Stacked hydrogel-based brain-on-chips utilizing capillary force flow pinning

The brain, as the most intricate organ in the human body,¹ characterized by a complex layered architecture,² structural heterogeneity, and regional stiffness variation.^3–5 These features underpin its functional capabilities. In vivo cellular processes such as migration, proliferation, morphology, and differentiation are shaped by mechanical cues,^6,7 highlighting the crucial interplay between the structure and function in brain tissue. Establishing a reliable in vitro model of brain tissue remains a significant challenge due to its complexity.^8,9

Addressing this challenge, we propose a microfluidic device approach to design and create relevant and use-dependent microenvironments for culturing neuronal cells by stacking two hydrogel layers of defined height in the device, thanks to capillary force flow pinning. Our previous study has focused on neuronal migration across varying material stiffness gradients. Confocal z-stack imaging was performed on 3-week differentiated SH-SY5Y human neuroblastoma cells labeled with β-tubulin III. The cells demonstrated migration from a stiffer to a softer substrate, with a distance of up to 60 μm.¹⁰ However, the hydrogel layers could not be controlled in height sufficiently simply by using a micropipette. Controlling the layer height is one of the key challenges of this work in addition to the challenges related to providing a microliter-scale 3D culture environment for highly shear-stress sensitive human induced pluripotent stem cells (hiPSCs)-derived neurons.

In this work, the microfluidic device design process providing pinning feature by easy access SLA printing of molds enabled us to replicate both the defined height of the layered hydrogel architecture with varying regional stiffness and working out the cell culture conditions fit for these small reservoir geometries with respect to cell density and media refreshment. Subsequently, the brain-on-chip (BoC) designs as introduced here offer promising advances in microengineered BoC models, which gained tremendous popularity in biomedical research over the last decade.^11–16 More specifically, stereolithography (SLA)-based 3D printing allows us to create versatile molds for these microfluidic devices. SLA-based 3D printing is increasingly adopted for creating microfluidic devices due to its numerous benefits over traditional manufacturing methods. These benefits include greater design flexibility, the ability to produce intricate designs, and the capability to construct complex architectures in a single step.^17,18

Demonstrating the versatility of the printing process, we designed and devised single- and dual-reservoir BoC devices. Both design configurations include specific pinning features aimed at creating a gel stack with two layers of defined height in the z-direction to mimic human physiology in the medial and sagittal planes.¹⁹ Polydimethylsiloxane (PDMS) replica molding subsequently allowed us to replicate the SLA-based 3D-printed mold structures and fabricate enough microfluidic devices for testing these two designs in BoC modeling applying human induced pluripotent stem cell (hiPSC)-derived neurons.

A single- and dual-reservoir BoC device was designed and fabricated, respectively.

In this design, the culture chamber is designed as a simple single well with a multilayered structure implementing two pinning rims. The outer dimensions of the part providing the cell culture reservoir of the chip are 28 × 39 mm² prescribed by the mold walls that have a width of 3 mm yielding a mold base of 34 × 42 mm². The mold base is 3 mm thick. The inverted shape of the reservoir is characterized by three distinct pinning rims. The first rim has a height of 0.2 mm and a diameter of 7 mm. The second and third rims each have a height of 0.4 mm, with the second rim having a diameter of 8 mm and the third having a diameter of 9 mm. Subsequently, the shape continues by a 75° bevel spanning from a diameter of 10–14 mm over a height of 2 mm. This configuration allocates an additional volume to the reservoir structure in the replicated device to accommodate media exchange atop the hydrogel layers during cell culture. Illustrations of the design are depicted in Sec. III [Figs. 1(e) and 1(f)]. A detailed design drawing encompassing both the overall dimensions and 3D printing instructions is provided in Fig. S1 in the supplementary material. Section III provides the functional characterization of the pinning rims for controlled hydrogel placement.

In this design, the outer dimensions of the chip are 30 × 34 mm² prescribed by the mold walls that have a width of 3 mm, yielding a mold base of 36 × 40 mm². The mold base is 2.5 mm thick. The culture chamber comprises a system of two 8 mm diameter wells linked by a straight channel with a width of 4 mm and a length of 8 mm featuring uniquely designed sidewalls with sawtooth-shaped structure. In addition, a rectangular shape with a width of 1 mm and a height of 0.2 mm is implemented in the culture chamber’s sidewalls, which can act as a microfluidic feeder channel encircling the replicated structure of the dual-reservoir device [Sec. III, Figs. 1(a)–1(c)]. The total height of the mold’s sidewalls is 2.5 mm (excluding the base plate). The inclusion of the sawtooth-shaped structure facilitates the formation of two stacked hydrogel layers along the straight linker channel with a prespecified height of 1 mm each. Illustrations of the design are depicted in Sec. III [Figs. 1(a)–1(c)]. A detailed design drawing encompassing both the overall dimensions and 3D printing instructions is provided in Fig. S2 in the supplementary material. Section III provides the functional characterization of the sidewall structure for controlled hydrogel placement and the feeder channel fluid manipulation.

The fabrication of BoC devices was performed in two steps, as depicted in Fig. 2. First, we made a CAD file (NX-12, Siemens) of the two designs and fabricated the molds by using an SLA-based 3D printer (Formlabs, Form 3+). Optimal printing settings as indicated in Table S2 in the supplementary material per design configuration were made by loading the NX-12 files into PreForm software. Both molds were printed with Clear V4 resin (Formlabs) in a printing orientation with 60°-to-print table pulling direction. Postprinting, the molds underwent a 30 min cleaning in an isopropanol bath, followed by 1 h UV postcuring at 405 nm, 5.5 mW cm⁻² (Formlabs, FormCure). Subsequently, built support structures were removed by cutting with a blade. Second, we performed PDMS soft lithography (SYLGARD 184, Dow Corning) using the molds to devise multiple chips for the culture experiments.

Prior to use, molds are cleaned with isopropanol and dried with pressurized nitrogen. Figure 3 illustrates the PDMS replica molding process in detail. Double-sided tape (Type 4965, Tesa) was cut with a laser cutter (Universal Laser Systems) in accordance with the pattern of the dual-reservoir shape and the three walls of the mold and mounted on the mold [Figs. 3(a) and 3(b)]. Ease Release 200 (Mann Technologies) was then sprayed from a distance of 15–20 cm on the mold surface, and the molds were left to dry for 5 min under a fume hood [Fig. 3(c)]. A 2 mm thick poly(methyl methacrylate) (PMMA) sheet measuring 40 × 36 mm² was cut using the same laser cutter as for the tape, and after removing the protective film from the tape on the mold, the PMMA and the mold were assembled and clipped together [Figs. 3(d)–3(g)]. This assembly was placed in an 80 °C oven, while the PDMS preparation took place. Increased temperature aids in softening the tape material, ensuring a strong bond without any gas bubbles forming between the PMMA and the 3D-printed mold. The PDMS base was mixed with a curing agent at a 10:1 ratio and degassed in a vacuum chamber (Kartell Labware). The assembled mold was put with the opening upright [Fig. 3(g)], and the PDMS mixture was then slowly poured into the mold. To reduce bubble formation, the PDMS was partially cured at room temperature overnight. Subsequently, the curing process was completed by placing the assembly in an 80 °C oven for an additional overnight curing step. The following day, the PDMS was peeled from the mold and soaked in 70% ethanol for 2 h. The PDMS parts were then air-dried with an air gun and placed back in the 80 °C oven overnight to ensure complete curing. Finally, PDMS parts are bonded to microscope glass slides 76 × 26 mm² (631-1552, WVR), respectively, using O₂ plasma pretreatment for both the glass and the PDMS by means of a plasma asher in an N₂-gas ambience (Emitech, K1050X, 20 W, 30 s). Section III elaborates on the results per fabrication step.

Neurogenin-2 (Ngn2+) hiPSCs were employed for the biological validation of the two design configurations of the cell culture chamber system with pinning features. The experimental procedure was adapted from the protocol established by Nael Nadif Kasri's group at Radboud University.²⁰ 

Expansion of hiPSCs was performed on Matrigel (1:15) coated six-well plates in a complete C8 medium, which was changed daily. Cells were ready for plating when they reached 80% of confluency.

For the induction of neuronal differentiation, hiPSCs were enzymatically dissociated using Accutase. The cells were plated in a dual-reservoir BoC device coated with 50 μg/ml poly-L-ornithine (PLO; Merck, P4957) and 20 μg/ml laminin (Sigma, L2020), as well as in a single-reservoir BoC device coated with 20 μg/ml laminin and 5% (w/v) gelatin methacryloyl (GelMA; Sigma, 900496). The cells were seeded at a density of 10 × 10⁴ cells/cm² and supplemented with a 1:100 dilution of RevitaCell on day 0 to enhance cell viability and promote neuronal differentiation (Table I). Half of the medium was exchanged every second day with dedicated medium of which the composition per timepoints in the protocol is provided in Table I. hiPSCs were differentiated into iNeurons, with cultures maintained for 3 weeks. Materials and suppliers are listed in the supplementary material section (Table S1).

After culturing iNeurons for 3 weeks in the two BoC configurations, cells were washed with Dulbecco’s phosphate buffered saline (DPBS, calcium, magnesium, Gibco, 14040133), and fixated for 15 min at room temperature with a solution of 4% formaldehyde (Thermo Scientific Alfa Aesar, 043368) and 4% w/v sucrose (Thermo Fisher, J65148.36). Subsequently, the cells were washed three times for 5 min at room temperature with DPBS again and permeabilized with 0.2% Triton X-100 (Sigma-Aldrich, 9036-19-5) solution for 15 min at room temperature. Thereafter, cells were incubated with blocking buffer solution of DPBS with 5% normal goat serum (Invitrogen, 31873) for 1 h at room temperature. Primary antibodies [microtubule-associated protein-2 (MAP2), Rabbit polyclonal, Abcam ab32454, 1:500; Homer-1, Chicken polyclonal, Synaptic systems, 160 006, 1:500; Synapsin 1/2, Guinea pig polyclonal, Synaptic systems, 106 004, 1:1000] were diluted in 1% blocking buffer were applied to the cells and incubated overnight at 4 °C. Then, cells were washed three times for 5 min with DPBS followed by incubation with secondary antibodies (Goat anti-Rabbit IgG Alexa Fluor 488, Thermo Fisher, 10236882, 1:2000; Goat anti-Guinea pig IgG Alexa Fluor 647, Thermo Fisher 10624773, 1:2000; Goat anti-Chicken IgG Alexa Fluor 405, Thermo Fisher 17121839, 1:2000) diluted in 1% blocking buffer for 4 h at room temperature. Finally, cells were washed three times for 5 min with DPBS followed by submerging the chips in the fluorescent mounting medium. (Mowiol 4-88, Polysciences, 25213-24-5). Then, cells were ready to be imaged using a fluorescent microscope (DMI8, Leica).

To characterize the pinning function in the BoCs for gel stacks with predefined height, colored thermoset gelatin (Dr. Oetker) was utilized. A solution of 20% w/v red and blue colored thermoset gelatin was prepared by combining gelatin powder with boiling water. Following preparation, the mixture was poured onto the top layer as the initial layer and, subsequently, cooled at 4 °C for 2 h to solidify. Once the first layer was set, the second layer was poured atop it. Gel stack formation was observed using a digital microscope (VHX-5000, Keyence).

PDMS BoC devices with pinning features, serving as a culture chamber system in combination with hydrogel scaffolds to mimic the layered architecture and regional stiffness variation in brain tissue in a brain-on-chip model, were successfully devised and assessed for their performance for both design configurations. Simple PDMS rings do not create a flat hydrogel surface, resulting in uneven heights and meniscus shapes across such simple wells. Our uniquely designed PDMS BoC devices facilitate a two-layer architecture of stacked gels, each with a certain height still by simple dispensing techniques abundantly available in established cell culture laboratories. Although oxygen plasma treatment can mitigate the meniscus effect for the culture well with a simple PDMS ring, it does not ensure consistent layer heights.

To address these challenges, we created two new BoC designs with (1) a single-reservoir and (2) a connected dual-reservoir design (Sec. II A 2). The choice of clear V4 resin was selected for its high resolution and compatibility with PDMS replica molding.

Our 3D multilayer hydrogel stack BoC modeling approach involves a straightforward microfluidic pinning effect implemented in two configurations of a single- and dual-reservoir design.

In the single-reservoir design, two pinning rims enable two layers of gel with a height of 0.4 mm each in the z-direction by standard pipetting techniques. These dimensions are selected to ensure that the medium can reach the bottom layer, which is essential for cell culture applications to maintain sufficient nutrient accessibility and waste removal by diffusion inspired by our previous microbioreactor-supported BoC design.^10,21 Here, it is the intention that two different gels are arranged in a sandwiched manner with neurons embedded in between the gels. Further brain research will then allow to mimic changing stiffness and (bio)chemical cues along the growth cone trajectory of neurons comparing such attractive scaffold properties and (bio)chemical interactions between cells and scaffolds against the effect of gravity.

To illustrate the concept, the first gel layer is loaded into the reservoir using a pipette, where the first rim pins the gel in place. After the first gel layer polymerizes, the second gel layer is added on top in the same manner pinning the gel precursor at the second rim [Figs. 4(a) and 4(b)]. Cut open by a scalpel [Fig. 4(c)] shows in an optical micrograph the cross section of the gel layers and the effect of the pinning rims in maintaining the structure.

The dual-reservoir design configuration incorporates additional microscale structures next to the sawtooth pinning features within its sidewalls to enhance fluidic functionality, such as facilitating a feeder channel. This advanced microfluidic device comprises a sophisticated microgrooved structure in its sidewalls with each sawtooth pinning line measuring 1 mm in height, which serve to secure the gel at specific levels by capillary flow, a phenomenon where adhesive forces between the liquid and the tube walls are stronger than the cohesive forces within the liquid. In this context, the adhesive forces pull the liquid upwards, resulting in a concave meniscus at the top. The height to which the liquid rises is inversely proportional to the tube’s diameter, with smaller diameters enhancing the capillary effect [Fig. 5(a)].^22–24 Here, we utilized this phenomenon in our microfluidic design to allow for an engineered solution of efficient capillary-driven nutrient flow by implementing the additional flow channel. The latter ensures precise positioning above a substrate, while the encircling groove allows additional microfluidic interactions. Previously, we used the direct extrusion printing of the microfluidic parts with the sawtooth structure and demonstrated that it allows us the establishment of a pinning effect for hydrogel layers with defined height along the linking channel of the dual-reservoir design. When hydrogel is introduced via standard pipetting, the sidewall structure effectively pins the hydrogels, forming defined layers [Figs. 5(b)–5(e)]. The larger reservoir area moderates the liquid’s rising level during filling, allowing for a more controlled input flow with sufficient precision of the hydrogel layer formation. To advance the dual-reservoir BoC device, a sophisticated design to harness capillary action, offering improved control and precision in hydrogel delivery and layer formation including a feeder channel was developed by converting the extrusion-printed prototype design into a mold design with the extended microfluidic feature of the encircling microgroove fabricated by SLA-based 3D printing. Subsequently, microfluidic devices by PDMS soft lithography were devised to observe the gel stacking effect in this dual-reservoir configuration of the BoC devices [Figs. 6(a)–6(c)].

In the dual-reservoir BoC device, each layer was 1 mm in height [Fig. 1(b)] due to the constraints of SLA-based 3D printing on mold fabrication. Thick gel layers can present a challenge due to potential nutrient deficiencies in the bottom layers when only fed from the top of the culture chamber system. Hence, we assessed whether the additional sidewall microfeature implemented in the dual-reservoir BoC configuration served as a microfluidic feeder channel with the potential to ensure nutrient- and waste management across the gels also at the bottom of the cell culture chamber system. We tested the functionality of the feeder channel concept using a syringe pump with a flow rate of 50 μl/min and colored microfluidic dyes (Darwin Microfluidics), confirming its ability to hold medium. Figure 7 shows the experimental setup and an overview of the results of this assessment.

Evaluating the results presented in Fig. 7, it appears easy to get the microfluidic function to work with an open channel at a liquid-to-air interface. However, one major issue was encountered with these feeder channels impeding their function. They tend to get filled with gel, too, by capillary action during the filling of the linker channel between the two reservoirs with hydrogel shown in Fig. S3 in the supplementary material. This problem remains to be resolved by optimizing the design dimensions and SLA printing process further to achieve a fully functional microfluidic BoC. Alternatively, one could fill the nutrient channel with a water-immiscible fluid inspired by two-phase flow microfluidics²⁵ adjacent to the hydrogel layers and subsequently wash out and replace the immiscible fluid with the nutrient fluid. To further characterize the feeder channel function, fluorescent tracer diffusion experiments similar to permeability studies performed for vascular modeling²⁶ could be done. Subsequently, cell studies with viability assays²⁷ must be performed to confirm the differentiation and maturation of cultured neuronal cell networks in the hydrogel layers of the BoC.

Similarly to other researchers,^28,29 we successfully generated hiPSCs-derived neurons by Ngn2+ overexpression. These excitatory neurons are an essential tool for studying neurodevelopmental processes and disorders.

After plating, hiPSCs-derived neuronal precursor cells strived in both microfluidic BoC design configurations and our neuronal cell culture proceeded for three weeks. To mitigate cell detachment issues, on the third culture day 3 (DIV3), the astrocyte conditioned medium was introduced to support the neuronal differentiation processes. Immunostaining at day 21 confirmed the presence of induced neurons (iNeurons) showing expression of MAP2, Homer-1, and synapsin-1/2 (SYN1/2) by fluorescence microscopy [Figs. 8(c)–8(f) and 9(c)–9(f)].

Our extended research aim is to integrate these devices with a mechanical actuator, moving beyond simple glass microscope slides for sealing the PDMS constructs. This actuator will employ a thin, deformable PDMS membrane layer, as utilized in our previous experiments.³⁰ With advancements in 3D printing technologies, we plan to use Femtoprint to micromachine channels in glass, enabling the development of such actuated BoC systems. These future efforts will extend the current work and complete our initial investigations into nanomechanical actuated dynamics in stem cell-derived neuronal cell networks.³¹ 

We demonstrated the design and fabrication of two distinct BoC devices, termed single-reservoir and dual-reservoir BoC devices, by integrating SLA-based 3D printing with PDMS replica molding. This technique enabled the creation of a brain-mimicking microenvironment, inspired by the native brain’s regional stiffness variations and layered architecture. The single-reservoir BoC device utilized pinning rims to generate multiple vertical gel stacks at certain heights. For the dual-reservoir BoC device with linker channel, we introduced an additional microfeature by designing a microgroove on the culture chamber’s sidewalls in addition to the sawtooth pattern that facilitates vertical gel stack formation through fluidic capillary action. This design could be used for improved microfluidic functionality by introducing a feeder channel in the dual-reservoir BoC, when the dimensions of the sidewall channel and the mechanical properties of the gel precursor during filling are further optimized.

Furthermore, we proved the differentiation of hiPSCs into iNeurons within both BoC configurations, confirmed by immunofluorescence staining of neuronal morphology and synapse formation in 21 days. These findings confirm the utility of the proposed BoC devices for neuroscience research.

See the supplementary material for detailed dimensions of the single-reservoir BoC mold design (Fig. S1) and dimensional overview of the dual-reservoir BoC mold design (Fig. S2); the challenge encountered during the filling of feeder channels with hydrogels (Fig. S3); materials with the suppliers used in cell culturing (Table S1); and 3D printing settings of the BoC molds (Table S2).

The research tasks of Regina Luttge are partially funded by the European Union’s Horizon 2020 research and innovation program H2020-FETPROACT-2018-01 under grant (Grant No. 824070). We extend our sincere gratitude to the Nael Nadif Kasri Lab at Radboud University for generously providing us with Ngn2+ hiPSCs.

The authors have no conflicts to disclose.

Gulden Akcay: Conceptualization (equal); Investigation (lead); Visualization (lead); Writing – original draft (lead); Writing – review & editing (equal). Jeroen van Venrooij: Investigation (equal). Regina Luttge: Conceptualization (equal); Funding acquisition (lead); Project administration (lead); Supervision (lead); Writing – review & editing (equal).

The data that support the findings of this study are available within the article or its supplementary material.