that increasing stiffness of the ECM increases neutrophil transmigration
Following transmigration, ability of neutrophils to infiltrate the ECM was hindered in collagen I compared to geltrex mix. This could again be due to mechanical differences between the ECMs, such as stiffness or pore size. Several groups have compared the structure of polymerized collagen I to BME gels, finding that although BME gels have smaller pores they are more deformable than collagen I matrices
We also explored the requirement for a chemoattractant gradient in the model. Absence of fMLP in chips with geltrex mix did not significantly affect numbers or distance of neutrophils transmigrated. In contrast, vessels grown against collagen I required the presence of fMLP for induction of neutrophil transmigration. Geltrex mix itself does not induce neutrophil infiltration, as no neutrophil migration was observed into geltrex mix without HUVEC or fMLP present. Therefore, the threshold for induction of transmigration differs between the ECMs, with transmigration into collagen I requiring an extra inflammatory signal to direct the neutrophils across the endothelium.
We next investigated a role for neutrophil proteases in navigating the ECM post-transmigration. Whereas some reports suggest that neutrophils move passively through the basement membrane, favouring ‘low expression regions’ with lower protein concentration
In contrast to collagen I, neutrophils easily migrated into geltrex mix without presence of a vessel. Moreover, inhibiting elastase and cathepsin G moderately reduced this infiltration, demonstrating a contribution from neutrophil proteases. When the inhibitors were tested with vessels present, a modest but consistent inhibition of migration was seen in geltrex mix vessels. This was not seen in vessels formed on collagen I. This suggests that elastase and cathepsin G are partially responsible for enabling migration in geltrex mix. Other proteases such as matrix metalloproteinases (MMPs) have also been found to play a role in neutrophil infiltration, and may also play a role here
Our results emphasise the role of the ECM in modulating cell phenotype and response to inhibitors. Previous studies, particularly in the field of cancer, have also found that 3D environment and ECM composition affect drug efficacy. For example, several groups have found cancer cells to be less responsive to chemotherapy in 3D cultures than in 2D monolayers
3D in vitro organ on a chip models have the potential to bridge the gap between current in vitro and in vivo models. Development of these models requires knowledge of the microenvironment being modelled, including the physical 3D structure and the cell types and signals present. Inflammatory diseases are often characterised by excessive neutrophil infiltration and associated ECM remodelling. Here, we have successfully established a neutrophil transmigration on a chip model where neutrophils transmigrate out of an endothelial vessel and through a 3D ECM environment, as they would in vivo. We found substantial differences in characteristics of neutrophil transmigration depending on the ECM used, including different responses to inhibitors. This work highlights how extracellular components influence cell behaviour, and demonstrates a need for accurate recapitulation of in vivo microenvironments when designing in vitro models. Further work is required to fine-tune ECM structures for use in models and to delineate exactly how the ECM alters cell phenotype.
All materials were obtained from Sigma (UK) unless specified otherwise.
All experiments involving human participants was reviewed by the Human Biology Research Ethics Committee, University of Cambridge, with fully-informed, written consent in accordance with the Declaration of Helsinki.
Human umbilical vein endothelial cells (HUVEC, PromoCell, Germany) were cultured in Endothelial Growth Medium (Media, PromoCell) with 30 µg/mL gentamicin, in T25 or T75 tissue culture flasks. Cells were grown to confluence (37 °C, 5% CO2) and then passaged using PromoCell Detach Kit. HUVEC were used at passage 4.
For collagen I ECM, 5 mg/mL rat tail collagen type I (ibidi, Germany) was neutralised by mixing at a ratio of 8:1:1 with 1 M HEPES and 37 g/L NaHCO3 producing a 4 mg/mL solution. For geltrex mix ECM, Geltrex™ LDEV-Free Reduced Growth Factor Basement Membrane Matrix (Gibco, A1413302) was diluted to 14.6 mg/mL in media and then mixed at a ratio of 75:25 (v/v) with 3 mg/mL collagen I. 3 mg/mL collagen I was made by neutralisation of 5 mg/mL collagen I with 1 M NaOH and 7.5% (w/v) NaHCO3 as per the manufacturer’s protocol. For each batch of geltrex, preliminary screening experiments were performed to confirm successful vessel formation. Some batches of geltrex required additional coating of the ECM with 10 µg/mL fibronectin for 2 days to support HUVEC adhesion.
Three-lane OrganoPlates® (Mimetas, 4003-400B, Netherlands) were used. 50 µL Hanks’ Balanced Salt Solution (without calcium and magnesium, HBSS, Gibco) was added to the observation windows to prevent chips from drying out. 1.75 µL geltrex mix or 1.85 µL collagen I was added to the middle lane of the chip via the gel inlet (Fig.
HUVEC were seeded into the top channel by addition of 2 µL cell suspension at 2 × 107/mL. Media was added to the top channel inlet and the plate incubated for 2 h on its side at a 70° angle (37 °C, 5% CO2) to allow cells to attach to the ECM. Media was then added to the top channel outlet and perfusion begun (8 min intervals, 7° angle) on the Mimetas OrganoFlow S plate rocker. Plates were incubated under perfusion for 7 days to enable optimal barrier function of the vessels. Media was refreshed every 2–3 days.
After 7 days of incubation, the barrier function of the vessels was tested. 0.5 mg/mL 70 kDa tetramethylrhodamine (TRITC)-dextran was added to the top channel for 30 min. Dye retention was imaged on a Leica SP5 confocal microscope. Barrier function was quantified by measuring the fluorescence in each channel in FIJI
Blood was taken from volunteers by venepuncture into 3.8% (v/v) sodium citrate-containing vacutainers (Greiner Bio-One). The volunteers were healthy and free of medication for at least 10 days prior to venepuncture. Fully-informed, written consent was obtained in accordance with the Declaration of Helsinki, and use of blood was reviewed by the Human Biology Research Ethics Committee, University of Cambridge. Nine mL blood was layered onto 5 mL histopaque-1077 and then centrifuged (428
On day 7 of culture, vessel lumens were stimulated with 1.6 ng/mL human recombinant tumour necrosis factor-alpha (TNF-α) (R&D Systems) overnight. Immediately before neutrophil perfusion, the bottom channels of the chips were filled with 200 nM
A cocktail of 100 µM sivelestat (Biotechne), 10 µM Cathepsin G Inhibitor 1 (Cambridge Bioscience), and 250 nM human recombinant secretory leukocyte peptidase inhibitor (SLPI) (Biotechne) was used. Neutrophils were incubated with the inhibitors or 0.2% (v/v) DMSO vehicle control for 15 min at room temperature before addition to the OrganoPlate®.
Vessels were fixed with 4% (w/v) paraformaldehyde (Santa Cruz Biotechnology) for 15 min, washed twice for five minutes with PBS and then blocked with 2% (w/v) BSA for 30 min. Anti-CD31 antibody (Abcam, ab9498) in 2% (w/v) BSA was added and incubated overnight at room temperature on a plate rocker. Vessels were washed twice with PBS and then anti-mouse-647 (ThermoFisher) secondary antibody, Hoechst-33342 (ThermoFisher), and phalloidin-FITC (ThermoFisher) were added and incubated for 1.5 h at room temperature in the dark, on a rocker. Two more PBS washes were performed. Images were obtained on a Leica SP5 confocal microscope, using a 20× objective. 3D reconstructions were created in FIJI
R&D Systems Cytokine Proteome Profiler Arrays were used to investigate cytokine production. Supernatants were taken from five chips 24 h after TNF-α stimulation and pooled together. Cytokine arrays were carried out as per manufacturer’s instructions, with the modified protocol for detection using AlexaFluor™ 790 Streptavidin (ThermoFisher). A LI-COR Odyssey Fc was used for detection of arrays. Cytokine levels were quantified by densitometry in FIJI. Cytokine expression was expressed as a % of the positive reference spots for each array.
Statistical analysis was performed in GraphPad Prism v9. All tests were carried out using matched analyses. Two-way ANOVA with Sidak’s multiple comparisons test was used to analyse differences in surface ICAM-1 levels, cytokine expression, or number of neutrophils transmigrated between collagen I and geltrex mix, and unstimulated versus TNF-α stimulated. One-way ANOVA with Tukey’s multiple comparisons test was used to compare differences in number of neutrophils transmigrated between unstimulated, TNF-α, and TNF-α + fMLP, and also to compare unstimulated, stimulated, and protease inhibitor-treated conditions. For differences in distance of neutrophil migration, paired two-tailed t tests were used. Values P < 0.05 were considered significant. In figure legends, N refers to independent experiments/neutrophil donors, and n refers to technical repeats (number of vessels) within donors.
R.R. designed, performed and analysed experiments and drafted the manuscript. K.J., K.H. supervised research direction and edited the manuscript. M.H. supervised research direction, designed experiments, and edited the manuscript.
This work was supported by an AstraZeneca-Cambridge PhD Studentship. Research in MTH lab is also supported by NC3Rs/BHF PhD studentship NC/N002350/1 and British Heart Foundation project Grant PG/20/12/34982.
Data are available from the corresponding author on reasonable request.
The authors declare no competing interests.
Supplementary Information 1. Supplementary Video 1. Supplementary Video 2. Supplementary Video 3.
The online version contains supplementary material available at
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.