A Modular Microfluidic Organoid Platform Using LEGO‐Like Bricks

Abstract

The convergence of organoid and organ‐on‐a‐chip (OoC) technologies is urgently needed to overcome limitations of current 3D in vitro models. However, integrating organoids in standard OoCs faces several technical challenges, as it is typically laborious, lacks flexibility, and often results in even more complex and less‐efficient cell culture protocols. Therefore, specifically adapted and more flexible microfluidic platforms need to be developed to facilitate the incorporation of complex 3D in vitro models. Here, a modular, tubeless fluidic circuit board (FCB) coupled with reversibly sealed cell culture bricks for dynamic culture of embryonic stem cell‐derived thyroid follicles is developed. The FCB is fabricated by milling channels in a polycarbonate (PC) plate followed by thermal bonding against another PC plate. LEGO‐like fluidic interconnectors allow plug‐and‐play connection between a variety of cell culture bricks and the FCB. Lock‐and‐play clamps are integrated in the organoid brick to enable easy (un)loading of organoids. A multiplexed perfusion experiment is conducted with six FCBs, where thyroid organoids are transferred on‐chip within minutes and cultured up to 10 d without losing their structure and functionality, thus validating this system as a flexible, easy‐to‐use platform, capable of synergistically combining organoids with advanced microfluidic platforms.

A novel fluidic circuit board is presented to facilitate the integration of complex 3D in vitro models on chip. The flexible platform consists of embedded microfluidic channels with innovative LEGO‐like fluidic interconnectors for plug‐and‐play connection with various cell culture bricks. The developed platform is validated as a flexible, user‐friendly system, capable of synergistically combining thyroid organoids with advanced microfluidic platforms.

1. Introduction

Organoid and organ‐on‐a‐chip (OoC) technologies have found tremendous promise as two distinct strategies to create human surrogates for a wide range of biomedical applications, from precision medicine and disease modeling to more predictive drug screening platforms.^{[ 1} , ² , ³ , ^{4 ]} Despite emerging as two separate strategies, the combination of organoids and OoC devices has been recently proposed and shown to be mutually beneficial.^{[ 5} , ^{6 ]} When incorporated in microfluidic cell culture devices, organoids can substantially increase the physiological relevance, while technical devices can help to overcome diffusion limitations and to control the size and shape of organoids.^{[ 1 ]} Moreover, microfluidic chips can be designed to provide spatially and temporally resolved biochemical and biophysical cues to further promote organoid maturation.^{[ 5} , ^{7 ]} Therefore, the synergistic combination of organoids with microfluidic systems is highly desired among researchers and industrial stakeholders.^{[ 5 ]} However, pioneering reports have shown that integrating large functional organoids featuring a more complex 3D architecture inside chip devices is a difficult endeavor as it further complicates the handling of microfluidic systems.^{[ 1} , ^{6 ]}

Conventional microfluidic devices still pose substantial challenges for organoid studies, in part owing to their difficult operability and lack of flexibility.^{[ 8} , ^{9 ]} Traditional OoC devices are typically laborious to set up, often requiring skilled personnel to integrate multiple external hardware, such as precision pumps, adaptors, tubing, and reservoirs. Consequently, microfluidic experimentation is normally associated with high failure rates and loss of cells during operation. This is especially inconvenient for organoid cultures, as they are considered a time‐consuming model typically generated from expensive and/or rare cells that one cannot afford to lose. While few microfluidic platforms have been successful in rapidly integrating organoids,^{[ 1} , ^{10 ]} these devices are difficult to develop and tailored for a specific application or tissue, which impede flexible alterations to the design.

Over the past decade, multiple attempts have been made to reduce operation complexity and increase versatility of microfluidic cell cultures. Modular microfluidic systems have been at the forefront of these efforts. Modular systems are created either by attaching fluidic blocks/modules to each other or by directly plugging them into a fluidic breadboard. For example, Ong et al.^{[ 8 ]} cultured multiorgan models, including 3D liver spheroids, in a planar perfusion system, which was created by connecting Tetris‐like fluidic modules via self‐aligning magnets. Similar planar systems have been reported by using 3D printed fluidic modules laterally connected by metal pins^{[ 11 ]} or by using casted polydimethylsiloxane (PDMS) chips fluidically connected by puzzle‐like studs.^{[ 12 ]} Alternatively and analogous to a printed circuit board, cell culture modules can be plugged into and connected with a universal manifold, which has been termed fluidic circuit board (FCB). A good example was given by Vollertsen et al.,^{[ 9 ]} who developed a FCB by thermal compression bonding of five polystyrene (PS) layers bonded with a styrene‐ethylene‐butylene‐styrene (SEBS) layer. The FCB connected multiple modules to obtain massive parallelization while providing design flexibility. This platform was tested by culturing human umbilical vein endothelial cells (HUVEC) for up to 5 d. Similarly, Vivas et al.^{[ 13 ]} created a thermally bonded polymethyl methacrylate (PMMA) FCB, which allowed reduction of the operation complexity when culturing fibrin‐embedded cardiomyocytes.

Although these modular systems can be quickly assembled and reconfigured, they use OoC modules that are often designed in a closed cell chamber layout and are irreversibly bonded to create a reliable, leak‐tight seal.^{[ 14 ]} By doing so, there is no direct access to the cell chamber and cell (un‐)loading is performed via enclosed, narrow channels. Advanced culture protocols to form organoids from single cells or cell suspensions make it difficult to carry them out within microfluidic chips, and transferring precultured and matured organoids is even more challenging, because they are usually too big or too sensitive against shear forces acting on them in the narrow channels.^{[ 1} , ¹⁵ , ¹⁶ , ^{17 ]} In addition, upscaled irreversibly bonded platforms might not be compatible with some extracellular matrix (ECM)‐like gels, such as Matrigel, which quickly solidifies and is difficult to flow through narrow microfluidic channels.^{[ 1 ]} Alternatively, OoC devices can be reversibly sealed via fasteners to facilitate direct incorporation and harvesting of large 3D cell models, such as brain organoids and 3D printed constructs.^{[ 1} , ^{18 ]} In these devices, precultured organoids can be easily transferred in/off chip using standard and even automated liquid handling tools, hence avoiding any cell loss during operations. Moreover, reversibly sealed devices allow easy loading of ECM‐like gels into the cell chambers through manual pipetting and straightforward protocols. Similarly, cells are often more efficiently harvested from reversibly sealed devices than those isolated using in‐channel techniques.^{[ 19 ]} This offers a great opportunity to perform multiple downstream analysis, including RNA sequencing and proteomics to examine comprehensively the tissue construct.^{[ 20 ]} Among different fasteners, lock‐and‐play (LnP) screws allow reversible tightening of a chip device in a single step and have found success to rapidly incorporate several advanced 3D in vitro models on chip, such as tumor^{[ 1 ]} and cardiac organoid models.^{[ 6} , ²¹ , ^{22 ]} Despite their promise, the combination of reversibly sealed OoC devices with modular microfluidic platforms has been largely unexplored.

To address this challenge, we combined a new tubeless FCB with OoC modular bricks featuring a reversible LnP sealing for direct access to the cell chamber. The developed FCB allowed real plug‐and‐play connection of various cell culture bricks to the FCBs via LEGO‐inspired fluidic interfaces without the need of tubing and side fasteners, while the LnP sealing facilitated culture of mouse embryonic stem cells (mESC)‐derived thyroid organoids in the OoC bricks. This synergistic approach was capable to reduce the time and complexity of the manual assembly operations, which ultimately helped integrating functional organoids on chip. Importantly, thyroid organoids were found to preserve their structure and functionality on chip, thus validating this platform as a next step toward the development of flexible, in vitro perfusion platforms tailored for advanced organoid applications.

2. Results

2.1. Design

The microfluidic system is composed of a tubeless FCB to which LEGO‐like cell culture module bricks can be plugged via novel fluidic interconnections (Figure  1A and Video S1, Supporting Information). The FCB is 39 mm in width and 137 mm in length, and comprises two PC layers: a top layer containing the fluidic interfaces and a bottom plate featuring surface‐machined microfluidic channels. Thermal bonding between the two plates closes off the microfluidic channels. The FCB allows recirculation of cell culture medium between the different cell culture bricks and combines them into a single microfluidic platform. The cell culture bricks have a universal size of 38 × 38 mm, while the height varies from 15 to 30 mm depending on their functionality. Bottom plates with different layouts of microfluidic channels can be bonded to the same top plate, enabling reconfiguration of culture module bricks in multiple ways and thus increasing the versatility of the system (Figure 1B,C). Two exemplary systems are illustrated in Figure 1C. The first system comprises a FCB that fluidically connects an OoC brick to a cell culture medium reservoir brick. Between these modules, a previously developed, cost‐effective mini‐microscope can be plugged for real‐time imaging of the microfluidic channels.^{[ 23 ]} Alternatively, an imaging module could be integrated directly on the OoC brick to allow continuous monitoring of cell populations.^{[ 24 ]} The second platform combines two reservoir bricks with an OoC brick to create two independent fluid compartments. Other cell culture systems can be established and are illustrated in Figure S1 (Supporting Information). All FCBs are connected to an external peristaltic pump via standardized Luer type inlet‐outlet (I‐O) ports positioned at the top and bottom corners (Figure 1C). Up to six FCBs could be installed in parallel for cell culture experiments, covering a total volume of ≈1.3 L (158 × 280 × 30 mm) (Figure 1D,E). All components were fabricated in‐house and can be sterilized and reused for multiple cell culture experiments.

Figure 1.

A modular microfluidic platform for dynamic culture of organoid models. A) 3D design concept of the FCB and the cell culture bricks. Plug‐and‐play insertion of bricks in the FCB is made by LEGO‐inspired fluidic interconnectors. B) The FCB is made from a top PC layer (which features the fluidic interfaces) thermally bonded against a bottom layer with engraved channels. The bottom layer can be customized to create different fluidic pathways. C) Schematics of different fluidic pathways using distinct bottom FCB layers. Photographs of FCBs plugged with reservoir and OoC bricks and injected with a red and blue food dye. A porous, membranous carrier can be integrated in the OoC device for various studies, such as transepithelial transport and cell migration assays across biological barriers as well as for studies of gradients of bioactive factors (Video S2, Supporting Information). D,E) Photographs of six FCBs in parallel before and during cell culture experiments.

2.2. Fabrication of the FCB

Prototypes of the fluidic channels were machined in the bottom FCB layer by a desktop CNC micromilling machine. Multiple channel geometries were engraved with sizes ranging from 0.3 to 1.0 mm (Figure S2A, Supporting Information). The type of cutting tool determined the channels’ morphology. End mills were used to create rectangular microchannels, while ball nose mills and engraving tools were used to create round and triangular‐shaped channels, respectively (Figure  2A). These geometries are attractive because they can be compatible with the design of microfluidic valves using, for example, round channels.^{[ 25} , ^{26 ]} Topographic analysis demonstrated that channels were successfully machined in PC plates with dimensional accuracies of 0.64% and 5.47% for width and height, respectively (Figure 2B). The lower height accuracy is due to manual setting of the Z₀ during milling and can be easily mitigated by using a more sophisticated milling machine. Chatter marks were observed on the bottom side of channels, which impaired the surface quality of channel walls. Chattering is a recurrent issue of milling processes and is mainly caused by vibrations in the system and suboptimal milling parameters. As good surface quality is desired for some cell culture applications, particularly to avoid excessive drug/compound adsorption and to limit the formation of leak spots, we evaluated different milling parameters to reduce surface roughness. As shown in Figure 2C, higher spindle speeds and lower feed rates created channels with smoother surfaces (Ra ≈ 0.5 µm), while climb milling (down cut) did not give further improvements over conventional milling (up cut). Despite improving the surface quality, higher speeds and lower feeds promoted the formation of burrs at the channels upper corners (Figure 2D). Moreover, the amount of burrs increased as the channels got smaller (Figure S2B, Supporting Information). After bonding, albeit burrs did not cause leakage per se (Figure S3A, Supporting Information), they were a source of irregularities at the channel edges and could potentially become nucleation points for air bubble formation inside the fluidic pathway. Therefore, several methods were conducted to remove unwanted burrs (Figure S3B, Supporting Information). Among these, post‐engraving of the channel edges using a deburring tool was found to be the only method able to reliably eliminate burrs. As a result, we fabricated microfluidic channels with 1.0 × 0.5 mm (W × H) at speeds and feeds of 40 000 rpm and 200 mm s⁻¹ for optimal surface quality and post‐deburred 0.05 mm in depth for burr removal.

Figure 2.

Micromilling of fluidic channels and thermal bonding of PC–PC FCB plates. A) Photograph of a PC test sample with an engraved fluidic channel (W = 1 mm, H = 0.5 mm). On the right, optic laser images of cross sections of channels with different geometries (round, square, and triangular). Scale bar, 500 µm. B) Topography analysis of engraved channels. The dimensions of the engraved channels were quantified by an optical profilometer (Keyence VK‐X100K). C) Quantification of the surface roughness (Ra) for test slots milled at different spindle speeds and feed rates (mean ± SD). Data are shown from three milled replicas (n = 3), and each point represents the average of multiple line Ra measurements across the samples surface area. D) Optic photographs of engraved microchannels and burrs formation. Images were taken with a 10× magnification. Scale bar, 250 µm. E) Schematic illustration of the system apparatus for bonding the two FCB plates (Al, aluminum). F) Fluorescent images of enclosed channels perfused with a FITC dye and prebonded at different bonding temperatures. Profile lines of the fluorescence intensity (a.u., arbitrary units) are presented below for each bonding temperature. For each image, three lines were drawn perpendicular to the channels length and placed equidistant from each other. G) Photograph of an enclosed channel of a FCB sample. I) Stereomicroscope images of two cross sections from channels bonded at 120 and 140 °C. Scale bar, 500 µm. J) Quantification of height and width changes of channels bonded at different bonding temperatures. All data are tested for statistical significance using Kruskal–Wallis test. ns, nonsignificant; p > 0.5; *p < 0.1; **p < 0.01. Data are shown from three measurements across four channel cross sections from two independent bonded samples (n = 12).

After successful machining, the bottom FCB plate was thermally bonded against the upper FCB plate in a hydraulic press (Figure 2E). The bonding was mediated by a polymeric adhesive foil (≈50 µm in thickness), which was placed between both PC plates. There were two holes for locating pins per FCB for easy alignment of the different bonding layers. We first investigated the ideal bonding temperature, as excessive temperature can cause deformation of the channels, while insufficient bonding temperature may ultimately cause leakage.^{[ 27 ]} We bonded PC plates at temperatures ranging from 70 to 140 °C, considering that the maximum operating temperature of the PC plates is 120 °C and the T _g of the adhesive foil is 88 ± 4 °C. Subsequently, the bonded FCB plates were perfused with a FITC dye at 0.5 mL min⁻¹. As illustrated in Figure 2F, temperatures equal or higher than 120 °C were required to avoid leakage. To quantify the extent of the deformation of channels, samples bonded at increasing temperatures were cut, imaged by a stereomicroscope, and three width and height measurements were made across each channel cross section (Figure 2G). We observed no significant changes in height and width as bonding temperature increased until 120 °C, after which, significant drops of 3% and 13% were measured for width (p < 0.1) and height (p < 0.01), respectively (Figure 2I,J). Therefore, we bonded PC plates at 120 °C to provide leak‐tight operations while avoiding significant deformation of channel walls. The sealing strength of thermally bonded PC plates was determined to be above seven bars. In total, 14 FCBs were fabricated and bonded for cell experiments, of which ten passed visual inspection.

2.3. Plug‐and‐Play Fluidic Interconnections

The fluidic interconnections were achieved by mating circular recesses (bores) of the FCB with a four‐stud LEGO‐like bottom part (plugs) (Figure  3A and Video S1, Supporting Information). The studs and recesses were concentrically aligned and had central I‐O holes (Ø0.8–1.0 mm) to allow passage of liquid between the modular bricks and the FCB. The sizes and geometry of the I‐O ports can be adjusted so that capillary forces keep fluids inside the modular bricks once detached from the FCB (Video S3, Supporting Information).^{[ 28 ]} Because the PDMS fluidic layers are kept under compression between the thermoplastic bodies of a modular brick, the internal liquid flow due to (un‐)plugging is considered neglectable. A radial groove was created in each stud, where an O‐ring (cross section Ø1.0 mm and inner Ø4.0 mm) is mounted to form a seal upon contact with the FCB recess walls (Figure 3B). O‐ring grooves were prelubricated and the corners of recesses were chamfered (20°) to avoid damaging the seal. A diametrical clearance gap (S) is required to create a reliable and reversible seal between the modular bricks and the FCB, thus enabling rapid system assembly. Determining the adequate clearance and O‐ring compression (%) is crucial for this application, as excessive O‐ring compression can lead to difficult unplugging of cell culture bricks, whereas insufficient compression or high clearance gaps may cause experimental failure by leakage (Figure 3C). For initial design iterations, studs were made from biocompatible clear MED610 dental resin and fabricated by Polyjet photopolymer technology (similar to inkjet printing). Later, parts were machined from PC. Various O‐ring groove dimensions were pre‐selected based on empirical data from the O‐ring supplier handbook.^{[ 29 ]} The sealing performance of each design was characterized by the maximum internal pressure that the fluidic interface could withstand. In brief, the inlet of the stud was connected to a pressurized air tank from which air was supplied to the mated stud‐recess system while the recess outlet was closed to allow pressure to build up until the sealing fails. We determined an optimal groove depth and width of 0.72 mm and 1.35 mm, respectively (Figure S4, Supporting Information). As expected, the lowest clearance gap (0.2 mm) achieved the highest sealing strength of up to 72 ± 15 kPa for the 3D printed parts (Figure 3D), which corresponds to 20% compression of the O‐ring. Lower clearance gaps were not studied, as S = 0.2 mm was found to be the minimal value allowed by the system to assure easy operability and no permanent damage to the O‐ring. For this S value, cell culture bricks could be (un‐)plugged easily (Video S1, Supporting Information) and correctly (Video S4, Supporting Information), while no leakage of liquids was detected for flow rates up to 0.5 mL min⁻¹ (Figure 3E and Videos S5 and S6, Supporting Information).

Figure 3.

Development of LEGO‐like fluidic interconnections. A) Photograph of a modular brick bottom part containing four studs. Studs are equidistant to each other similar to those of commercially available LEGO bricks. B) 3D model of the fluidic interface. A modular brick is inserted into a circular recess of the FCB. An O‐ring is located radially on the stud. Flow connection is realized between the modular brick and the FCB via centric and aligned microchannels. C) Illustrations of O‐ring failure scenarios. A high clearance gap (S) and low compression over the O‐ring can lead to leakages, while excessive compression may cause O‐ring extrusion and damage. D) Quantification of the maximum burst pressure values as measure of the sealing strength of studs either made from 3D printed resin (MED610) or from machined PC (mean ± S.E.M.). E) Photograph of a modular brick plugged in a FCB and being perfused with a food dye at 0.5 mL min⁻¹. F) Percentage of successful plugging of O‐ring studs made from 3D printed material and machined PC. G) Photographs of the sidewalls and the O‐ring grooves with and without O‐rings for 3D printed and machined PC studs. Scale bar, 1.5 mm. H) 3D heat map of O‐ring groove surfaces for 3D printed and PC machined studs. J) Measurement of the surface finishing of 3D printed and machined studs (mean ± S.E.M.).

3D printing of studs allowed rapid prototyping of complex geometries at low costs with good dimensional accuracy (0.35%). However, once the design was stablished, we machined parts from PC to avoid leakage of toxic compounds to the cell culture medium, which was found to decrease viability of thyroid organoids in case of the 3D printed resin (Figure S5, Supporting Information). Interestingly, we observed that machined PC fluidic interfaces revealed higher sealing strengths than their 3D printer counterparts, yielding a maximum pressure burst of 111 ± 15 kPa for 20% O‐ring compression (Figure 3D). Moreover, PC parts were easier to plug into the recesses and without causing frequent squeezing of the O‐ring out of its groove (Figure 3F). Appropriate surface quality is known to be crucial for successful O‐ring seals. Therefore, we investigated the main differences between the surface quality of PC and 3D printed studs. Both devices exhibited surfaces of similar finishing (Ra ≈ 2 µm); however, grooves were extensively damaged for the 3D printed parts (Figure 3G,H). In O‐ring groove regions, milled PC samples showed ≈100‐fold smoother surfaces than their 3D printed counterparts (Figure 3G,I), which, together with the inherent blocking behavior of PC, could explain the higher sealing performance of machined parts.

2.4. Module Brick and LnP OoCs

Multiple LEGO‐like cell culture bricks were designed and fabricated to perform a variety of functionalities, including cell culture medium reservoirs, pH and oxygen sensor modules, and reversibly sealed OoCs (Figure  4A). The fluidic pathways of all LEGO‐like bricks were defined by a specific set of fluidic layers made by PDMS casting. This method allowed the fabrication of PDMS fluidic layers, which were transparent, easy to clean, and maintained their thickness after several cycles of sterilization (Figure S6, Supporting Information). These PDMS fluidic chips were stacked vertically inside the PC housing and sealed by mechanical clamps. The detailed designs of each module are illustrated in Figure S7 (Supporting Information). The cell culture bricks shared the same outer dimensions of 38 × 38 mm and the four‐stud connector layout for easy interfacing with the FCB.

Figure 4.

Cell culture modular bricks and development of a LnP chip device. A) 3D model views of an organoid‐on‐a‐chip, a cell culture medium reservoir module, a sensor module for continuous measurement of dissolved oxygen and pH values, and a mini‐microscope. B) 3D explosive model view of the LnP organoid‐on‐a‐chip device. The device consists of a PC housing (grey), fluidic chips (cyan), and a thermoformed membranous carrier (yellow). The carrier can be easily transferred in/out of the chip with tweezers. Brightfield picture of Matrigel‐embedded mouse ESC‐derived thyroid organoids cultured in the membranous carriers. Image taken with a 4× magnification. Scale bar, 250 µm. C) Photographs of the clamped versus unclamped organoid‐on‐a‐chip. D) 3D view of the fluidic pathway through the organoid‐on‐a‐chip device. E) Extracted still images from a time‐lapse recording of the filling of the LnP organoid‐on‐a‐chip with a passive bubble trap tank. The device was perfused with a blue food dye at 0.5 mL min⁻¹.

Recently, LnP clamps have been proposed as promising fasteners to facilitate quick and easy (un)loading of organoids into microfluidic OoCs.^{[ 22 ]} Here, we explored a single‐screw clamping to allow incorporation of mouse ESC‐derived thyroid follicles (Figure 4B). The device was composed of three PDMS fluidic layers stacked with a polymeric carrier and sealed by tightening a screw cap against a threaded female body. The polymeric carrier is a PC membrane (50 µm in thickness) that can be microthermoformed to include different patterns, such as trays, pockets, or microcavities.^{[ 30} , ^{31 ]} Moreover, multiple cell types can be directly cultured onto the PC carrier^{[ 32 ]} and on both sides.^{[ 33 ]} For example, this offers the opportunity to create a vascular barrier beneath the tissue model off‐chip, which once matured, can be transferred inside the OoC device.

In this study, thyroid organoids were embedded in basement membrane extract (Matrigel) on thermoformed membranous carriers with a central culture tray (5 × 5 mm). The membranous carrier allowed easy on‐ and off‐chip transfer of organoids with the help of tweezers, thus enabling culture in standard well plates prior to on‐chip perfusion experiment. The screw LnP clamp allowed quick loading and sealing of organoids with assembly times of less than four minutes, while providing leak‐tight conditions under clamping load (Figure 4C).

Once the cell culture bricks are plugged into the FCB, they are the highest point of the fluidic pathway. This may cause accumulation of air bubbles, for example inside the cell culture chambers. To address this issue, we inserted a simple passive bubble trap, which consisted of a third silicone layer with an elliptical‐shaped tank. The bubble trap was found to be able to entrap air bubbles in the tank, thus avoiding disruption of the flow in the cell chamber (Figure 4C).

2.5. Validation of the FCB Platform for Organoid Culture

After establishing a working microfluidic platform, we assessed its suitability for organoid culture. In a previous work, we have demonstrated that perfusion promotes the functionality of mESC‐derived thyroid organoids.^{[ 30 ]} However, on‐chip culture of organoids was difficult to execute, in part due to the high complexity of the microfluidic apparatus and time‐consuming cell culture protocols. In the present work, Matrigel‐embedded thyroid organoids (30–100 µm), were pre‐formed and cultured in microthermoformed carriers for three days under static conditions.^{[ 34} , ^{35 ]} Subsequently, these organoid carriers were loaded into six LEGO‐like OoC modules and connected to cell culture medium reservoirs to set up six FCB platforms. A plastic manifold was used to assemble the FCB platforms in parallel (Figure 1D). For proof of concept, a sensor module brick for continuous monitoring of DO₂ and pH levels was plugged into one FCB downstream to the OoC module (Figure  5A). All platforms were fluidically connected to a peristaltic pump and the flow rate was set to 12 µL min⁻¹ for 10 d of culture (Figure 5A,B). Perfusion was provided to thyroid organoids through the top chamber of the OoC modules. The FCBs occupied a total volume of 1.3 L (excluding pump) and allowed a minimal used tubing length of 34 cm per FCB, which corresponded to a 64% drop as compared to our previously used OoC platform.^{[ 30 ]} This decrease in tubing and connectors yielded 15× shorter assembly times of microfluidic peripherals (Figure S8, Supporting Information), facilitating the assembly of the total fluidic system by two operators within 2 h.

Figure 5.

Validation of the developed microfluidic platform for organoid culture. A) Photograph of the experiment set‐up at day 0. FCBs were connected with six reservoirs, six OoC module bricks, and one sensor module brick for monitoring of DO₂ and pH. B) Top view of most of the OoC module bricks at day 3 of perfusion. Some bubbles were entrapped inside the bubble trap. C) Side view of the fluidic system at day 10. The photograph shows no leakages at the fluidic interfaces. D) Bottom view of the sensor module brick. Yellow arrows indicate the flow direction, which passes through a bubble trap, a pH sensor dot (white), and finally through a black dot (DO₂). E) Continuous oxygen and pH measurement. F) Brightfield and live/dead assay pictures of thyroid organoids exposed to flow or kept in static wells for 3 and 10 d of culture. In green, Tg promoter‐driven GFP. Brightfield images taken with a 4× magnification. Scale bar, 150 µm. Fluorescence images of a live/dead staining of organoids cultured in the OoC module brick versus static controls. In red, dead cells stained with ethidium homodimer‐1 (EthD). Scale bar, 250 µm. G) Measurement of the equivalent diameter of thyroid organoids. All data are presented as mean (n = 35) and tested for statistical significance using Kruskal–Wallis test. ns, nonsignificant; p > 0.05. H) Confocal Z‐projections of thyroid organoids cultured for 3 d in OoC module bricks. Immunofluorescence images of organoids marked with GFP (green), and stained for Tg (red), ZO‐1 (white), and counterstained with DAPI (blue). Images taken with a magnification of 25×. Scale bar, 100 µm and insets, 25 µm.

The fluidic system was validated for perfusion consistency and accuracy using an in‐line flow sensor. The sensor was fluidically connected in series between the OoC module brick and the pump. The measured flow rates of three independent FCBs varied ≈8.4% from the nominal value with a precision between platforms of ± 0.33 µL min⁻¹ (Figure S9, Supporting Information).

The fluidic system was assessed at day 3 and 10 for technical inspection and cell harvesting. No leakage was detected in the system, including at the fluidic connectors, during the 10‐d perfusion period (Figure 5C). By day 3, the bubble tank was partially filled with air, with some air bubbles escaping into the cell chambers (Figure 5B). The sensor module brick allowed continuous monitoring of the DO₂ and pH levels throughout the culture period (Figure 5D and Video S7, Supporting Information). We observed a slight, gradual decrease of DO₂ over time, while pH values remained roughly steady at 7.3–7.6 (Figure 5E). Interestingly, the DO₂ sensor detected a drop in DO₂ when the incubator was opened for regular checkups.

We evaluated thyroid cell morphology and viability on chip and compared it to static controls. The controls consisted of thyroid organoids cultured in carriers and placed in static multiwell plates. At day 3 and 10, some OoC devices were unplugged from the FCB and imaged to assess any cell morphological changes. The remaining devices were opened and cells were harvested and stained using live/dead assay. No morphological changes were observed between static and flow conditions (Figure 5F,G). In addition, cell viability was similar between static and dynamic conditions with most cells being found alive (Figure 5H).

To evaluate the functionality of our thyroid‐on‐a‐chip model, we performed immunostainings against thyroglobulin (Tg) and ZO‐1. In the native gland, thyroid cells are required to be organized into follicular structures for Tg synthesis and successful iodination of precursor molecules into the thyroid hormones T3 and T4 in the luminal compartment. In the present study, organoids in both static and flow conditions showed polarization and follicular organization characterized by intracellular deposition of Tg and apical expression of ZO‐1 outlining the luminal compartment, suggesting recapitulation of the native thyroid‐like morphology and functionality in vitro (Figure 5I).

3. Discussion

As organoid technology rapidly advances, the combination of flexible and easy‐to‐use chip platforms with reversibly sealed OoC devices is desired to seed, host, and harvest these 3D multicellular models with increasing levels of complexity. To address this challenge, we demonstrated the design and fabrication of a new modular fluidic platform that allows insertion of organoid‐compatible culture bricks in a quick and flexible manner. Moreover, the developed FCB greatly reduced the amount of tubing and fluidic connectors required to integrate different microfluidic equipment. The small microfluidic footprint further allowed the full system to be placed inside a standard lab incubator, thus reducing the risk of bubble formation.^{[ 36 ]} Instead of combining module bricks on a common breadboard, previous studies have created alternative modular architectures by interconnecting module bricks to each other.^{[ 8} , ¹¹ , ^{12 ]} Although this strategy can achieve similar degrees of modularity and flexibility to those of our system, we decided to follow a breadboard approach as it enables culture modules to be assembled into a single, planar platform with a pre‐defined size. Importantly, the platform dimensions can be standardized to make it compatible with robotic lab equipment and automated liquid handling machines.^{[ 9} , ³⁷ , ^{38 ]}

A variety of manufacturing techniques has been employed to fabricate microfluidic devices. In this project, micromilling was selected for prototyping the FCBs, owing to its easy access, versatility, and dimensional quality. Desktop CNC micromilling allowed quick fabrication of channels in PC plates with great dimensional accuracy and sizes down to 300 µm. Moreover, milling offered a vast set of cutting tools to create channels with different geometries. Some limitations of milling microfluidic channels were identified. Milling resulted in a rough finish (Ra < 1 µm) on the surfaces of machined channels, which, albeit recurrent for milled plastics,^{[ 39 ]} can cause detrimental effects for cell experimentation. For example, high surface roughness can increase drug/compound adsorption to the channel walls and create new places for nucleation of air bubbles on the channel walls.^{[ 36 ]} Many finishing methods exist to reduce roughness; however, it is still difficult to achieve a surface finishing close to that obtained by photolithography processes. Another opportunity for improvement is in burr prevention and removal. We observed extensive formation of burrs in milled channels. These burrs are undesired as they create defects and irregularities in fluidic channels. Although the extent of burr formation could be controlled by refining the feed and speed parameters of the milling machine, elimination of burrs was only achieved by chamfering the edges with a cutting deburring tool. This process is suboptimal as it permanently changed the pre‐defined geometry of the channels. Other less invasive techniques have been reported to prevent the formation of burrs based on the deposition of a sacrificial layer.^{[ 40} , ^{41 ]} Future experiments could be carried out to elucidate the compatibility of such methods with cell culture. Nevertheless, micromilling revealed to be an accessible way for prototyping µm‐to‐mm sized channels while circumventing the necessity to use complicated processes and expensive equipment. Notably, our designs are fully compatible with standard replication techniques, such as injection molding for mass production of the different modules.

FCBs have been built by bonding a variety of materials together, including glass^{[ 42 ]} and thermoplastics, such as PC,^{[ 27 ]} PMMA,^{[ 13} , ^{43 ]} and PS.^{[ 42 ]} Thermoplastics are an attractive choice as they are amenable for mass production and offer the possibility to construct PDMS‐free FCBs, thus avoiding some limitations of PDMS in drug development and in industrial settings.^{[ 44 ]} In this study, we only used PDMS to create the fluidic chip layers of each cell culture module. To obtain a completely PDMS‐free system, fluidic chips could be made from alternative elastomers or thermoplastic elastomers (TPE), such as SEBS.^{[ 44} , ^{45 ]} Among thermoplastics, we decided to use PC for FCB fabrication, due to its low cost, transparency, and good machining properties. In addition, our PC platforms could be sterilized using standard methods and reused for multiple cell experiments, thus saving time and fabrication costs. Many techniques have been reported to bond thermoplastic plates with the potential to create FCBs. These methods include thermal compression bonding of PS,^{[ 46 ]} surface plasma modification and ultrasonic welding of PMMA,^{[ 47} , ^{48 ]} and lamination of COC plates.^{[ 49 ]} For PC–PC bonding, chemical modification followed by lamination of sheets as well as vapor solvent bonding have been reported.^{[ 27} , ^{50 ]} In our study, we presented an alternative PC–PC thermal bonding by compressing two PC sheets together with an adhesive foil. The main drawback of thermal bonding lies in the use of elevated temperatures, which can deform the geometry of the bonded channels. By refining our bonding settings, we demonstrated the formation of enclosed microchannels that were leak‐tight and had minimal channel deformation. Furthermore, the adhesive bonding was capable of enduring up to six consecutive cell experiments. Compared to previously published methods, our system offers the possibility to bond channels at relatively faster times than vapor solvent bonding and does not require chemically induced surface modification. On the other side, it was difficult to control the melting of the adhesive foil and, in some bonding regions, the adhesive material was displaced too much and intruded into the channels, ultimately creating blockages. This might stem from uneven load distribution during the bonding process and improvements on this aspect need to be made.

A crucial aspect of our design is the plug‐and‐play fluidic interconnections. These interconnections are key in designing FCB as they enable easy and reversible connection between the modular bricks and the breadboard. A multitude of plug‐and‐play fluidic interfaces have been reported and are thoroughly revised elsewhere.^{[ 51 ]} One of the most relevant examples was given by Vollertsen et al.,^{[ 9 ]} in which a PS FCB was developed for highly parallelized cell culture. The interface between the cell culture modules and the FCB was established by screwing the culture modules against the FCB, keeping under compression a set of O‐rings located in the I‐O ports. This design enabled simple integration of culture modules and leak‐tight operations. Contrarily to this approach, our O‐rings are placed radially in the LEGO‐like studs, and, once connected to the breadboard, the recess walls kept the O‐ring under compression. This is an improvement as it excludes the need for any fasteners to keep the O‐ring compressed, enabling faster and simpler assembly operations. Apart from this, our system revealed high reliability with no reported leakages during perfusion cell experiments for up to ten days. We believe that the reliability here achieved stems from the tight geometric tolerances, the quality control during manufacturing, and the use of a lubricant. Any of these aspects revealed to be indispensable for the success of this system.

To achieve a successful interface, the studs and recesses were carefully designed and tested, first using 3D printing and later by micromilling. In this study, 3D printed parts had lower consistency and reliability than milled PC ones, mainly owing to the poor surface quality of the 3D printed O‐ring grooves. After printing, the supporting printed material had to be manually removed from the O‐ring grooves, a process that caused permanent marks at the surface. These defects were detrimental for the O‐ring functionality and thus should be avoided. Nevertheless, 3D printing allowed us to conduct several initial design iterations at low cost and fast pace before testing PC parts for cell culture validation.

Once the FCB was established, we focused on facilitating the incorporation of a large amount of 3D organoids in the modular cell culture bricks. Previous examples of modular bricks have reported successful integration of 2D and 3D cell models on chip by seeding cells through narrow channels into cell chambers featuring sizes below 1 mm².^{[ 8} , ⁹ , ¹¹ , ^{13 ]} These in‐channel cell seeding methods can be challenging and inefficient, especially when using organoid models with a large volume or too sensitive to shear forces. Recently, we introduced a new concept, termed LnP, which englobes any device that enables reversibly sealing of an OoC in a single step.^{[ 22 ]} These devices allow opening and closing of cell chambers when needed and thus are interesting to directly culture organoids with a higher degree of complexity and in enough quantity for multiple downstream analysis. In this project, the new LnP device allowed incorporation of a large number of thyroid organoids (30 µL of Matrigel containing 3000 follicles) without the necessity for delicate and time‐consuming loading and harvesting protocols, such as in‐channel cell seeding or lysis. By doing so, the efficiency of cell culture was improved, hence providing the opportunity to conduct multiple analysis. Previously, we have shown that perfusion is beneficial for thyroid modelling by promoting T₄ synthesis of mESC‐derived thyroid organoids on‐chip.^{[ 30 ]} However, the dynamic culture was difficult to execute, due to the size and complexity of the microfluidic set‐up (e.g., entanglement of multiple tubing, leaking fluidic connections between reservoirs and microfluidic devices). Alternatively, in this study, we demonstrated that the developed FCB achieved simple microfluidic set‐ups and straightforward protocols while being suitable for dynamic organoid culture. It substantially reduced the amount of tubing and thus allowed faster (dis‐)assembly times. Therefore, this advanced platform offers new opportunities to test drug responses under flow and to study the effects of flow exposure on complex dynamic biological models.

Along with the LnP OoC modular brick, we have validated designs of reservoirs and sensor bricks, however a larger variety of functional, standardized and/or customized bricks is possible. The modularity of our system and the use of the same fluidic interfaces between different modules facilitates the designing process of new components to be applicable to the FCB.

4. Conclusion

We reported the development of a new modular microfluidic platform that reduces operation times and facilitates the culture of 3D thyroid organoids. This system achieves high versatility as it allows the fluidic combination of various modular bricks with different functionalities. Importantly, the system contains embedded microfluidic channels, which reduce the amount of tubing and connectors required, thus enabling easier operations and saving costs. Milling was demonstrated to be an accessibly way to create microfluidic channels with great dimensional accuracy and sizes in the range of 300–1000 µm. However, the surface quality of milled channels was not ideal for applications requiring a smooth finishing. Moreover, we reported the development of new fluidic interconnectors, which are capable of creating a real plug‐and‐play interface between the cell culture bricks and the FCB. Finally, an OoC brick was designed to be reversibly sealed by a LnP clamp, which facilitates the rapid and simple incorporation of thyroid organoids on chip. Overall, we believe that this platform presents novel innovative and valuable solutions toward the design of next‐generation microphysiological systems capable of synergistically combining organoids with microfluidic devices.

In the next steps, we aim at improving some technical aspects of our system. Desktop CNC milling of microfluidic channels showed some limitations, including difficult removal of burrs and high surface roughness. Methods to prevent burrs will be investigated based on deposition and milling of a top sacrificial layer.

As performed in this work, most of the reported FCBs are built by irreversibly bonding of two or more plates together to create a reliable, leak‐tight sealing of microfluidic channels. We intend to investigate the feasibility of reversibly sealing the top and bottom PC layer, instead of making a permanent bond. This might bring some advantages, including direct reconfiguration of the channel layout, by simply disassembling the bottom layer from the FCB and changing it with another one featuring a different channel layout. This opens an opportunity to provide a higher degree of flexibility and to reduce costs further.

5. Experimental Section

Micromilling of FCB Platforms

The FCB design was modeled using Solidworks (2015, Dassault Systems, Vélizy‐Villacoublay, France), converted into a g‐code file using VisualCAM software (2015, MecSoft Corporation, Irvine, CA), and fabricated by a CNC milling machine (Minitech Machinery Corporation, Norcross, GA). PC plates (Ensinger GmbH, Nufringen, Germany) with 1.5 and 6 mm in thickness were used to create the bottom and top layers of the FCB, respectively. Initial milling parameters were obtained from the mill tool manufacturer (Datron Dynamics, Inc., Milford, CT) and then optimized for surface smoothness. The final cutting parameters are reported in Table S1 (Supporting Information). After milling, the bottom layers were deburred in the milling machine to remove burrs at the edge of the channels, while I‐O ports of the top plate were reamed to a Luer geometry in a drill press using a carbide reamer tool (RLUER, Drill Service, Horley Ltd., UK). All holes and edges of PC components were chamfered using a 45° engraving tool.

PC–PC Thermal Bonding

After machining, PC layers were bonded together with an adhesive foil in a hydraulic press under a mechanical load of 0.75 tons. First, a 50 µm‐thick adhesive foil made from ethylene vinyl acetate (Polaflex Typ 80, T _g = 88 ± 4 °C) (Kunststoffwerk Lahr GmbH, Germany) was laser cut (80 W CO₂ laser, Trotec Speedy 300) to replicate the fluidic pathways into the foil. The cut was made with a bidirectional 1.5 mm offset relatively to the channel edges in order to avoid inflow of melted polymer inside the channels. Subsequently, the cut foil was placed between the PC layers, which in turn were sandwiched between two 0.5 mm‐thick silicone sheets. The rubber sheets acted as cushions to spread uniformly the load over the PC layers. All layers were stacked vertically and aligned with two stainless steel pins located at opposing corners of the FCB. The system was allowed to be heated up to 120 °C for 10 min, after which plates were cooled down and removed from the press. Any FCB samples with clogged channels or misalignments between I‐O ports and channels were discarded.

Fabrication of Cell Culture Module Bricks

Housing components were outsourced and machined by either lathe or CNC milling machines (Maastricht Instruments B.V., The Netherlands). The membranous carriers were fabricated in‐house by microthermoforming process.^{[ 52} , ^{53 ]} In brief, 50 µm‐thick PC films (it4ip, S.A., Louvain‐la‐Neuve, Belgium) were compressed between a prefabricated 0.7 mm‐thick brass mold (CW508L, Gemmel GmbH, Tuttlingen, Germany) and a counter‐plate in a hydraulic hot press. Temperature was set to 153 °C and compressed nitrogen was applied through openings in the counter‐plate (15 bar). The softened PC film stretched into the brass mold and after cooling down, nitrogen pressure was released, and the film was peeled off. Subsequently, the carrier was cut into an Ø29 mm circle featuring two Ø1 mm I‐O ports and a thermoformed, central tray with 5 mm × 6 mm (W × L). Prior to cell culture, carriers were air cleaned, immersed in 70% ethanol (EtOH) for 2 h, and dipped in a series of ethanol solutions of decreasing concentrations until 100% ddH2O. Fluidic chips were fabricated by PDMS casting. PDMS (10:1) (Sylgard 184, Midland, USA) was degassed, poured into milled negative molds, and closed with a smooth top lid. The system was let to cure at 80 °C for 2 h, after which the PDMS gaskets were peeled off while particles were removed with an adhesive tape.

Metrology

The surfaces of studs and FCB parts were thoroughly characterized using a 3D laser scanning microscope (Keyence VK‐X200K, Osaka, Japan). The geometry and dimensions of the milled microfluidic channels were evaluated after milling. Profiles of channel cross sections were obtained from a Z‐stack image taken with a 10× magnification and analyzed by Keyence MultiFileAnalyzer software. For refinement of milling parameters, slots with 20 × 20 mm were milled in a 1.5 mm‐thick PC plate at different feeds and spindle speeds. The bottom surface was profiled with a 10× magnification, Z‐pitch 0.5 µm. A multiline Ra value was obtained from each image using the Keyence MultiFileAnalyzer software (n = 3). Similarly, for quantification of the surface roughness on 3D printed and PC studs, devices were imaged and multiline Ra values were obtained from the O‐ring groove area and from the sidewalls (n = 3).

Deformation Analysis

Small, FCB blocks of 35 × 35 mm featuring a single‐channel pathway were bonded at different temperatures (no bonding, 70 °C, 90 °C, 120 °C, and 140 °C). Bonded blocks were injected with black‐colored PDMS and let to cure at 37 °C, O/N. Next, FCB samples were cut with a circular saw in two different places and channels cross sections were analyzed with a stereomicroscope (Nikon SMZ25, Tokyo, Japan). The PDMS filling reduced burr formation at cut edges of the channels, thus improving image quality. Images were taken with a 1× magnification and quantification of the width and height of the channels was performed using Fiji‐ImageJ 1.53t version software (National Institutes of Health, Bethesda, MD). For each cross section, three straight, equidistant lines were drawn across the channels area and perpendicular to the base of the channel (for height measurement) or to the sidewalls (width). The function “Analyze > Measure” quantified the length of each line and values were plotted individually. Data for each temperature condition represent measurements of two independent blocks cut at two different areas (n = 12).

Maximum Burst Pressure Tests

Burst pressure measurements were performed to evaluate the maximum internal pressure that the fluidic interfaces could withstand without leakage. In brief, studs were mated with correspondent recesses and outlets were closed with tape. Mated parts were immersed in a water tank and inlets were connected to a pressurized air tank from which air was supplied. A pneumatic pressure relief valve (Festo, Esslingen, Germany) was connected in‐line to prevent the internal pressure to build up more than seven bar. The burst pressure was recorded as the maximum pressure that causes leakage in the form of visible bubbles.

Organoid Culture

Thyroid organoids were differentiated from a recombinant murine embryonic stem cell (ESC) line (A2LoxNkx2‐1‐Pax8) as previously described.^{[ 30} , ³⁴ , ^{35 ]} In brief, ESCs were cultured on γ‐ray irradiated mouse embryonic fibroblasts (MEF) feeders in DMEM supplemented with 15% ES Cell qualified FBS (Sigma Aldrich, St. Louis, MO), IK0701 LIF (1000 U mL⁻¹) (ORF Genetics, Kopavogur, Iceland), non‐essential amino acids (0.1 × 10⁻³ m), sodium pyruvate (1 × 10⁻³ m), penicillin and streptomycin (50 U mL⁻¹), and 2‐mercaptoethanol (0.1 × 10⁻³ m). ESCs were cultured in hanging drops (1000 cells per droplet) in differentiation medium containing DMEM supplemented with 15% FBS, vitamin C (50 µg mL⁻¹), non‐essential amino acids (0.1 × 10⁻³ m), sodium pyruvate (1 × 10⁻³ m), penicillin and streptomycin (50 U mL⁻¹), and 2‐mercaptoethanol (0.1 × 10⁻³ m). After 4 d, embryoid bodies (EBs) were collected and embedded in Matrigel Growth Factor Reduced (354230, Corning, New York, USA) (50 µL Matrigel drops, containing 30 EBs per drop). EBs were differentiated into thyroid lineage using differentiation medium supplemented with doxycycline (1 µg mL⁻¹) for 3 d, followed by 14 d in differentiation medium supplemented with 8‐Br‐cAMP (10 × 10⁻⁶ m, B 007, Biolog, Hayward, CA). For flow experiments, differentiated thyroid organoids were extracted from Matrigel drops. In brief, cells were incubated with collagenase type IV (17104019, Gibco, Waltham, MA) (100 U mL⁻¹) and dispase II (04942078001, Roche, Basel, Switzerland) (4 U mL⁻¹) in HBSS (14175095, Gibco, Waltham, MA) at 37 °C for a maximum of 1 h 30 min. Subsequently, the cell suspension was filtered with a 100 µm cell strainer followed by reverse filtering using a 30 µm cell strainer (pluriSelect Life Science GmbH, Leipzig, Germany), resulting in a suspension predominantly composed of thyroid follicles (30–100 µm). Meanwhile, microthermoformed carriers were placed in petri dishes, washed with PBS, and precoated with Matrigel (5 µL) for 5 min at 37 °C to prevent organoid attachment to the bottom surface. Next, 3000 follicles in Matrigel (35 µL) were seeded per carrier. Carriers were then cultured with differentiation medium supplemented with 8‐Br‐cAMP (10 × 10⁻⁶ m) and TGF‐𝛽RI inhibitor SB431542 (10 × 10⁻⁶ m, 1614, Tocris, Bristol, UK) in static condition. After 3 d, carriers were either transferred to the OoC bricks or left in static conditions until day 10 post‐enrichment.

Flow Experiments

Cell culture bricks and FCB platforms were cleaned and disinfected with 70% ethanol followed by immersion in PBS and air‐drying in a cell culture hood. The remaining microfluidic peripherals (adaptors, needles, tubing, etc.) were autoclaved (120 °C, 30 min). Six FCBs were used for the flow experiments. Prelubricated O‐rings (EPDM 70, 55985, ERIKS, The Netherlands) were inserted into the studs of reservoir module bricks, which in turn were plugged into each FCB. Next, OoC bricks were assembled by stacking vertically gaskets, a carrier, and a plastic coverslip with the help of two Ø0.9 mm syringe needles, which acted as alignment pins. The system was made leak‐tight by manually rotating clockwise the OoC lid. After clamping, the needles were removed and the OoC brick was loaded with O‐rings and plugged into the FCB. All platforms were connected to an IPC‐N Ismatec peristaltic pump (78000‐47, Metrohm, The Netherlands) via PharMed BPT transfer (Ø0.8 × 4.0 mm, Saint‐Gobain, France) and two‐stop pump tubing (Ø0.89 mm, Saint‐Gobain, France). Prior to experimentation, the flow rate in the FCB was evaluated by a liquid flow rate sensor (SLF3S‐0600F, Sensirion AG, Stäfa, Switzerland). The sensor was inserted in‐line and downstream to the OoC brick. Data were acquired for 10 min and processed using GraphPadPrism 9 software. All FCBs were placed inside an ICO240 incubator (Memmert GmbH + Co. KG, Schwabach, Germany) with active humidity control. Cells were incubated at 37 °C, 70% relative humidity, and perfused with a flow rate of 12 µL min⁻¹ for up to 10 d.

Oxygen and pH Sensors

Sensor dots for oxygen and pH measurements (PreSens Precision Sensing GmbH, Regensburg, Germany) were glued onto a 1.5 mm‐thick PC coverslip. Prior to cell experiments, sensor dots were sterilized by autoclave (121 °C, 15 min) and batch calibrated using a two‐point calibration method. An oxygen‐free standard was obtained by dissolving Na₂SO₃ (1 g) and Co(NO₃)₂ standard solution (50 µL) (𝜌(Co) = 1000 mg L⁻¹; in nitric acid 0.5 mol L⁻¹) in water (100 mL). To obtain air‐saturated water, air was blown into a stirred water‐filled beaker for 20 min followed by 10 min without air supply and under agitation. No substantial differences in signal acquisition were detected due to sterilization. Next, the sensor dot‐loaded coverslip was inserted in a sensor module brick, which in turn was plugged into a FCB platform and downstream to an OoC brick. Two POFs were inserted into the custom‐made POM lid of the sensor brick and fixed by fastening side M3 × 3 mm grub screws. Oxygen and pH values were recorded every minute and for 10 d of perfusion culture period.

Organoid Morphology Assessment

Images of thyroid organoids exposed to flow versus control for 3 and 10 d were obtained with a brightfield microscope (Nikon Eclipse TS100). Quantification of follicle sizes was conducted using Fiji‐ImageJ version 1.53t software. In brief, follicles were manually outlined and ROIs were automatically analyzed for size by using the “Analyze>Measure” function. The results were representative of 35 follicles randomly selected from three different OoC devices.

Cell Viability

3D printed and machined PC four‐stud bottoms were incubated in differentiation medium for 3 d in petri dishes. Subsequently, mESC‐derived thyroid follicles were incubated with the conditioned medium of each bioreactor component, supplemented with 8‐Br‐cAMP (10 × 10⁻⁶ m), for 3 d at 37 °C. Cell viability was assessed by live/dead staining according to manufacturer's instructions. Dead cells were stained with ethidium homodimer‐1 (EthD‐1; Molecular Probes, Invitrogen, Waltham, MA), whereas all cells were counterstained with Hoechst (H3570, Thermo Fisher Scientific) according to manufacturer's instructions. Fluorescence images were taken with a Nikon Inverted Research Microscope ECLIPSE Ti.

Immunostaining and Confocal Analysis

OoC module bricks were manually disassembled and Matrigel‐embedded organoids were harvested from the membranous carriers. Organoids were washed twice with PBS, fixed with 4% v/v PFA (30 min, RT), collected into Eppendorf tubes, and stored at 4 °C in PBS until staining. For immunostaining, organoids were blocked with 0.2% Triton X100 supplemented with 3% w/v bovine serum albumin (BSA) and 5% w/v donkey serum (1 h, RT). Primary antibodies were diluted in 0.1% Triton X100 containing 3% BSA and 1% donkey serum (dilution buffer) and added to samples O/N at 4 °C. The following primary antibodies were used: mouse anti‐ZO‐1 (ZO1‐1A12, Invitrogen, Waltham, MA) (1:100) and rabbit anti‐thyroglobulin (A025102‐2, Agilent, Santa Clara, CA) (1:500). Samples were washed three times (15 min each wash) in PBS and incubated (2 h, RT) with the secondary antibodies (1:500 in dilution buffer): Alexa Fluor 568 goat anti‐mouse (A11008) and Alexa Fluor 647 goat anti‐rabbit (A11020). Cells were washed with PBS and counterstained with 4′,6‐diamidino‐2‐phenylindole (DAPI) (0.2 µg mL⁻¹) for 10 min at room temperature. Stained samples were mounted with Fluoroshield Mounting Medium (ab104135, Abcam, Cambridge, UK) onto a glass‐bottom petri dish and were imaged using a confocal laser scanning microscopy (Leica TCS SP8). Confocal Z‐stack projections were acquired and processed with Fiji‐ImageJ 1.53t software.

Statistics

All statistical analysis was conducted using GraphPad Prism 9 software. D'Agostino and Pearson test normality test was conducted to assess the distribution normality of all data sets. For data on quantification of the dimensions of bonded channels and the equivalent diameter of thyroid organoids, statistical significance was tested using Kruskal–Wallis test. ns, nonsignificant; p > 0.5; *p < 0.1, and **p < 0.01.

Conflict of Interest

S.G. is co‐founder and shareholder of 300MICRONS GmbH.

Disclaimer

LEGO is a trademark of the LEGO Group, which did not sponsor, authorize, or endorse this work.

Supporting information

Supporting Information

Supplemental Video 1

Supplemental Video 2

Supplemental Video 3

Supplemental Video 4

Supplemental Video 5

Supplemental Video 6

Supplemental Video 7

Carvalho D. J., Kip A. M., Tegel A., Stich M., Krause C., Romitti M., Branca C., Verhoeven B., Costagliola S., Moroni L., Giselbrecht S., A Modular Microfluidic Organoid Platform Using LEGO‐Like Bricks. Adv. Healthcare Mater. 2024, 13, 2303444. 10.1002/adhm.202303444

Data Availability Statement

The data that support the findings of this study are available from the corresponding author upon reasonable request.