Robust and Seamless Integration of Thiol‐Ene‐Epoxy Thermosets with Thermoplastics and Glass as Hybrid Microfluidic Devices Suitable for Drug Studies

Facile, durable and biocompatible integration of hybrid Lab‐on‐a‐Chip devices has remained an important challenge in microengineering. Here, the robust and seamless integration of off‐stoichiometry thiol‐ene‐epoxy (OSTE+) thermosets with a diverse range of thermoplastic and glass materials is demonstrated. While direct bonding offers sturdy proprieties for the majority of the material hybrids (up to 1.5 MPa), further surface priming enhances the bonding strength for those of lower surface energy (up to seven‐fold). Using microfluidic devices with a unique configuration of chambers and microchannels down to 100 µm, the impact of various bonding methods on their fluidic functionality is further shown. To investigate the utility of hybrid devices for pharmacological and toxicological studies, both pristine and primed devices are benchmarked against polydimethylsiloxane (PDMS) with regard to cell biocompatibility and drug absorption. While being on par with PDMS in terms of cell viability, pristine OSTE+ devices show an 11‐fold lower absorption of hydrophobic drug molecules. Although surface priming enhances the bonding strength, it compromises other criteria in device performance and biological applications. Combined, results demonstrate a novel integration approach for facile manufacturing of hybrid microfluidic devices using a material toolbox suitable for cell and pharmaceutical studies.


Introduction
[3] LoC has adapted microand nanofabrication techniques such as photo [4] and nanoimprint lithography [5] in tandem with the less conventional soft lithography [6] to create well-defined structures for micro-and nanofluidic applications.Unlike the semiconductor field, which is heavily reliant on silicon processing, [7] microfluidics has gradually shifted to glass [8] and polymer-based devices. [9]Glass has remained a strong candidate material due to high optical transparency, surface inertness and resistance to solvents. [10]Polymers have also attracted attention due to the ease of processing and low cost of fabrication. [11]Specifically, thermoplastic, [12] thermosetting [13] and elastomer [14] polymers have provided a rich toolbox of materials for nanoengineers.Microfluidic devices are typically subject to back-end processing such as sealing and bonding, world-to-chip interfacing, sensor integration and electrical connection [15] and such processes can constitute a major part in overall cost estimation, even up to 80% in manufacturing. [16]or sealing and bonding purposes, a diverse range of integration techniques have been developed that can be categorized into direct and indirect bonding.In direct bonding, no additional materials such as surface modifiers or chemical reagents are used.Thermal fusion [17] and solvent-assisted bonding [18] are two well-known examples of this category.However, high temperature and pressure regimens in thermal bonding commonly lead to the deformation and even collapse of structures, [19] while the use of solvents can entail the distortion or blockade of microchannels due to solvated polymers. [20]Other examples of direct bonding include surface modification-based techniques that create polar functional groups on the surface using plasma or UV exposure. [21]However, the bonding strength resulting from such methods deteriorates over time. [22]In indirect bonding, additional chemical reagents, modifiers or tapes are used.Particularly epoxy resins [23] and adhesive tapes [24] as intermediate materials to seal and bond different layers of microfluidic devices are becoming increasingly prevalent as they provide a simple and easy way to interface different material types.These methods, however, entail the risk of introducing new substances that can alter biocompatibility, surface energy, monomer leaching and optical properties. [25,26]The uneven coating and unwanted filling of microchannels by the interfacing materials are further obstacles commonly associated with such methods. [27]ith the recent emergence of microphysiological systems (MPS) that are based on compartmentalized perfused devices of often hybrid nature, [28,29] the need for appropriate back-end processing techniques has been brought to the spotlight.In such settings, device surfaces and bonding materials should not manifest a high level of monomer leaching, nor any toxicity toward different organotypic cell models. [30,31]Moreover, they should ideally exhibit minimal drug absorption to enable accurate pharmacological and toxicological studies. [32]To that end, the biological repercussions of deploying surface modifiers and chemical bonding aids in manufacturing of MPS devices need to be thoroughly investigated.
Microfluidic materials such as polymers and glass further offer different properties in terms of optical transparency, solvent resistance, polarity, gas permeability, biocompatibility and ease of manufacturing which can each be leveraged independently or combined into the targeted applications including hydrophilic-in-hydrophobic microwell sensors, [33] thermoplastic porous membrane integration, [34] protein analysis, [35] 3D cell culture [36] and multi-layer organ-on-a-chip devices. [37]ere, we use thiol-ene-epoxy [38] chemistry to provide direct, robust and seamless integration of hybrid glass and polymer microfluidic devices.We demonstrate that various thermoplastic materials commonly used in microfluidic settings (PMMA, PET, PS and PC) can be effectively bonded to thiol-ene-epoxy without additional surface modification.Furthermore, the hybrid of directly assembled polymer-glass devices shows a bonding strength up to 1.5 MPa, by far exceeding the operational requirements of microfluidic devices.We demonstrate the fabrication and bonding of multi-layered microfluidic devices embedded with complex sets of chambers and channels of different aspect ratios.Furthermore, a commercial acrylate-based surface modifier developed for hybridization of thiol-ene significantly enhances bonding strengths of most devices and allows for the fabrication of thiol-ene-PEEK, -PP and -COC hybrids.The unique hydraulic resistance configuration of the microfluidic design enables us to characterize and compare the impact of both direct and indirect bonding on the integrity of micro-replicated features while subjecting to flow.We also examine the compatibility of different microfluidic device fabrication methods including CNC micromilling, casting and reaction injection molding, resulting in different surface roughness, with the direct integration method.We then fully characterize the sealing properties of the fabricated and integrated chips.Finally, we investigate the implications of different bonding approaches using both native and primed surfaces for cell and drug studies.

Thiol-Ene-Epoxy-Based Reaction Enables Strong Direct Bonding to a Wide Range of Materials
We first characterized the bonding strength of pristine thiol-eneepoxy directly bonded to glass and thermoplastic materials using the pull-off test method.In direct bonding, functional groups on the OSTE+ surface enable the integration of the material without the use of chemical additives.The stable operation of microfluidic devices in mild conditions typically requires pressures below 200 kPa. [39,40]Notably, the bonding strengths of pristine OSTE+ to PET, PMMA, PS, PC and glass exceeded this threshold with the latter being the strongest (Figure 1).For coupling to COC, PEEK and especially PP, bonding strengths to pristine OSTE+ were too weak and further surface modification protocols are required to generate such hybrid microfluidic devices.

Priming Surfaces Can Significantly Increase the Bonding Strength
To enhance bonding strengths, OSTE+ surfaces were treated with the acrylate-based, UV-activated surface modifier MCAP 212.Such surface modification imparts strong bonding through producing a thin layer (≈20 μm) that is anchored covalently onto OSTE+ while retaining acrylate groups of primer on the surface (Figure 2).Surface priming resulted in a material-specific modification of bonding strengths.Enhancements were particularly significant for PEEK (bonding strength to primed OSTE+ = 1 MPa, seven-fold increase compared to pristine OSTE+), PP (0.2 MPa, six-fold increase) and COC (0.5 MPa, 2.5-fold increase).While the pristine bonding on PET and PMMA already met the requirements, adding primer increased bonding strength by two-fold.For PS, PC and glass, however, surface modification did not significantly impact bonding strengths.
We further examined the long-term stability of the devices, approximately stored for 12 months under wet conditions at room temperature, using the blister test method.The majority of devices demonstrated a strong bonding exceeding the limit of the experiments (5 bar) with the exception of bonding of pristine OSTE+ to PEEK, PP, and COC as expected.Also, the bonding of primed OSTE+ to PP did not endure the test.

XPS Characterization Confirms the Availability of Surface Functional Groups
The X-ray photo electron spectroscopy (XPS) measurements were performed to investigate the functionality of the surfaces for both pristine and primed devices.The Figure S1a-c (Supporting Information) demonstrates the raw and fitted C 1S spectra of different samples including pristine OSTE+ after the first curing process, primed fully-cured OSTE+ before UV exposure of the primer and primed fully-cured OSTE+ after UV exposure of the primer.In Figure S1a (Supporting Information), the C─S and O─C═O peaks could be attributed to the presence of the thiol functional groups and the C─O peak to the epoxide groups on the surface of pristine OSTE+.In Figure S1b-c (Supporting Information), peaks assigned to O─C═O and C─O might indicate the presence of active acrylate groups [41] on the surface of the primer while C═C presumed to be related to the ene functional groups.It is worth noting that the lack of a notable reduction in C═C ratio after UV exposure is expected based on the fact that the depletion of the ene groups through thiol-ene reaction occurs adjacent to the OSTE+ surface situated at the bottom of the thick micron-sized layer of the primer which is far beyond the XPS reach.The data are normalized to the equivalent measurements for pristine surfaces and therefore demonstrate the fold change in bonding strength of primed versus pristine samples.Columns in red show materials with significant enhancement in bonding.OSTE+: off-stoichiometry thiol-ene-epoxy, PMMA: poly(methyl methacrylate), PC: polycarbonate, PP: polypropylene, PEEK: polyether ether ketone, PS: polystyrene, PET: polyethylene terephthalate, COC: cyclic olefin copolymer.

The Bonding Method Can Crucially Impact the Function of Microfluidic Devices
OSTE+ polymers entail a well-controlled two-step polymerization process in which the intermediate time window can be effectively used for back-end processing.This can be of significance in the microfluidic field for which bonding different parts is often cumbersome.Here, we utilized UV-based micro reaction injection molding (μRIM) previously developed for fabrication of OSTE+ microfluidic and organ-on-a-chip devices. [42,43]Separate micro-precision milled aluminum molds were designed: Microstructured mold to replicate the main fluidic features and the tube connector mold for world-to-chip interfacing (Figure 3a).OSTE+ polymer precursor was injected into the molds, followed by UV curing and demolding of the replicas.Subsequently, the tube connectors were directly bonded to the microfluidic layers.To realize hybrid OSTE+ devices, we chose different substrate materials (glass, thermoplastics and OSTE+) to enclose the fluidic cavity.By bonding the substrates to microfluidic layers featuring pristine and primed bottom surfaces, we then assembled all the layers together (Figure 3b) via both direct and indirect bonding proceeded by a heat-assisted polymerization process.Although we expect a lower coefficient of thermal expansion in thiol-ene-epoxy systems due to their thermosetting nature compared to those of thermoplastics, we did not observe any noticeable warpage next to the thermal polymerization process as demonstrated in the Figure S2 (Supporting Information).This could be due to the fact that the second polymerization step of OSTE+ continues after the heat treatment window and therefore the crosslinking network has time to readjust and release the thermally induced stresses at room temperature until it is fully polymerized.
The bonding step in back-end processing could lead to certain complications including collapse and blockage of microstruc-tures due to the excess of bonding pressure and unintended flow of adhesive materials respectively.Such problems might disqualify microfluidic chips, especially when microchannels are affected.One way to verify the intactness of such microchannels is to examine the cross-sectional slices of the chip under a microscope next to the bonding process.High-quality crosssectional slices, especially for polymer and glass-based devices, require cumbersome processing due to the lack of monocrystalline properties of the materials. [44,45]In addition, invasive cutting or milling procedures can lead to deformation or destruction of delicate microchannels, possibly leading to misinterpretation of processing artifacts.To circumvent these problems, we designed a microfluidic device to investigate the post-bonding quality of targeted microchannels (Figure 4a).We utilized the fact that parallel microchannels make the microfluidic network prone to bubble entrapment, which, in turn, can be used to amplify the effect of clogging or collapse in microchannels.To achieve this, we devised four groups of microchannels (C 1j …C 4j ) with heights ranging from 100 to 400 μm, each comprising four microchannels (C i1 …C i4 ) with widths ranging from 100 to 400 μm respectively (indices i and j represent group and channel numbers cycling from 1 to 4).The hydraulic resistances of the microchannels (R ij ) were approximated based on Equations (1-3): [46] R ij = 12L wh 3 ( 1 − 0.63 where μ, L, w and h represent the dynamic viscosity of the fluid, length, width and height of the microchannels, respectively.As shown in the equivalent circuit (Figure 4b), the resistances were configured to satisfy the relations stated in Equations ( 4) and ( 5): 3 Which means that resistances are equal in each group.Furthermore, the equivalent resistance of the series of groups 2 and 3 in the upper branch is equal to the overall resistance of the series of groups 1 and 4 in the lower branch.This ensures the simultaneous and even advection of the flow through the channels.Importantly, any deviation from the designed hydraulic resistances favors the entrapment of bubbles, thereby acting as a sensor for structural defectivities.
We then tested the theoretical calculations by infusing pristine and primed microfluidic devices with a colored dye (Figure 4c,d).Notably, primed devices are subject to the formation of bubbles in proximity to the critical microfluidic features which were intentionally designed to highlight such malfunctions.In contrast, the microchannels in pristine devices show a smooth flow without bubble entrapment.Although the cross-sectional images show that microstructures in pristine device (see Figure 4e) are less affected by defectivity compared to the primed device (see Figure 4f), this analysis is limited to a section of the microchannels and reveals little regarding the whole network of structures in comparison to what can be achieved through the bubble entrapment test.
While none of the microstructures with different widths and heights in pristine devices show signs of structural defectivity and remain intact, primed microchannels exhibit different levels of clogging depending on their width and height (Figure 4g,h).The width of microchannels appears to be the determinant variable affecting their structural integrity as the narrowest channels (w = 100 μm) show the highest level of defectivity irrespective of their height, while the widest channels (w = 400 μm) remain unaffected.The impact of channel height is more pronounced for channels with 100 μm < w < 400 μm where the probability of blockage increases with the reduction of height.

Hybrid Microfluidic Devices Show Moderate To Excellent Level of Sealing Properties
As primed devices are not expected to be significantly prone to leakage, we mainly focused the experiments on pristine-based hybriddevices which were designed to evulate the sealing properties The equivalent electrical circuit of the microfluidic design.c) Macroscopic images of a realized pristine device, injected with dyed water for bubble entrapment evaluation; the magnified images show smooth and bubble-free flow in sensitive regions.d) The equivalent images for a primed device; bubble entrapments are highlighted in the counterpart sensitive regions.Cutting line (above) and the corresponding cross-sectional images (below) for pristine e) and primed f) devices.g,h) Assessment of feature integrity through bubble entrapment analysis for pristine and primed devices, respectively.Note that microchannels with varying widths and heights demonstrate different degrees of blockage.across microbarriers ranging from 250 to 1000 μm.As demonstrated in Figure 5, OSTE+/OSTE+ and PC hybrid demonstrate an excellent sealing property with zero leakage across all micro-barriers.COC and glass hybrids show very good sealing with minor leakage events.PMMA hybrid also exhibits a good sealing property, and lastly, PEEK and PS manifest more frequent leakages through various width sizes.

Cell Viability and Drug Absorption Can Vary Depending on Material and Bonding Mechanism
One of the most common applications of microfluidic devices is their use for pharmacological and toxicological studies.Importantly, PDMS, the most widely used material for devices in preclinical drug testing [47] , is prone to high absorption of hydrophobic small molecules, which can have major impacts on drug response assays. [48,49]To assess the suitability of OSTE+ for such applications, we thus first benchmarked small molecule absorption of OSTE+ to PDMS.Consistent with previous observations, PDMS strongly absorbed the hydrophobic test compound with less than 2% remaining in solution after 2 h of incubation (Figure 6a).Plasma treatment had only marginal effects, increasing the amount of free compound after 2 h from 1.3% to 3%.After 24 h, absorption was so extensive that compound levels were below the limit of quantification in both untreated and plasmatreated PDMS.Strikingly, we find that absorption is 75-fold and 11-fold lower in pristine OSTE+ compared to PDMS after 2 and 24 h, respectively (Figure 6b).However, upon priming, absorption drastically increases and <1% of the compound remains free in solution as early as 2 h after incubation.To evaluate whether these effects are invariant for different OSTE formulations, we also measured absorption in pristine and primed OSTE+flex (Figure 6c).Importantly, absorption in pristine OSTE+flex was negligible throughout the 24 h incubation time (<3%).While priming again increased the loss of free compound, absorption of primed OSTE+flex remained > 4-to 23-fold lower than in primed OSTE+; this could be explained by the higher sorption of the primer itself into the porous network of OSTE+flex leading to the formation of a thinner layer of primer, as the main drug absorber, on the surface.
To constitute a useful material for biological assays, we tested the overall biocompatibility of the different materials (Figure 6d,e).As expected, PDMS did not exert cytotoxicity, neither in its pristine form, nor upon plasma treatment.Similarly, cells attached well on OSTE+ surfaces, forming a confluent monolayer without any noticeable decrease in viability.Upon priming, however, viability dropped significantly and only a few alive cells were found to be attached.On pristine OSTE+flex surfaces, mild cytotoxicity could be detected whereas all cells died upon priming.These results suggest that pristine thiol-ene-epoxy systems provide good biocompatibility and reduced drug absorption compared to conventional PDMS, rendering them useful platform materials for pharmacological and toxicological assessments.

Discussion
Back-end processing is known to be a labor-intensive and costly step in microfluidic device manufacturing.While the emergence of 3D printing-based fabrication techniques initially created many hopes to reduce complexities regarding the assembly of different microfluidic layers, the incompatibility of printable resins for biological applications, [50] monomer leaching, [51] high surface roughness, low feature resolution [52] and low level of optical transparency [53] have impeded the wide-spread dissemination of such methods for direct device manufacturing.Despite the fast growth of 3D printing technologies, the majority of applications are only compatible with small-scale prototyping or fabrication of test molds for PDMS soft lithography. [54]Therefore, robust bonding and integration of microfluidic devices have remained to be a major frontier in the field.While countless direct and indirect bonding methods have been developed to facilitate the process, these strategies often come with tradeoffs.Importantly, feature distortion and microchannel blockage are some of the common consequences of these methods.Off-stoichiometry thiol-ene-epoxy-based materials with high temporal control over the polymerization steps [38,55] hold promises to facilitate robust and heterogeneous bonding.
Initial reports using such a platform have shown its potential for silicon wafer [56] and porous membrane dry bonding. [57]Here, we integrated OSTE+ materials, both with pristine and primed surfaces to a broad range of thermoplastic materials as well as glass for the first time.After the first polymerization step, pristine OSTE+ polymers offer reactive functional groups including thiol and epoxide on the surface that enable strong bonding to glass and the majority of thermoplastics including PMMA, PC, PET, and PS, thus alleviating the need for additional surface modification.This potentially ensures that surfaces remain uniform and intact when subject to the bonding process.The results of the XPS measurements also confirm the availability of both thiol and epoxide functional groups on the surface of the pristine OSTE+ which is also in agreement with data obtained by Fourier transform infrared spectroscopy (FTIR) previously demonstrated. [38]owever, in the case of materials with low surface energy such as PEEK, PP, and COC, robust bonding is not readily achievable.We therefore evaluated the adaptability of both physical and chemical surface modification approaches to improve bonding.
Different physical and chemical surface modification methods have been developed and utilized to promote the bonding strength of microfluidic devices.Here we show that applying MCAP 212 facilitates bonding of OSTE+ with most materials which renders such indirect bonding methods suitable for potential high-pressure applications such as microreactor chips [58] and liquid chromatography-based devices. [59]The availability of active acrylate functional groups on the surface of the primer layer to feature such bonding properties is also substantiated by XPS data.
The impact of such surface priming on the chip functionality is not always well investigated.The stereotypic characterization methods in terms of device integrity such as cross-sectional imaging are limited to specific parts and features and therefore not thoroughly informative.Here, we created a special microfluidic design to examine the whole microfluidic network of various microchannels and chambers to benchmark the structural integrity of pristine versus primed devices.Whereas pristine-based microfluidic devices exhibit a full degree of microfluidic functionality, the performance of primed devices varied based on the proximity of the flow to defect-prone structures.We speculate that the recurrent clogging among narrower microchannels is mostly a ramification of the capillary filling during the primer application.As the primer was applied through a swift and low-pressure contact, we reason that its pressure-driven flow into the microchannels throughout the process is negligible.Although the contact pressure and duration of the subsequent bonding process are notable, we also do not expect any significant pressure-driven flow of the primer nor any capillary filling into the microchanbecause of the high viscosity of the primer by the time of bonding as a main consequence of its UV-polymerization.This lack of primer flow into the wide structures both during the application and the bonding process could be the cause that wider microchannels show more immunity to a high degree of blockage regardless of their height.Overall, the results indicate that while the primed devices display robust bonding, they are more prone to defectivity and malfunction compared to pristine thiolene-epoxy devices.
Our sealing characterization shows different degrees of leaktightness for hybrid-pristine-based devices.We intentionally designed our leakage test chips with dense structures separated by thin micro-barriers rendering them more prone to leakage.Despite this, we overall obtained zero to medium levels of leakage in the integrated devices.The unintuitive and arbitrary nature of the results in terms of barrier thicknesses, especially for the PEEK, PS, and PMMA hybrids can be attributed to a number of uncontrollable variables such as surface roughness of the manu-factured thermoplastic sheets, different material pliability, micro milling residual stress and the resulting surface deformations.Nevertheless, OSTE+/OSTE+, PC, COC, and glass hybrids, in their pristine forms, are robust enough to show excellent sealing properties regardless of the noted intruding factors.
Absorption of small molecules constitutes an important limitation for pharmacological and toxicological in vitro studies that can contribute toward the translational gap in preclinical research.[62][63] Absorption is primarily caused by hydrophobic interactions and pore-filling effects. [64]The reduced loss of compound in OSTE+ compared to PDMS could be expected, given the high porosity and hydrophobicity of PDMS.Despite the low crosslinking density of OSTE+flex, [65] it exhibited lower drug absorption than OSTE+ owing to its different chemical composition; the flex product contains a polyether thiol which is not present in OSTE+ and a cycloaliphatic epoxy as opposed to the aromatic epoxy resin in OSTE+.This could possibly result in higher polarity/polar group density for the flex material leading to a higher absorption of water instead of the hydrophobic drug.Priming drastically increased the drug absorption, possibly due to the aliphatic structures in the primer network which are inherently prone to absorb hydrophobic chemicals.Importantly, pristine OSTE+flex showed no detectable absorption of hydrophobic chlorpromazine (>100 fold lower than PDMS), but albeit minor, exhibited some cytotoxicity after 24 h.Thus, pristine OSTE+flex might be best suited for short-term pharmacological cell assays or the fabrication of devices or parts that are in contact with a medium, but not directly with cells.The primed devices evidently suffer from cytotoxicity due to the plausible leaching of uncured monomers into the biological samples.In contrast, pristine OSTE+ was highly biocompatible without evident cytotoxicity in agreement with previous results [43,66] and resulted in detectable compound absorption that was however >11-fold lower than with conventional PDMS, suggesting that this material is suitable for long-term pharmacological and toxicological cell-based applications, as it significantly reduces absorption without impacting viability.
Combined, these results indicate that thiol-ene-epoxy systems can be directly bonded to a wide spectrum of thermoplastic materials as well as glass, resulting in the facile manufacturing of robust hybrid devices with minimal defectivity.By combining their ease of fabrication with high biocompatibility and significantly reduced drug absorption, thiol-ene-epoxy provides a useful tool kit for the fabrication of MPS devices in preclinical drug development.Indirect bonding using primed OSTE reduces feature integrity and biocompatibility of microfluidic devices but further increases bonding strength, which can be useful for cell-free, high-pressure applications.

Experimental Section
Materials: OSTEMER 322 (henceforth referred to as OSTE+) or OS-TEMER 324 (henceforth referred to as OSTE+flex) as well as the primer MCAP 212 was acquired from Mercene Labs AB (Stockholm, Sweden).OSTE+ contains a multifunctional aromatic epoxy resin, a tri-functional allyl, an aliphatic thiol and initiators with an approximate viscosity of 1000 mPa.s.OSTE+flex contains an aliphatic epoxy resin, a tri-functional allyl, an aliphatic thiol and initiators with an approximate viscosity of 1800 mPa.s.MCAP 212 contains liquid, non-volatile monomers and oligomers with two or more unsaturations and a UV initiator with an approximate viscosity of 100 mPa.s.
Sample Preparation: Pull-off measurements on OSTE+/thermoplastic interfaces to characterize their bonding strength were performed.Each substrate material was milled and cut into squares of 50 × 50 mm.OSTE+ precursor was cast into aluminum molds of 20 × 50 × 2 mm; the first stage of polymerization was performed using a 365 nm UV lamp (Firefly FF200, PhoseonTM Technology, USA) with an output power of 1.5 w/cm 2 for a duration of 25 s.The resulting flat replicas were then demolded and used either as pristine for which the native surface of the material remained intact after UV-polymerization step or as primed for which the material underwent additional surface modification via application of the primer, followed by UV exposure with similar previously stated settings.For glass, priming was implemented through oxygen plasma treatment due to the compatibility of glass with this surface modification approach.Substrate bonding was implemented using both pristine and primed surfaces with a sample size of four per material.Paper clamps in conjunction with Teflon vices were used to ensure a mild and even contact between OSTE+ and other substrates.The second stage of OSTE+ polymerization was performed at 90 °C for 2 h.
Pull-Off Test: The samples were stored at room temperature for 5 min prior to being fixed onto 20 mm diameter aluminum dollies through applying an epoxy glue (Epoxylim Casco Strong Epoxy Professional).The sample-mounted dollies were left at room temperature for 12 h to allow the glue to harden prior to pull-off measurements.The bonding strength measurements were performed using a pull-off adhesion tester (Elcometer 506, UK) by pulling the dollies until the breaking point of the sample layers.
X-Ray Photoelectron Spectroscopy Analysis: An X-ray photoelectron spectroscopy (XPS) was performed on several samples including pristine OSTE+ polymer after the first curing process, primed fully cured OSTE+ prior to UV exposure of the primer and primed fully-cured OSTE+ post UV exposure of the primer.XPS measurements were performed in an ultrahigh vacuum (<1 × 10 −8 Torr) using a monochromated (h = 1486.69eV) Al K X-Ray source (Kratos Axis Supra+) at a power of 225 W with charge neutralization.Data was fitted using ESCape software (Kratos) with standard background correction and multi-component curve-fitting settings.The binding energies were calibrated to Carbon 1s peak at 285.0 eV.
Microfluidic Device Fabrication and Characterization: Molding approach (casting and reaction injection molding): aluminum molds were micromilled using a micromilling machine (Minitech Machinery Corp., USA).Micro reaction injection molding [42] was then performed to produce 75 × 25 mm OSTE+ microfluidic layers.The same process was used for the fabrication of OSTE+ tube connectors.Similarly, the microfluidic parts were also fabricated using the casting technique by pouring thermosetting resin inside the open mold cavity, bubble removal, subsequent UV curing and demolding process.To assemble the hybrid microfluidic devices, similar to the pull-off samples, the fluidic layers were enclosed through bonding a series of substrates (OSTE+, thermoplastics and glass) and their ports were directly bonded to the tube connectors.The assembled devices underwent thermal curing using the same temperature setting as used during the pull-off tests.The resulting devices feature different mmsized chambers and μm-sized channels with sectional dimensions down to 100 × 100 μm.To drive the microfluidic devices, dyed water was infused with a syringe pump set at a constant flow rate of 0.08 ml/min.The devices were also subject to milling for further cross-sectional microscopic imaging.
Direct CNC micromilling approach: Thermoplastic polymer substrates including PMMA, PC, COC, PS, PEEK and PET were subject to micromilling for direct fabrication of microfluidic parts in the microscopy glass slide format (75 × 25 mm).Except for PET which did not endure the heat during the milling process, the rest of the thermoplastic materials were successfully microstructured.
Characterization of Device Sealing: Leakage from microchannels was investigated through specially designed microfluidic chips featuring bonded barriers with various widths of 250, 350, 500, 700, and 1000 μm.For each hybrid material, nine samples per barrier size were used to quantify the leakage.
Blister Test: To evaluate the long-term stability of bonding, the microfluidic devices were subjcted to blister test by blocking their outlets, connecting their inlets to pressurized air and submerging them in water in order to monitor air bubbles leaking from the devices as a measure of bonding failure.The pressure source could apply the maximum pressure of 5 bars into the devices.
Drug Absorption Measurement: Chlorpromazine (CAS no.69-09-0) was prepared in DPBS pH 7.4 at a concentration of 1 μMm (with 0.1% DMSO).60 μL of the solution was incubated in either OSTE+, OSTE+flex, or PDMS chips with identical surface-to-volume ratios.After 24 h of incubation at room temperature, 10 μL of the solution was mixed with 20 μL of chilled stop solutions comprising of 99% acetonitrile, 1% formic acid, and 0.03% (v/v) of 10 mm pruvanserine (Merck, Darmstadt, Germany).10 μL of working solution was sampled at the start as a baseline.The samples were then analyzed by liquid chromatography with tandem mass spectrometry (LC-MS/MS) as previously. [63]iocompatibility Testing: HEK293A cells (passage 10) were grown until confluence in Dulbecco's Modified Eagle Medium supplemented with 100 U mL −1 penicillin, 100 μg mL −1 streptomycin, and 10% fetal bovine serum (FBS, HyClone).For ATP measurements, cells were seeded at a seeding density of 12000 cells/well in serum-starved (1% FBS) media.After 24 h, cellular ATP was measured using the CellTiter Glo Luminescence ATP reagent (Promega) according to the manufacturer's instruction with addition of 0.1% SDS to facilitate cell lysis.For live/dead staining, slabs of untreated polydimethylsiloxane (PDMS), plasma-treated PDMS, pristine or primed OSTE+ or OSTE+flex were placed in six-well plates.After sanitation, the slabs were coated with 50 μg mL −1 poly-L-lysine for 1 h at room temperature (RT), then dried for 1 h and equilibrated with culture media prior to cell seeding.After overnight incubation at 37 °C and 5% CO2, cell viability was assessed with LIVE/DEADTM Viability/Cytotoxicity Kit (Thermofisher) on a LSM 880 confocal microscope (Zeiss).

Figure 1 .
Figure 1.Investigation of direct bonding mechanism between pristine OSTE+ strip and polymer/glass substrates via pull-off test.a) OSTE+ strip featuring thiol and epoxide reactive groups and glass/thermoplastic substrate including PEEK, PC, PET, COC, PMMA, and PS prior to bonding.b) Bonded glass-OSTE+ surfaces with associated bonding mechanism (before and after bonding inside inserts).c) Mounting dollies onto OSTE+ strips.d) Pull-off force measurement.e) Breaking point and separation of the bonded surfaces.f) Bonding strength measurements of OSTE+ to different materials; the dashed line shows the average pressure required for common microfluidic devices.OSTE+: off-stoichiometry thiol-ene-epoxy, PMMA: poly(methyl methacrylate), PC: polycarbonate, PP: polypropylene, PEEK: polyether ether ketone, PS: polystyrene, PET: polyethylene terephthalate, COC: cyclic olefin copolymer.

Figure 2 .
Figure2.Indirect bonding using primed OSTE+ surfaces.a) Results of pull-off measurements using primed OSTE+ surfaces bonded to similar thermoplastic/glass substrates.b) The data are normalized to the equivalent measurements for pristine surfaces and therefore demonstrate the fold change in bonding strength of primed versus pristine samples.Columns in red show materials with significant enhancement in bonding.OSTE+: off-stoichiometry thiol-ene-epoxy, PMMA: poly(methyl methacrylate), PC: polycarbonate, PP: polypropylene, PEEK: polyether ether ketone, PS: polystyrene, PET: polyethylene terephthalate, COC: cyclic olefin copolymer.

Figure 3 .
Figure 3. Process flow of fabrication and integration of hybrid microfluidic devices using pristine and primed thiol-ene-epoxy systems.a) Fabrication of microfluidic layer and tube connector molds (left), micro reaction injection molding (μRIM) process including polymer precursor injection and the subsequent UV polymerization (middle), and demolded replicas of the microfluidic layer and tube connectors (right).b) Integration of different layers: direct bonding of tube connectors and direct/indirect bonding of the substrates followed by thermal exposure.

Figure 4 .
Figure 4. Assessment of feature integrity and fluidic function in both pristine and primed-based devices.a) Microfluidic design aimed to investigate the post-bonding qualities of the device using chambers and microchannels of different aspect ratios.b)The equivalent electrical circuit of the microfluidic design.c) Macroscopic images of a realized pristine device, injected with dyed water for bubble entrapment evaluation; the magnified images show smooth and bubble-free flow in sensitive regions.d) The equivalent images for a primed device; bubble entrapments are highlighted in the counterpart sensitive regions.Cutting line (above) and the corresponding cross-sectional images (below) for pristine e) and primed f) devices.g,h) Assessment of feature integrity through bubble entrapment analysis for pristine and primed devices, respectively.Note that microchannels with varying widths and heights demonstrate different degrees of blockage.

Figure 5 .
Figure 5.A radar graph presenting the sealing results of pristine-based hybrid devices across a variety of micro-barriers ranging from 250 to 1000 μm denoted as the radial axes.Data points depict the number of leakage events read by the gridlines 0 to 9.

Figure 6 .
Figure 6.OSTEs are superior compared to PDMS in terms of low drug absorption, but primer compromised their chemical inertness and biocompatibility.Amount of remaining compound (chlorpromazine; shown as % relative to t 0 ) after incubation in PDMS a), OSTE+ b), and OSTE+flex devices (c; n = 3 independent devices for each group).The average amount of drug absorbed in each group is indicated.d) Viability measurements of HEK293A cells cultured on PDMS, OSTE+ or OSTE+flex show significant toxicity upon priming (n = 6 replicates).Data are shown as mean ± s.e.m. *** p < 0.001, **** p < 0.0001 (two-tailed unpaired heteroscedastic t-test).e) Representative images from live/dead staining on PDMS and OSTE+ surfaces.Scale bar = 100 μm.