Generation of a Simplified Three-Dimensional Skin-on-a-chip Model in a Micromachined Microfluidic Platform

Summary

Here, we present a protocol to generate a three-dimensional simplified and undifferentiated skin model using a micromachined microfluidic platform. A parallel flow approach allows the in situ deposition of a dermal compartment for the seeding of epithelial cells on top, all controlled by syringe pumps.

Abstract

This work presents a new, cost-effective, and reliable microfluidic platform with the potential to generate complex multilayered tissues. As a proof of concept, a simplified and undifferentiated human skin containing a dermal (stromal) and an epidermal (epithelial) compartment has been modelled. To accomplish this, a versatile and robust, vinyl-based device divided into two chambers has been developed, overcoming some of the drawbacks present in microfluidic devices based on polydimethylsiloxane (PDMS) for biomedical applications, such as the use of expensive and specialized equipment or the absorption of small, hydrophobic molecules and proteins. Moreover, a new method based on parallel flow was developed, enabling the in situ deposition of both the dermal and epidermal compartments. The skin construct consists of a fibrin matrix containing human primary fibroblasts and a monolayer of immortalized keratinocytes seeded on top, which is subsequently maintained under dynamic culture conditions. This new microfluidic platform opens the possibility to model human skin diseases and extrapolate the method to generate other complex tissues.

Introduction

Recently, advances have been made toward the development and production of in vitro human skin models for the analysis of the toxicity of cosmetic and pharmaceutical products¹. Researchers in pharmaceutical and skin care industries have been using animals, mice being the most common, to test their products^2,3,4,5. However, testing products on animals is not always predictive of the response in humans, which frequently leads to drug failure or adverse effects in humans and consequently to economic losses^5,6. The UK was the first country that prohibited the use of animals for cosmetic testing in 1998. Later, in 2013, the EU banned the testing and approbation of cosmetics in animals (EU Cosmetics Regulation No. 1223/2009)⁷.

This prohibition is also being considered by other countries such as in 'The Humane Cosmetics Act' in the USA⁸. In addition to ethical concerns, the anatomical differences between animal and human skin make animal testing time-consuming, expensive, and often ineffective. Furthermore, the global in vitro toxicology testing market size is expected to reach USD 26.98 billion by 2025⁹. For these reasons, there is a need to develop new methods and alternatives for those in vitro studies, such as bioengineered human skin models, that enable testing for safety and toxic effects of cosmetics and drugs without the use of animals.

There are two different kinds of commercially available, in vitro, human skin models. The first type consists of stratified epidermal equivalents containing multiple layers of differentiating keratinocytes that are seeded on different materials. Some of them have been approved by the Organization for Economic Co-operation and Development (OECD) and validated by the (European Centre for the Validation of Alternative Methods (ECVAM) for skin corrosion and irritation testing, such as EpiDerm or SkinEthic^10,11,12. The second type are full-skin equivalents with a layer of differentiating human keratinocytes seeded on a three-dimensional (3D) scaffold that contains fibroblasts, such as T-Skin and EpiDerm-FT. However, these models are cultured under static conditions, which makes them unable to accurately represent human physiological conditions.

Recent interest has focused on generating in vitro 3D skin models in cell culture-insert (CCI) formats with dynamic perfusion^{13,14,15,16,17,18,19}. However, these systems cannot be considered stricto sensu as microfluidic skin-on-chips as per their classical definition in the field. Ingber's definition for organs-on-a-chip states that the organ must be placed inside the microfluidic channels, which is a condition that only a few devices fulfil^20,21. Skin-on-chips have so far modelled mostly simple epithelia as single-cell layers and/or dermal cell layers separated by a porous membrane^22,23. Although there have been some advances modeling skin in microfluidic systems^16,24, there is currently no literature showing an organ-on-a-chip system that fits Ingber's definition, capable of producing a multilayered skin in situ and including both epithelial and stromal components.

In this work, a new, cost-effective, robust, vinyl-based microfluidic platform for skin-on-a-chip applications is presented. This platform was produced by micro-machining, which provides more simplicity in the fabrication process, as well as increased flexibility and versatility in the layout of the device, overcoming some of the limitations of PDMS²⁵. A way to introduce a simplified skin construct through a parallel flow controlled with syringe pumps was also designed. Parallel flow allows two fluids with very different viscosities (a buffer and fibrin pre-gel in this case) to be perfused through a channel without mixing with each other. As a proof of concept, a dermo-epidermal construct containing fibroblasts embedded in a fibrin matrix mimicking the dermis was introduced in the device, on top of which a monolayer of keratinocytes was loaded to emulate the undifferentiated epidermis. The dermal compartment height can be modulated by modifying the flow rates. The main novelty of this work, compared to previously described models^{22,26,27,28,29}, is the development of a 3D construct inside a microchamber by means of microfluidics. Although this article presents a simplified undifferentiated skin, the long-term goal is to generate and characterize a fully differentiated skin construct to demonstrate its viability and functionality for drug and cosmetic testing purposes.

Protocol

1. Chip design and micromachining parameters

1. Design the microfluidic chip layers with FreeCAD open-source design software; refer to Table 1 for the dimensions of the channels. Include four 2.54 mm diameter holes in the design to use a custom-made aligner for a correct layer superposition.

	Length (μm)	Width (μm)
Lower chamber	28,400	800
Upper chamber	31,000	800

Table 1: Dimensions of the upper and lower channels of the device.

2. Cut 95 µm-thick, adhesive, transparent vinyl sheets into 30 cm x 30 cm squares to fit in the plotter properly.
3. Use Brother CanvasWorkspace software to create a 30 cm x 30 cm workspace, and fill it with the designed patterns for the different layers of the chip (Figure 1A). Store it in a .svg file.
4. Cut the 30 cm x 30 cm vinyl sheets with the edge plotter (Figure 1B-D).
   1. Stick the vinyl sheet to a low tack adhesive mat, and eliminate all the air bubbles if necessary.
   2. Upload the .svg file to the plotter, and set the cutting parameters: cutting blade: level 3; cutting pressure: level 0; cutting speed: level 1. Place the adhesive mat with the vinyl into the plotter, and start the cutting process.
5. Cut the top channel pattern on 12 µm-thick double-sided tape vinyl by following the previous steps.

Figure 1: Chip design and micromachining process. (A) Software layout showing the working space filled with both the top and bottom patterns designed for the chip. (B) Edge plotter during cutting process; cutting blade, whole vinyl sheet, and adhesive mat are shown. (C) Patterned vinyl being detached from the cut sheet. (D) Sample of an adhesive vinyl layer patterned with the top channel design. Please click here to view a larger version of this figure.

2. PDMS layer fabrication

1. Mix the PDMS and curing agent in a ratio of 10:1 (v/v), and place the mixture under vacuum for 15 min to remove air bubbles. Pour 55 mL of the mixture into a 55 cm² square culture dish to obtain a 2 mm-thick layer. Remove bubbles with a needle.
2. Cure the mixture (step 2.1) in an oven for 1 h at 80 °C. Unmold the PDMS and cut it into rectangles with the chip dimensions. Make holes for the tubing with an 18 G syringe needle.

3. Chip assembly

NOTE: For better understanding, see Figure 2.

1. Assemble the whole device using an aligner to adjust channels, inlets, and outlets properly. Pile up four vinyl layers (with the corresponding bottom micropattern) for assembling the lower channel, keeping the cover tape of the bottom layer to avoid sticking to the aligner.
2. Cut and place the polycarbonate (PC) porous membrane on top of the lower channel to separate it from the upper one. Be careful not to cover the inlets of the lower channel.
3. Add ten vinyl layers with the upper chamber design. Stick a double-sided tape vinyl layer with the top-channel pattern on top. Remove the chip from the aligner and stick it on the glass slide.
4. Place a 2-mm-thick PDMS sheet on top of the double-sided tape vinyl layer to provide appropriate anchoring for the tubing and to avoid leakage. Leave a weight on top of the chip overnight to ensure the chip is completely watertight. Sterilize the chip by flushing 70% v/v ethanol for 5 min, and then wash with distilled H₂O.

Figure 2: Microfluidic chip assembly. (A) General scheme of the assembly of the device. Lower and upper chambers are composed of four and eleven superimposed vinyl sheets, respectively. (B) Top and lateral views of the microfluidic chip. Top and bottom channels are represented in pink and blue, respectively. (C) Image of the chip assembly using a custom-made aligner. (D) Chip image after complete assembly. Please click here to view a larger version of this figure.

4. Pump connections

NOTE: The graphical representation of pumps connections is shown in Figure 3.

1. Connect Pump 1 to the Upper Chamber Inlet 1 (UCi1).
2. Connect Pump 2 to the Upper Chamber Inlet 2 (UCi2).
3. Connect Pump 3 to the Lower Chamber Inlet (LCi).
4. Connect Upper Chamber Outlet (UCo) and Lower Chamber Outlet (LCo) to a waste tube.
5. Connect the syringes to each inlet using polytetrafluoroethylene (PTFE) tubes and 18 G stainless steel connectors.

Figure 3: Pump connections and inlets/outlets location. (A) Diagram showing the connection of the three different pumps to their respective inlets. Outlets connect to a waste container. (B) Chip image with labeled inlets and outlets. Abbreviations: LCi = lower chamber inlet; LCo = lower chamber outlet; UCi1 = upper chamber inlet 1; UCi2 = upper chamber inlet 2; UCo = upper chamber outlet. Please click here to view a larger version of this figure.

5. Cell culture

NOTE: The HaCaT cell line has a commercial origin. Human primary fibroblasts come from healthy donors and were obtained from the collection of biological samples of human origin registered in "Registro Nacional de Biobancos para Investigación Biomédica del Instituto de Salud Carlos III".

1. Work in a cell culture hood, previously sterilized under ultraviolet light and wiped with ethanol.
2. Thaw H2B-GFP-HaCaT cells (human immortalized skin keratinocytes, hKCs) and GFP-human primary fibroblasts (hFBs) at 37 °C, add 2 mL of culture medium, and centrifuge for 7 min at 20 °C at 250 × g.
   NOTE: H2B-GFP-HaCaT cells are human immortalized keratinocytes modified to express a hybrid histone H2B-green fluorescent protein (GFP), providing their nuclei with green fluorescence. GFP-hFBs are human primary fibroblasts transformed with the vector pLZRS-IRES-EGFP to express cytoplasmic green fluorescence. These cells were modified following previously published protocols^30,31
3. Culture both hKCs and hFBs in 1x DMEM supplemented with 10% fetal bovine serum and 1% of antibiotic/antimycotic solution. Pre-warm the culture medium at 37 °C before use.
4. Detach the cells by washing them with 1x phosphate-buffered saline (PBS), adding 2 mL of trypsin/ethylenediamine tetraacetic acid (EDTA) and incubating them for 10 min at 37 °C.
5. Inactivate trypsin adding 4 mL of culture medium. Resuspend the cells, and transfer them to a 15 mL tube. Remove 10 µL to count cells on a Neubauer chamber and determine the appropriate concentration.
6. Centrifuge the 15 mL tube for 7 min at 20 °C at 250 × g. Remove the supernatant, and resuspend the pellets at the desired concentration: hFBs at 50,000 cells/mL and hKCs at 5·× 10⁶ cells/mL.

6. Fibrinogen pre-gel preparation

1. Activate thrombin by adding 1mL of CaCl₂ (1% w/v in NaCl) to the vial.
2. Add the following components to obtain 1 mL of a fibrin hydrogel at a final concentration of fibrin of 3.5 mg/mL: 59 µL of activated thrombin (10 NIH units/mL), 59 µL of tranexamic acid (Table of Materials, 100 mg/mL), 764 µL of culture medium containing 50,000 hFBs/mL, 118 µL of fibrinogen (20 mg/mL in NaCl (0.9% w/v)).
   ​NOTE: Fibrinogen must be added at the last moment.

7. Parallel flow protocol

1. Pump 1x PBS with pump 3 through the LCi at 50 µL/min during the whole process.
2. Pump sacrificial fluid (1x PBS) with pump 2 through the UCi2 at 100 µL/min.
3. Load the syringe with the pre-gel, rapidly place it into pump 1, and run it at 200 µL/min (Figure 4A).
4. Stop pumps 1 and 2 once the pre-gel exits the UCo.
5. Leave the chip without removing the tubing at 37 °C for at least 10 min to allow gelation.
6. Pump culture medium at 50 µL/h with pump 3 through UCi2 overnight.
7. Block UCi1 with a cap.

8. hKCs monolayer seeding

1. Check under the microscope that hFBs are spread 24 h after the generation of the dermal compartment.
2. Introduce the hKCs with pump 2 through UCi2 at 5 ×·10⁶ cells/mL at 40 µL/min for 1 min (Figure 4B).
3. Leave the chip overnight at 37 °C in a humidity-saturated incubator for cell attachment.
4. Pump fresh culture medium with pump 3 only through LCi at 50 µL/min (Figure 4C).

Figure 4: Microfluidic protocol for the generation of the dermo-epidermal construct. (A) Transverse cross-section showing the parallel flow process to generate the dermal compartment. (B) Keratinocyte monolayer seeding 24 h after dermal compartment generation. (C) Cell culture maintenance inside the microfluidic device. (D) Cross-sectional recreation of the skin inside the chip. Please click here to view a larger version of this figure.

9. Cell viability assay

NOTE: Live/Dead kit stains cells with green or red fluorescence depending on their live or dead state. For proper viability differentiation, non-fluorescent hKCs and hFBs must be used in this step. All the steps in the procedure are carried out through UCi2 with pump 2.

1. Wash the top channel with 1x PBS for 5 min at 50 µL/min to remove culture medium.
2. Pump air to remove the 1x PBS at 50 µL/min.
3. Prepare Calcein AM/Ethidium homodimer-1 Kit (Live/Dead) solution by following the manufacturer's instructions.
4. Pump Live/Dead solution at 50 µL/min for 2 min.
5. Incubate 30 min at 37 °C in the dark.
6. Wash top channel by pumping 1x PBS at 50 µL/min for 2 min to remove any remaining reagent.
7. Observe the sample under the confocal microscope. Use an excitation wavelength of 495/590 nm and an emission wavelength of 519/617 nm for live and dead cells, respectively.

Representative Results

The designed chip is composed of two fluidic chambers separated by a 5 µm pore size PC membrane that allows the growth of the cell by allowing the passage of growth-promoting molecules from the lower chamber. The upper chamber holds the tissue construct, in this case, a monolayer of hKCs on a fibrin hydrogel containing hFBs.

The height of the channels is determined by the number of adhesive sheets added to each channel. The lower chamber is composed of 4 layers (380 µm) and the upper one of 10 one-sided tape layers and a double-sided one (962 µm). The dimensions of the chip are 4 cm x 2 cm, which enhances its manipulation. The adhesive vinyl sheets provide water-tightness and transparency for visual inspection of the device. The PDMS layer was useful for the proper anchoring of llthe tubing to avoid any leakage from the holes where the tubes were fixed.

According to published literature, the injectability of cell-containing hydrogels into microfluidic chambers using syringe pumps has not been reported to date. For this reason, the injectability of the fibrin pre-gel had to be assessed. We observed that under a flow of 50 µL/min, the syringe was blocked. However, flow rates higher than 200 µL/min could damage the cells. Rheological studies were performed to test the shear thinning behavior of the fibrin pre-gel, obtaining viscosities from 10 to 50 cP within the selected flow rate range (50-200 µL/min). These results helped to establish the working conditions of this system.

In this work, a parallel flow method has been developed based on the generation of two superimposed laminar flows the lower one being the pre-gel, while the upper one was sacrificial fluid (PBS). Numerically solving Navier-Stokes equations and imposing the appropriate boundary conditions, we found that there were multiple possible solutions to obtain the desired height. Considering the shear rate limits established earlier, to achieve a hydrogel of approximately 500 µm height, the flow rates were 104 and 222 µL/min for the sacrificial PBS and the pre-gel, respectively. In practice, microflow rates of 100 and 200 µL/min were used for simplicity. When re-introduced in the model, these values were found to result in a gel height of 576 µm, very similar to the expected values (Figure 5B). To experimentally check the functioning of the proposed method, the height of the hydrogel along the upper chamber was measured. An average height of 550 µm was observed (Figure 5A), quite similar to the prediction of our mathematical model.

Figure 5: Parallel flow mathematical solutions to choose the appropriate pre-gel and sacrificial fluid flows to obtain the desired dermal height. (A) Front view of the confocal image of hKCs seeded on top of the fibrin gel to assess its height. (B) Mathematical solution for different heights depending on the pre-gel and PBS flow rates. Abbreviations: hKCs = human immortalized skin keratinocytes; PBS = phosphate-buffered saline. Please click here to view a larger version of this figure.

When establishing the protocol at the beginning, PBS was not flushed through the lower channel, leading to discrepancies between the theoretical height of the gel and the one measured. This difference in height was ~40% compared with the estimated one (Figure 6). Once the protocol was optimized and PBS pumped through the lower chamber, this loss in height was resolved (Figure 5).

Figure 6: Confocal view of hKCs seeded on top of the hydrogel to measure its height when PBS was not pumped through the lower chamber. Abbreviations: hKCs = human immortalized skin keratinocytes; PBS = phosphate-buffered saline. Please click here to view a larger version of this figure.

Figure 7A shows a fluorescent top view image of the upper chamber containing a fibrin hydrogel with embedded GFP-hFBs, demonstrating that 24 h after loading, cells are uniformly distributed along the chamber and well spread. Figure 7B shows confluent GFP-hKCs seeded on top of the hydrogel.

Figure 7: Fluorescence images of the upper channel showing different cells seeded in the device. (A) Top view of the upper chamber 24 h after parallel flow protocol showing the hFBs embedded in the fibrin gel. (B) Confluent GFP-hKCs seeded on top of the hydrogel 24 h after hydrogel generation. Dashed red line indicates channel walls. Scale bars: 400 µm. Abbreviations: hFBs = human primary fibroblasts; GFP-hKCs = green fluorescent protein-expressing human immortalized skin keratinocytes. Please click here to view a larger version of this figure.

It is important to keep the system closed overnight until the hKCs sediment and attach to the hydrogel surface. When the tubing is removed before cell attachment, air bubbles enter the channel and displace cells, leading to a nonuniform confluent monolayer, as shown in Figure 8.

Figure 8: Top view of the hydrogel surface when removing tubing immediately after hKCs seeding. Dashed red line indicates channel walls. Scale bar: 400 µm. Abbreviations: hKCs = human immortalized skin keratinocytes. Please click here to view a larger version of this figure.

The cell viability test performed on hFBs embedded in the hydrogel was carried out 24 h after loading using the parallel flow method, showing a cell viability of ~95%. The same test performed on hKCs cells 24 h after seeding them on the fibrin hydrogel showed similar results. The next step was to generate a dermo-epidermal construct and study its structure using confocal microscopy after 24 h (Figure 9).

Figure 9: Reconstructed 3D confocal image of the undifferentiated skin model in the microfluidic chip. Yellow dashed line indicates the surface of the hydrogel separating hKCs (top) from the hFBs embedded in the gel (bottom). Abbreviations: hFBs = human primary fibroblasts; hKCs = human immortalized skin keratinocytes. Please click here to view a larger version of this figure.

It is crucial to find an equilibrium between the shear thinning behavior of the fibrinogen gel and its gelation time: if it takes too long to establish the parallel flow, it coagulates and blocks the system; if the gelation process is too slow, hFBs in the hydrogel will sediment as shown in Figure 10. The behavior of this transient state can be regulated by varying the thrombin concentration.

Figure 10: Confocal image of a fibrin hydrogel showing sedimented hFBs (in red) due to a slow fibrin gelling time. A hKCs monolayer (in blue) is shown on top of the gel. Abbreviations: hFBs = human primary fibroblasts; hKCs = human immortalized skin keratinocytes. Please click here to view a larger version of this figure.

During the device planification and posterior experimental practices, some complications arose that had to be solved to obtain an optimally functioning device and well-structured tissue. These problems are shown in Table 2, along with the solutions for troubleshooting.

Problems	Solutions
Introducing a 3D hydrogel in the upper chamber, leaving free space above to seed keratinocytes on top after dermal generation	Apply a parallel flow of two fluids (PBS and pre-gel) with different viscosities to create a dermal compartment with a controlled height.
Channel misalignment during vinyl sheet stacking	Use of a custom-made aligner to pile up all the sheets in the correct place
Achieve a confluent keratinocyte monolayer on top of the fibrin gel to simulate the epidermis	Allow cell attachment to the gel prior to removing the tubes from the chip, preventing air bubbles from entering the channel and displacing the cells.
Hydrogel height loss along the channel due to pre-gel leaking to the lower channel during parallel flow protocol through the porous membrane.	Pump PBS through the lower channel during parallel flow establishment.
Fibroblast sedimentation inside the gel	Thrombin concentration was slightly increased to accelerate fibrin gelation

Table 2: Issues found during the development of the current work and solutions applied.

Discussion

The motivation to develop this method was the desire to model skin diseases and study the effects of new and innovative therapies in a high-throughput platform. To date, this laboratory produces these dermo-epidermal equivalents by casting-either manually or with the help of the 3D bioprinting technology-the fibrin gel with fibroblasts into a cell culture insert plate and seeding the keratinocytes on top of it. Once the keratinocytes reach confluence, the 3D culture is exposed to the air-liquid interface, which induces keratinocyte differentiation, generating a stratified epidermis and consequently, a fully developed interfollicular human skin^32,33,34. However, these 3D cultures, albeit very relevant clinically, are expensive, time-consuming, and do not mimic physiological conditions.

Tissue-on-a-chip technology provides a platform to emulate physiological conditions in cell culture, thus enabling a better understanding of the toxicity, efficacy, and delivery of drug candidates^{13,14,15,16,17,18,19,35}. Most of the tissues modeled on microfluidic "classical" chips are composed of single layers of cells, mostly epithelial cells^{22,26,42,27,28,36,37,38,39,40,41}. The modelling of more complex, multi-layered tissues has been hampered by the difficulty of generating homogeneous and superimposed tissue layers.

The main novelty of this work, as compared to previous devices^22,43, is the development of a method to generate an undifferentiated and simplified skin construct by means of microfluidics. Moreover, another important innovation with respect to other published technologies⁴⁴ is the design of a relatively simple, cost-effective, easy-to-use disposable chip system made of biocompatible vinyl sheets allowing ad hoc fabrication. This system avoids the use of silicon wafers and complicated plasma bonding procedures needed with PDMS²⁵. Chips based on this material require the generation of wafers specific for each different chip design, which raises the price of the prototyping process. All of this makes the current "classical" technology expensive, complex, and not very flexible.

To overcome these limitations, using skin as a tissue model, we present a very flexible, cheap, and robust microfluidic platform based on vinyl layers, produced by micromachining. We also present a new methodology based on the parallel flow of laminar fluids that allows the in situ generation of a bilayer skin construct with a lower dermal compartment containing hFBs and an upper epidermal compartment composed of a monolayer of hKCs. The chip consists of two chambers separated by a porous PC membrane. The upper channel contains the skin construct and leaves free space to allow hKC differentiation and stratification and/or perfusion of culture medium, air, or even drugs in the future. The lower chamber is continuously perfused with culture medium to promote cell growth. The use of a porous membrane may lead to the loss of a part of the pre-gel from the upper chamber to the lower one when subjected to high pressure due to pumping force. Introducing a PBS flow in the lower chamber compensates this pressure difference and avoids this leakage.

The distribution of hKCs on the hydrogel surface is an important aspect when generating a skin model. When they are not homogeneously spread, the generation of a uniform monolayer and therefore, the epidermal differentiation could be hampered. In the same way, hFBs must be equally distributed within the hydrogel to resemble their natural location in which they are found in real skin. To avoid cell sedimentation or tubing blockage, hydrogel composition (especially thrombin concentration) must be carefully studied to control gelation times and shear thinning behavior.

Despite these encouraging findings, future work is needed to demonstrate the long-term functioning of the system necessary to promote the correct proliferation and differentiation of the hKC monolayer to form a well differentiated human skin including a stratum corneum. Besides skin, this system would allow the generation of other complex, multilayered tissue constructs.

Materials

Name	Company	Catalog Number	Comments
Amchafibrin	Rottafarm		Tranexamic acid
Antibiotic/antimycotic	Thermo Scientific HyClone
Calcium chloride	Sigma Aldrich
Culture plates	Fisher
DMEM	Invitrogen Life Technologies
Double-sided tape vynil	ATP Adhesive Systems		GM 107CC, 12 µm thick
Edge plotter	Brother		Scanncut CM900
FBS	Thermo Scientific HyClone
Fibrinogen	Sigma Aldrich		Extracted from human plasma
Glass slide	Thermo Scientific
GFP-Human dermal fibroblasts	-		Primary. Gift from Dr. Marta García
H2B-GFP-HaCaT cell line	ATCC		Immortalized keratinocytes. Gift from Dr. Marta García
Live/dead kit	Invitrogen
PBS	Sigma Aldrich
Polycarbonate membrane	Merk TM		5 µm pore size
Polydimethylsiloxane	Dow Corning		Sylgard 184
Sodium chloride	Sigma Aldrich
Syringes	Terumo		5 mL
Thrombin	Sigma Aldrich		10 NIH/vial
Transparent adhesive vinyl	Mactac		JT 8500 CG-RT, 95 µm thick
Trypsin/EDTA	Sigma Aldrich
Tubing	IDEX		Teflon, 1/16” OD, 0.020” ID