Rapid Manufacturing of Multilayered Microfluidic Devices for Organ on a Chip Applications

Microfabrication and Polydimethylsiloxane (PDMS) soft-lithography techniques became popular for microfluidic prototyping at the lab, but even after protocol optimization, fabrication is yet a long, laborious process and partly user-dependent. Furthermore, the time and money required for the master fabrication process, necessary at any design upgrade, is still elevated. Digital Manufacturing (DM) and Rapid-Prototyping (RP) for microfluidics applications arise as a solution to this and other limitations of photo and soft-lithography fabrication techniques. Particularly for this paper, we will focus on the use of subtractive DM techniques for Organ-on-a-Chip (OoC) applications. Main available thermoplastics for microfluidics are suggested as material choices for device fabrication. The aim of this review is to explore DM and RP technologies for fabrication of an OoC with an embedded membrane after the evaluation of the main limitations of PDMS soft-lithography strategy. Different material options are also reviewed, as well as various bonding strategies. Finally, a new functional OoC device is showed, defining protocols for its fabrication in Cyclic Olefin Polymer (COP) using two different RP technologies. Different cells are seeded in both sides of the membrane as a proof of concept to test the optical and fluidic properties of the device.


Introduction
During the last two decades, the vast majority of microfluidic systems were built by soft lithography, a technique based on PolyDiMethylSiloxane (PDMS) micro-molding [1]. PDMS features some key properties that have made it to become the material of choice in microfluidics labs. To name some of them: (1) biocompatible; (2) optical clearance; (3) gas-permeability; and (4) patternable at high resolution [1].
Nevertheless, in the last few years, a few limitations have emerged, raising concerns about PDMS as an appropriate material for future production of Organ-on-a-Chip (OoC) devices, particularly, regarding its suitability for cell culture and the need of an easy manufacturability for both research and commercialization.
One of the main concerns is related to its bulk absorption of small molecules compared to the surface-only adsorption of glass and thermoplastics. PDMS was reported to absorb and deplete both estrogens [2] and administered drugs [3] in the media, altering their layer, or withdrawing it from a bulk block. Subtractive manufacturing was implemented in microfluidics since the very beginning when the first devices were produced using methods derived from microsystem technologies such as silicon or glass etching [25][26][27]. There are subtractive techniques, like Xurography or laser-cut, by which, although used as part of an additive process, the final part is obtained by assembling different patterned layers. This process is often referred as Laminated Object Manufacturing (LOM) or sheet lamination. Employing Xurography for the fabrication of microfluidics devices, features down to 6 µm width and 25 µm thick can be obtained in minutes with no masks [28]. CO 2 laser-cut is used for processing different materials, like PMMA (Poly-Methyl-MethAcrylate) [29], PDMS [30], Mylar [31][32][33][34][35][36], and Poly-lactic acid (PLA) [37,38]. Among the techniques used to bond two different layers, we can find heat press [38], heat lamination [28], Pressure-Sensitive Adhesive (PSA) [39,40], or solvent bonding [41]. Therefore, in this paper, we propose some strategies for an easy implementation, costefficacy, and fast realization (~15 min) of multi-layered OoC devices minimizing the uses of cleanroom facilities and high-end equipment. Some thermoplastics alternatives to PDMS, such as PC (Poly-Carbonate), PS, PMMA, and Zeonor ® 1420R COP, have been explored together with pattern fabrication techniques based on both laser-cut and Xurography. Finally, we have evaluated different layer bonding strategies. Results on cell seeding and culture confirm the appropriateness for the selected technology for OoC applications.

Direct Polymer Bonding and APTES Functionalization
O 2 plasma bonding, UV Ozone (UVO 3 ), solvent bonding, and adhesive were explored as bonding strategies. COP, PC, PS, PMMA were tested as bulk material. In addition to them, a bonding test against PC membranes was also performed. APTES-functionalized PC membranes were used for plasma and UVO 3 . It consisted of modifying the surface of the material with a silane composite which could be later oxidized when exposing the hydroxyl (OH) groups. APTES functionalization and plasma activation bonding were reported by Aran et al. [42] as a method for embedding porous polymer membranes to a microfluidic device. This chemical was used as a crosslinker, showing a strong and irreversible bond between the membranes and microfluidic structure, with no significant alteration of membrane transport function or pore morphology. O 2 plasma processing was performed, using a Harrick Plasma Cleaner PDC-002-CE (Harrick Plasma, Ithaca, NY, USA) with O 2 gas. Samples were treated with O 2 plasma (t = [1, 2, 5] min, 29.6 W, 0.2 Torr) after cleaning, brought into contact, and left at 65 • C for at least 10 min. UVO 3 treatment was performed using UV/Ozone Cleaner (BioForce Nanoscience, Ames, IA, USA, t = [2, 10, 20] min). After treatment, samples were brought into contact and left at 65 • C for at least 15 min.
Cyclohexane for COP and Trichloromethane for all other materials were used for solvent bonding. Different material combinations were tested: (1) COP on COP, (2) COP on PC membrane, (3) PC on PC membrane, (4) PMMA on PC, (5) PMMA on PMMA, (6) PS on PS, (7) PS on PC membrane. Substrates were suspended 2 3 mm above a glass Petri dish filled with the solvent. Another glass Petri dish was flipped on top to cover the set-up. After exposing to vapors at 65 • C, the substrates were pressed together and left at 65 • C.
APTES-functionalized PC membranes were prepared following similar protocol as described by Aran et al. [42]. Briefly, a solution of 5% APTES in DeIonized (DI) water was pre-heated in a glass Petri dish at 80 • C and covered with another Petri dish to prevent evaporation. The membrane was activated using O 2 plasma (1 min, 29.6 W, 0.2 Torr) and immediately submersed in the solution. After 20 min, the membrane was washed thoroughly using DI water, deposited on top of a PTFE (Poly-Tetra-Fluoro-Ethylene) flat substrate, and dried using N 2 .

COP Contact Angle Measurement
Measurements were performed using OCA15 Pro system from DataPhysics (Filderstadt, Germany). After focusing the sample, 2-µL drop of de-ionized water was released on top of the surface for digital image capture. Contact angle measurements were performed using SCA 20 software from DataPhysics.

OoC Design and Fabrication
We designed a multilayered device using Autodesk Inventor ® . The device consists of a PC membrane sandwiched between two COP layers, each containing network of 4 parallel microfluidic channels (Figures 1 and A1). Flow is distributed through these parallel channels (layers II and IV) thanks to a fluidic splitter on layer (V). Cross-section dimensions for individual channels are 2 mm × 180 µm, length is 20 mm. Design includes Mini Luer fluidic ports in a 3-mm PMMA layer (VII), to facilitate OoC couple to standard connectors.
After removing the protective liner of a 180-µm thick COP sheet, the substrate was rinsed with IPA (2-propanol) and dried using N 2 . Double-sided PSA was laminated over the COP and PMMA surfaces with the help of a hand-roller, avoiding bubble formations.
Layers were customized using Graphtec ® CE-5000-40 vinyl cutter (Graphtec Corporation, Tokyo, Japan). Cutting was performed using 2 passes at a speed of 30 mms −1 with a mechanical precision of 5 microns, allowing a lateral resolution of 25 microns.
Assembly was performed in a custom-made aluminum aligner. It consisted of a thick base with 4 press-fitted steel dowel pins and a thick cover with 4 holes (both 60 × 60 × 10 mm). Layers located below the membrane were placed in the aligner base ( Figure A1). Aligner cover was placed on top and the whole stack was pressed using an Atlas T8 hydraulic press (Specac Ltd., Orpington, UK) at 2 × 10 3 kg for 30 s. This procedure was repeated for the layers located above the membrane. Finally, a thicker PMMA layer with mechanized Mini-Luer ® ports was assembled to obtain the complete device. After removing the protective liner of a 180-μm thick COP sheet, the substrate was rinsed with IPA (2-propanol) and dried using N2. Double-sided PSA was laminated over the COP and PMMA surfaces with the help of a hand-roller, avoiding bubble formations.
Layers were customized using Graphtec® CE-5000-40 vinyl cutter (Graphtec Corporation, Tokyo, Japan). Cutting was performed using 2 passes at a speed of 30 mms with a mechanical precision of 5 microns, allowing a lateral resolution of 25 microns.
A Trotec Speedy 100 laser cutter (Trotec Laser Inc., Marchtrenk, Austria) was used for laser-cut fabrication. Layers were cut using 2-mm lens ( 40 W, 5000 Hz , speed: 0.15 ms ). Thick layers of PMMA were cut with adhesive side facing up. ¼-28 UNF threads were formed using a threading plug tap.
Assembly was performed in a custom-made aluminum aligner. It consisted of a thick base with 4 press-fitted steel dowel pins and a thick cover with 4 holes (both 60 × 60 × 10 mm). Layers located below the membrane were placed in the aligner base ( Figure A1). Aligner cover was placed on top and the whole stack was pressed using an Atlas T8 hydraulic press (Specac Ltd., Orpington, UK) at 2 · 10 kg for 30 seconds. This procedure was repeated for the layers located above the membrane. Finally, a thicker PMMA layer with mechanized Mini-Luer® ports was assembled to obtain the complete device.

Cell Culture for OoC Application
Each device was activated using the Bioforce UVO 3 tip cleaner (20 min) under a fume hood. The device was connected to a 10 mL disposable syringe (BBraun, Rubi, Spain) using Luer 23 G flat needles (Instechlabs, Leipzig-Markkleeberg, Germany), Tygon ® tubing (VWR International Eurolab, Barcelona, Spain), and 1 /4-28 UNF fluidic connectors (microfluidic ChipShop, Jena, Germany). Inlet reservoirs were filled sequentially with Ethanol (EtOH) 70% and PBS solution to rinse the device microchannels by negative pressure-induced flow. Then, the device was placed onto a hotplate in order to maintain the temperature at 37 • C.
Subsequently, the microdevices were functionalized. The inlet reservoir was filled with 500 µL of coating solution, connecting the pump in refill mode (volume: 200 µL, flowrate: −200 µL min −1 ). Lower channels were functionalized, incubating them with 1:20 collagen solution in PBS (1 h, 37 • C), and upper channels were functionalized with a 2 mg·mL −1 gelatin solution in PBS (15 min, 37 • C). Afterwards, cell suspension of hCMEC/D3 cells in endothelial cells culture medium was inserted at the bottom channel (2.8 × 10 6 cells·mL −1 ) and the devices were left incubating in a humidified CO 2 incubator at 37 • C for at least 2 h, flipped upside-down to promote cell adhesion onto the membrane. Then, a cell suspension of pericytes in pericyte culture medium was inserted at the upper channel (2 × 10 6 cells·mL −1 ). The devices were left incubating in the incubator at 37 • C for at least 2 h before filling each reservoir with the corresponding medium and covering them with a lid.
The devices were left in a humidified incubator at (37 • C, 5% CO 2 ) with no flow and new media was perfused 50 µL/min every 24 h through negative pressure-driven flow.
After 5 days in vitro, the devices were fixed in a 4% PFA solution for 30 min at room temperature, then rinsed in PBS and kept in PBS with 0.02% of sodium azide.
To perform the immunostaining, all solutions were perfused with a negative pressuredriven flow of 50 µL/min. Cells were permeabilized with a 0.1% Triton in PBS solution. After incubating with blocking solution-PBS with 0.1% Triton and 10% FBS-for 2 h, cells were washed and incubated with primary antibody (rabbit polyclonal anti-VE-cadherin antibody, 1:1000) overnight at 4 • C. Then cells were washed and incubated with secondary antibody (Alexa Fluor 568 goat anti-rabbit, 1:1000) for 1 h at room temperature protected from light, washed again, and incubated with the primary monoclonal antibody (Alexa Fluor 488 ZO-1 monoclonal antibody, 1:100) for 3 h at room temperature. After washing, samples were stained with 32.4 µM of Hoechst 33,342 for 10 min, washed, and kept in PBS-azide for imaging in an inverted confocal Leica SP5 microscope. Afterward, images were processed with ImageJ software.

Direct Polymer Bonding
Effective bonding between various layers of the device is fundamental to avoid liquid leakages, which would compromise the entire device functionality. We tested COP bonding to different materials using Plasma treatment, UVO 3 treatment, and solvent bonding. Results are summarized in Table 1. An extended study evaluating other thermoplastics bonding is shown in Table A1 of Appendix A. O 2 plasma and UVO 3 gave generally poor bonding or no-bonding results. Using O 2 plasma to seal two samples of the same bulk polymer failed with all materials. UVO 3 treatment gave almost the same result, with exception of COP-COP bonding. Good strength COP-COP bonding was obtained with higher UVO 3 exposure times and thermal postcuring. Both surface activation methods failed to bond PC membranes to bulk materials. Good results were obtained between COP and APTES-functionalized PC membranes. Sandwiching between two COP layers was not possible, as one of the layers always showed poor bonding.

COP Contact Angle Measurements
Contact angle measurements ( Figure 2) were performed on native COP after O 2 plasma or UVO 3 treatment to evaluate treatment effects on the material. Table 2 presents how the contact angle is significantly reduced after treatment in both cases, turning into a more hydrophilic COP surface. Solvent bonding gave stronger results between same-material combinations (COP, PS, PMMA).

COP Contact Angle Measurements
Contact angle measurements ( Figure 2) were performed on native COP after O2 plasma or UVO3 treatment to evaluate treatment effects on the material. Table 2 presents how the contact angle is significantly reduced after treatment in both cases, turning into a more hydrophilic COP surface.

Design and Fabrication
We selected COP to fabricate our multilayered device using Xurography and lasercut to pattern them. After the poor results obtained with the different solvent bonding approaches, adhesive bonding through 467 MP and adhesive layers were evaluated. Assembly was successful in both cases, though optical quality was better with ARSeal® 8026, (Figure 3). Final assembled devices were successfully tested against leakage for 24 h. They were placed under an optical microscope and PBS/water plus colorant was pumped through the microchannels at 2 mL min −1 .

Design and Fabrication
We selected COP to fabricate our multilayered device using Xurography and lasercut to pattern them. After the poor results obtained with the different solvent bonding approaches, adhesive bonding through 467 MP and adhesive layers were evaluated. Assembly was successful in both cases, though optical quality was better with ARSeal ® 8026, (Figure 3). Final assembled devices were successfully tested against leakage for 24 h. They were placed under an optical microscope and PBS/water plus colorant was pumped through the microchannels at 2 mL min −1 .

Sensing Integration: O2 Concentration Sensing
PreSens® flow-through cells were enclosed in the setup, before and after the chip outlets, to measure the oxygen input and output to/from the device through the tubing. Input concentration was controlled by the microscope incubator chamber and a silicone

Sensing Integration: O 2 Concentration Sensing
PreSens ® flow-through cells were enclosed in the setup, before and after the chip outlets, to measure the oxygen input and output to/from the device through the tubing. Input concentration was controlled by the microscope incubator chamber and a silicone tubing coil used as gas and temperature interchanger. Response of oxygen concentration vs. incubator environment was characterized, as well as flowrate dependence of oxygen concentration ( Figure A2).

Cell Culture for OoC Application
The microfluidic prototype can be used to mimic biological barriers and to analyze its permeability. Brain endothelial cells and pericytes were cultured on opposed sides of the device membrane, mimicking a simplified blood-brain barrier ( Figure A3). The culture of both cell types was successful. After five days in vitro, the bottom channel was covered by an endothelial cell monolayer, and the top channel by pericytes as shown in confocal images of Figure 4. The staining of tight junctions, nuclei, and adherens junctions was performed from the bottom of the device using a confocal microscope, as a routine test for the blood-brain barrier. The devices showed excellent optical characteristics.

Sensing Integration: O2 Concentration Sensing
PreSens® flow-through cells were enclosed in the setup, before and after the chip outlets, to measure the oxygen input and output to/from the device through the tubing. Input concentration was controlled by the microscope incubator chamber and a silicone tubing coil used as gas and temperature interchanger. Response of oxygen concentration vs. incubator environment was characterized, as well as flowrate dependence of oxygen concentration ( Figure A2).

Cell Culture for OoC Application
The microfluidic prototype can be used to mimic biological barriers and to analyze its permeability. Brain endothelial cells and pericytes were cultured on opposed sides of the device membrane, mimicking a simplified blood-brain barrier ( Figure A3). The culture of both cell types was successful. After five days in vitro, the bottom channel was covered by an endothelial cell monolayer, and the top channel by pericytes as shown in confocal images of Figure 4. The staining of tight junctions, nuclei, and adherens junctions was performed from the bottom of the device using a confocal microscope, as a routine test for the blood-brain barrier. The devices showed excellent optical characteristics.  We have compared the results obtained in our device with the classical Petri dish culture ( Figure 5), observing the excellent biocompatibility of our prototype. We have compared the results obtained in our device with the classical Petri dish culture ( Figure 5), observing the excellent biocompatibility of our prototype.

Discussion
The work presented here explores digital manufacturing technologies for OoC device fabrication to enable low-cost chip production for research purposes. Particularly, our fo cus was on LOM techniques, using Xurography and a CO2 laser-cut. After identifying materials following their biocompatible characteristics, we tested possible bonding strat egies. It was not possible to achieve irreversible and solid bonding using O2 plasma on any pristine material, probably due to the relatively low energy of equipment. Although

Discussion
The work presented here explores digital manufacturing technologies for OoC device fabrication to enable low-cost chip production for research purposes. Particularly, our focus was on LOM techniques, using Xurography and a CO 2 laser-cut. After identifying materials following their biocompatible characteristics, we tested possible bonding strategies. It was not possible to achieve irreversible and solid bonding using O 2 plasma on any pristine material, probably due to the relatively low energy of equipment. Although PMMA bonding with UVO 3 and O 2 plasma was previously reported [43], likely higher plasma power and pressures (0.2 1 Pa) were used. Plasma bonding of PC membranes to COP was successfully achieved only after APTES treatment. UVO 3 and solvent bonding results on COP-COP were better, but the optical quality was low. Consequently, double-sided Pressure-Sensitive Adhesive is used as the bonding method between layers.
COP is the selected material for device layers fabrication due to its top-grade cell culture properties after mild plasma treatment (contact angle in 19 36 • range) [14], discarding PC considering its flammability characteristics when exposed to CO 2 laser processing and PS or PMMA for their poor optical properties. Contact angle measurements confirmed that similar surface activation was achieved after 1-min O 2 plasma or 10-20 min UVO 3 treatment.
Only a few works can be found in the microfluidics literature regarding processing COP with cutting plotter [41,44] or laser cutters [45], and even a Lab-on-Chip (LoC) device fabricated in COP using cutting plotter and PSA with an embedded white polyester membrane [46]. COP has a relatively low permeability to water vapor compared to other gases such as oxygen, limiting sample evaporation and favoring the culture of respiring cells [42,44]. COP as well as COC also show other benefits for microfluidics applications, like its high transparency in the visible and near-ultraviolet regions of the spectrum [43], low autofluorescence [45], enabling long-term stable surface treatments, and the deposition of metals by sputtering directly on the substrate.
We showed how both LOM fabrication strategies described here bring the benefits of RP into microfluidics fabrication. Throughput is improved by the reduction of process timing as well as reducing the number of intermediate steps. We reduced the overall process cost and especially costs related to future modifications of the device. We find adhesive bonding to be a versatile solution for rapid prototyping in the research environment, as it enables to extend the process to different membrane materials. On the contrary, solvent or surface activation strategies might limit the process to a smaller range of compatible materials or might require adding chemical functionalization to achieve bonding of different materials. Finally, the OoC technology developed was tested in biological applications to reproduce the blood-brain barrier in vitro by means of co-culture of two different types of cells.

Conclusions
Rapid prototyping methods showed here overcome most of the limitations of PDMS soft lithography. Thanks to the material of choice, COP, and the used methodology, the produced device shows excellent optical characteristics, and low autofluorescence. This fabrication process is rapid (~15 min), easier, low-cost, and enables modifications of the design, including scalability. On the other hand, the process allows an easy solution for embedding commercial established semi-permeable membranes to interconnect different channels or chambers. The technological process is compatible with membrane functionalization and cell culture, allowing to carry out Organ-on-a-Chip devices in an easier manner.