Monolithic microfluidic platform for exerting gradients of compression on cell-laden hydrogels, and application to a model of the articular cartilage

Abstract Movement is essential to our quality of life, and regulates cell behavior via mechanical stimulation. Here, we report a monolithic microfluidic platform, in which engineered tissues composed of cells in a hydrogel are exposed to gradients of mechanical compression. Mechanical stimulation is applied through the deflection of a thin polydimethylsiloxane (PDMS) vertical membrane. The device design and all actuation parameters were optimized in this work to produce physiologically relevant compression on a cartilage model (strain of 5–12 %), as well as gradients of compression ranging from healthy to hyper-physiological conditions in the same device, as evidenced by the measured gradients in cell deformation. While this work focuses on mechanical compression of engineered tissues, we also demonstrated that our platform allowed creating more sophisticated multi-modal stimulation patterns. As the membrane is actuated by three independently addressed yet connected pressurized chambers, a variety of programmable deflection patterns and various cell stimulation modalities can easily be created by tuning the pressure applied in the different chambers (positive vs. negative, and amplitude). Advantageously, the fabrication of this monolithic platform is straightforward, with a single-step process. Moreover, the vertical membrane configuration allows for real-time imaging of cells encapsulated in the hydrogel matrix. The herein reported platform is highly versatile and of great interest to model other types of tissues, which also experience complex mechanical actuation patterns in vivo.


Introduction
The microenvironment all cells experience within the body is highly dynamic. In particular, mechanical cues are essential, since they regulate a variety of biological functions and cell processes in a spatiotemporal manner, as recently reviewed [1,2]; e.g., activation of mechanoreceptors in the cell membrane, rearrangement of the cytoskeleton, gene expression profiles, expression of cell-cell and cell-extracellular matrix (ECM) contact proteins [3], cell fate (proliferation, differentiation, phenotypic change, and apoptosis), and cell shape [4]. At the tissue and body levels, these mechanical cues have essential roles in, e.g., embryogenesis [5], tissue morphogenesis, angiogenesis [6][7][8], tumor progression [9], and reproductive biology [10].
Mechanical cues, which can be exerted in a continuous, temporary, cyclic, or pulsatile manner in the body, can be classified as, e.g., surface strain, fluid-based shear, substrate topography, matrix stiffness, stretching, bulk shear stress and compression. For instance, the epithelium and endothelium are exposed to fluid flows, which enhances their barrier function and induces changes in the cell configuration and cytoskeleton [11]. Epithelial cells undergo periodic stretching, for instance due to breathing in lung and the process of digestion in the gut, [12], which significantly influences effects of drugs and is primordial for the formation of a columnar epithelium. Compressive forces in the form of mechanical loads are experienced by bones and joints; changes in the load distribution in these tissues can impact their physiological response and possibly lead to impairment and/or fracture [13]. Furthermore, cartilage in the knee joint slowly degrades during long immobilization periods, due to the absence of mechanical cues, which regulate the ECM formation [14].

J o u r n a l P r e -p r o o f
Organ-on-a-chip technology [15] has become a game-changer in the field of biomechanics: it provides exquisite control on a cell microenvironment and allows studying in situ the impact of mechanical stimulations on the cell and organ function, as reviewed by the Kamm group [16]. Using a microfluidic format, the influence of fluid-induced shear stress has been examined on both the endothelium and epithelium, grown as monolayers on planar substrates [17,18], porous membranes [19], hydrogels [20] or in tubular geometries [21]. Exposure to physiological flow levels led to endothelial cell alignment [22], cytoskeleton rearrangement, and epithelial cell polarization and differentiation [23]. Stretching can be implemented by deforming elastomeric membranes prepared from polydimethylsiloxane (PDMS) through application of positive or negative pressures, to expose cells on these membranes to welldefined surface strains [24,25]. Noteworthy, such strain and fluid flow mechanical cues have been combined, to create more biomimetic environments, and to study the synergistic effect of various stimuli [22,[26][27][28]. Microfluidic platforms also lend themselves well to 3D cell cultures in hydrogel matrices [29]. By taking advantage of the deformability of these soft matrices, compressive forces have been exerted on 3D cell cultures using an elastomeric membrane: e.g., to expose cartilage cells to compressive mechanical cues [30,31], fibroblasts to a gradient of compressive forces [32], and mesenchymal stem cells to cyclic compression [33]. However, in these previous reports, horizontal membranes were employed, which presents two main limitations. First, the device fabrication can become tedious when multiple layers must be aligned and assembled together. Secondly, imaging of both the membrane and cells, notably to examine their response to the mechanical stimulation, implies using confocal microscopy. In short, real-time imaging of the engineered tissues is limited, or even precluded.

J o u r n a l P r e -p r o o f
In this paper, we present a monolithic PDMS-based platform for exerting mechanical stimulation on a 3D engineered tissue using a thin vertical membrane located next to the cell culture chamber, a configuration which was inspired by a module developed for microbial pressurization [34]. We first thoroughly examined the influence of various parameters on the membrane deflection level. Next, we mapped the deformation generated in an agarose matrix, upon application of compressive forces of different amplitudes. Following this, agarose supplemented with chondrocytes was exposed to physiological gradients of strains, and cell deformation and viability were characterized using fluorescence microscopy, as a function of their position in the hydrogel matrix. Finally, we explored the applicability of our platform to generate more sophisticated mechanical stimulation patterns allying compression and bulk shear forces, as found in articular cartilage during movement, through the sequential pressurization of the three actuation chambers. The herein reported platform opens new avenues for replicating complex physiological mechanical stimulation patterns on engineered tissues in a microfluidic format.

Platform design
The device incorporates a rectangular culture chamber (1260-µm length and 4000-µm width), separated from a perfusion channel (4400-µm length and 500-µm width) by an array of pillars (100 µm x 100 µm cross-section; inter-pillar distance 100 µm) (Fig. 1). Three actuation chambers are placed on the other side of the culture chamber, a thin vertical membrane (50 or 100-µm thick) physically separating the culture from the actuation chambers. Each actuation chamber has one inlet (1-mm diameter), and a 50-µm gap along the membrane links the 3 individual actuation chambers. All microfluidic structures are 200 or 220 µm in height.

Fabrication
Designs were drawn with Clewin5 software (WieWeb software, The Netherlands).
Microfluidic devices were produced from PDMS using soft-lithography [35] on a SU-8 mould. SU-8 100 photoresist (MicroChem, Westborough, MA, USA) was spin-coated on a <100> silicon wafer (Okmetic, Finland) to yield a 200 or 220-µm thick layer, exposed to UV light through a chromium-coated quartz mask, baked and developed according to the manufacturer's specifications. The resulting mould was treated in the vapour phase with trichloro(1H,1H,2H,2H-perfluorooctyl)silane (PFOTS, Sigma-Aldrich, Zwijndrecht, The Netherlands) during 20 min after plasma activation. A mixture of PDMS pre-polymer and curing agent (Sylgard 184, Dow Corning, USA) was prepared with a weight ratio of 20:1, thoroughly degassed, poured on the mould and once more degassed, before being placed in an oven for curing at 60°C for 24 h. Thereafter, inlets and outlets were created using 1-mm and 2-mm diameter biopsy punchers. The PDMS was subsequently cleaned with 70% ethanol and dried using pressurised air.
J o u r n a l P r e -p r o o f Two types of lids were considered: microscopy glass slides (1-mm thick) with or without a ca.
4-mm thick PDMS layer. To prepare the latter, uncured 20:1 weight ratio PDMS mixture was poured in a Petri dish, degassed, and cured as described above. This PDMS lid substrate was plasma-bonded (Cute, Femto Science, South Korea) to a glass slide. Both the lid (PDMScoated or non-coated glass) and the microfluidic structures were activated using plasma treatment, before assembly. The resulting microfluidic devices were placed in an oven at 60°C for 24 h before use.

Microfluidic set-up
For compressive stimulation, the three actuation chambers were connected to a positive pressure controller (MFCS-EZ, pressure output from 0 to 2000 mbar, Fluigent, France). For multi-modal actuation, the actuation chambers were connected to both the positive pressure controller and a negative pressure source (-350 mbar), via 2-way valves Fluigent) allowing rapid pressure switching for each actuation chamber using an automated script (Fluigent). The microfluidic platform was placed on the stage of an inverted microscope (IX51, Olympus, Japan) equipped with a camera (ORCA-flash 4.0 LT, Hamamatsu Photonics, Japan) and directly connected to a computer. A schematic representation of the entire set-up is provided in Supplementary information S1.

Device characterization
The membrane deflection was characterized for homogeneous compression (all three actuation chambers exposed to the same positive pressure). In each experiment, the membrane deflection was measured at four different locations along the device with ImageJ software (NIH, Bethesda, Maryland, USA), and at least three devices were characterized separately.
Values obtained from each device were averaged, and a standard deviation was calculated. All To quantify the hydrogel deformation upon deflection of the membrane, agarose 2% w/w was supplemented with polystyrene microbeads (15-µm diameter, amino-coated, Kisker, Germany) at a final concentration of 61 µg/ml. Specifically, a 2% w/w agarose solution was prepared by dissolving 0.5 g agarose powder (UltraPure agarose, Invitrogen, Thermo Fisher Scientific, Waltham, MA, USA) in 25 ml PBS after heating at 60°C. For each experiment, 30 µl of this solution was injected in the culture chamber. After 1 min gelation, culture medium at room temperature was added to the perfusion channel. It should be noted that the agarose hydrogel was not covalently attached to the PDMS. For homogeneous compression, the bead displacement normal to the membrane was measured as a function of their distance from the membrane. For the multi-modal stimulation, heat maps were generated for both the normal (compressive) and bulk shear strains; for this, the considered part of the culture chamber was divided in zones of 122 µm x 122 µm, and an average microbead displacement determined in each zone. All details regarding data analysis and normal and shear strain calculations are provided in the Supplementary information (S2 and S3).

Chondrocyte isolation and culture
The collection and use of human cartilage from patients undergoing total knee replacement was approved by a medical ethical committee (METC) of Zorggroep Twente, The Netherlands. Written informed consent was received from all patients. Chondrocytes were isolated from macroscopically and histologically healthy looking cartilage and expanded in T175 flasks (Cellstar®, Greiner bio-one, Germany) containing 20 ml of chondrocyte proliferation medium (DMEM with 10% FBS (Foetal Bovine Serum), 0.2 mM ascorbic acid J o u r n a l P r e -p r o o f 2-phosphate, 0.1 mM non-essential amino acids, 100 U/ml penicillin and 100 µg/ml streptomycin, 4 mM proline) [36,37]. Medium was refreshed every three days. When chondrocytes reached 80% confluence, they were passaged using a trypsin-EDTA (1X) solution (0.25%/0.1 mM, Invitrogen) in PBS in a new flask at a concentration of 500,000 cells per 175 cm 2 . Here, experiments were conducted with chondrocytes between passages 1 and 4.

Cell culture in the device and mechanical stimulation
Chondrocytes were cultured in agarose in the device. To that end, 1 ml of the 4% w/w agarose solution at 50°C was mixed with 1 ml of proliferation media at room temperature containing

Cell viability
Chondrocytes were cultured in agarose in the platform for up to three days, with or without mechanical stimulation (800 or 1000 mbar for 1 h per day, 1 Hz frequency). Medium was refreshed every day by pipetting manually twice 70 µl of medium in the perfusion channel.
After three days, cell viability was assessed using NucBlue® Live and NucGreen® Dead

Microfluidic platform for the mechanical stimulation of an engineered articular cartilage model
The goal of this study was to develop a microfluidic platform able to generate gradients of compressive mechanical stimulation and multi-modal actuation patterns on an engineered tissue, consisting of a cell-laden agarose matrix. This tissue model was confined in a rectangular culture chamber flanked on one side by a series of pillars, and on the other side by a thin PDMS membrane, acting as a mechanical actuator (Fig. 1). The pillars confined the hydrogel in the culture chamber by capillary action, while supporting the nutrient delivery from the perfusion channel into the cell-laden hydrogel. As a proof of concept, we demonstrate the applicability of our platform to create a cartilage model. Cartilage is a waterrich, relatively soft, elastic tissue, mainly composed of collagen (type II) fibers and proteoglycans, with a low percentage of specialized cells, called chondrocytes, in this extracellular matrix. In vitro cartilage models have been proposed using various synthetic or natural hydrogels such as collagen [38], alginate [39], and agarose [40 ,Buschmann, 1992 #74, [41][42][43][44], as used in this work.

Characterization of the device: parameters influencing the membrane deflection
First, we thoroughly characterized the PDMS membrane deflection to identify optimal parameters to emulate the compression level found in articular cartilage. For this, a positive pressure ranging from 0 to 1200 mbar was applied in all three actuation chambers concomitantly, and the influence of the PDMS formulation, the membrane thickness and height, and the nature of the lid substrate (glass vs. PDMS on glass) on the membrane deflection was quantified.
J o u r n a l P r e -p r o o f PDMS is an elastomeric material whose properties can be tuned by varying its composition and/or curing parameters [45]. More specifically, the lower the amount of cross-linker, the longer the PDMS chains between two cross-linking points, and the lower the material Young Modulus [46]. Here, we used a 20:1 prepolymer:curing agent weight ratio, which gave rise to larger deflection than a more traditional 10:1 ratio (Supplementary information S5). We next hypothesized that the lid substrate could impact the membrane deflection, and we compared glass substrates with and without a thick PDMS coating (Fig. 2a) To assess the device stability, membrane deformation was examined over a period of one month. No significant variation was found at different time points (days 0, 6, 14, 26 and 34) in individual devices (Fig. 2b). Moreover, even for high deflections (> 150 µm), the membrane remained entirely elastic, and the deflection amplitude was unchanged after 27,000 cycles of stimulation (1.5 h at a frequency of 5 Hz, 800 mbar applied pressure) (data not shown).
Noteworthy, implementing a post-curing step after device assembly was found to be essential to avoid large changes in the PDMS elasticity over time due to incomplete initial curing of the material.
Yet, all these characterization steps were performed in air. Therefore, we also examined the influence of the content of the culture chamber on the membrane displacement, and compared air, PBS buffer and agarose supplemented with 15-µm size microbeads at a concentration of 61 µg/ml. The nature of the material placed in the cell culture chamber had little to no influence on the membrane deflection (Fig. 2a). Consequently, the comprehensive device characterization performed in air can apply to agarose-laden culture chambers.

Creation of gradients of compressive forces as found in vivo
Next, we examined the agarose deformation upon application of homogeneous compression (Fig. 3a) by tracking the displacement in agarose of microbeads having a size (15 µm) similar to that of chondrocytes. Both the displacement and strain orthogonal to the membrane were quantified in front of the central actuation chamber along the full width of the culture chamber, for applied pressures of 100-1300 mbar. Overall, the applied pressure impacted the microbead displacement: the higher the applied pressure, the greater the bead displacement, and the closer the microbead to the membrane, the higher the displacement (Fig. 3b). Next to the membrane, the microbead displacement was comparable to the membrane deflection (i.e., 207.2 µm vs. 201.4±3.7 µm at 1000 mbar), and regularly decreased across the chamber width to less than 25 µm close to the pillar array. Interestingly, such a behavior is reminiscent of the deformation profile found in vivo in the knee cartilage, in which the superficial layer of this tissue experiences much higher deformation than layers close to the bone. Here, the agarose is not covalently attached to the PDMS membrane and to the roof and bottom of the culture chamber. When the membrane is deformed, the whole hydrogel slide inside the chamber. As a consequence, while the agarose in contact with the membrane is probably less compressed close to the roof and bottom, due to the shape of the membrane, there is no difference in displacement across the entire hydrogel height already about 100 micrometers from the membrane.
The strain was evaluated and found to be almost constant across the entire culture chamber for applied pressure lower than 1000 mbar (Fig. 3c). An applied pressure of 700 mbar yielded a strain value of 5-12%, which corresponds to physiological strain in healthy knee cartilage, while applying a 1000-mbar pressure allowed emulating hyper-physiological compression (> J o u r n a l P r e -p r o o f 20% close to the membrane), as observed in diseased knee cartilage [48]. Altogether, both healthy and pathological compression profiles in knee cartilage could be reproducibly generated in our platform, with the same gradients of deformation and strain, as found in vivo.

Deformation of chondrocytes under compression
The mechanical stimulation of chondrocytes can lead to a wide range of biological responses and increased or decreased viability [48]. Here, in a first instance, we characterized the chondrocyte deformation under homogeneous compression (800 mbar applied pressure; 1 Hz frequency). Since at that short timescale, cells are incompressible [49], which has also notably been demonstrated for chondrocytes [50], here, we specifically measured changes in the cell projected area, and not in their volume, which would not be appropriate in this configuration.
Changes in cell volume are typically observed after continuous and prolonged compression of cells [51], which is not the case here. The projected surface areas of a total of 24 chondrocytes across the culture chamber were measured, both at rest (before stimulation) and under compression, and the projected surface area decrease (%) plotted as a function of the chondrocyte distance to the membrane (Fig. 4). Under physiological compression (800 mbar), the chondrocyte projected surface area decreased by up to 13% at 200 µm to the membrane, and gradually decreased across the chamber from the PDMS membrane to the pillar array (2%). Altogether, cell deformation followed the same gradient as agarose displacement.

Chondrocyte culture and mechanical stimulation in the device
To assess the impact of this gradient of hydrogel compression on cells, human chondrocytes were cultured for three days in agarose in the device, and daily stimulated by cycles of homogeneous compression (1 Hz for 1 h), using parameters chosen based on previous reports [42,[52][53][54][55]. Two pressures were included to mimic healthy (800 mbar) and hyper-J o u r n a l P r e -p r o o f physiological (1000 mbar) compression, and static culture (no applied pressure) as a negative control. For each condition three individual devices were considered. Cell viability was evaluated at the end of the culture using NucBlue® Live reagent (Hoechst 33342), which stains all cell nuclei and NucGreen® Dead reagent that only stains the nuclei of cells, whose plasma membrane integrity is compromised. Two zones were considered, based on the cell deformation results and the presence of a gradient of strain, to evaluate cell viability (Fig. 5a).
Noteworthy, chondrocytes do not proliferate in agarose [56], which was confirmed in our experiments, in which no cell aggregate was formed during this 3-day culture period, as would occur upon cell proliferation.
After three days of culture, chondrocytes were mainly NucGreen®-negative (80-90% viability) for the static (no compression) and healthy dynamic conditions (800 mbar) in the entire culture chamber, demonstrating that cell viability was not compromised in our platform (static control) and by the healthy mechanical stimulation. In contrast, using hyperphysiological stimulation (1000 mbar), high level of NucGreen®-positive cells were found close to the membrane, indicating compromised cell membrane integrity, and possibly cell death (Fig. 5a). This variation in cell fate between static, dynamic healthy, and pathophysiological conditions was found to be significantly different in zone 1 next to the membrane (P<0.05) (Fig. 5b), but not in zone 2 closer to the pillars (Fig. 5c). Noteworthy, the static control confirmed that agarose is a suitable matrix for the culture of chondrocytes, and that nutrients were successfully brought to the entire engineered cartilage via the perfusion channel, either across the agarose or via above and below this matrix, since agarose was not attached to the PDMS. Interestingly, the latter configuration would not suffer from any possible changes in the material properties (e.g., decrease in mesh size) upon repeated compression.

J o u r n a l P r e -p r o o f
Only short-term changes in chondrocytes have been examined. Arguably, longer stimulation would increase the protein and matrix deposition by these cells, which would most probably translate into its stiffening of the agarose matrix, as notably reported by Bian et al. [57]. Yet, by continuously monitoring the membrane deflection, such changes could be identified, and the applied pressure increased accordingly to counterbalance those changes and ensure the same compression level is continuously applied to the engineered tissue in our microfluidic platform.

Exerting compressive and bulk shear forces on an engineered tissue
During the articulation of diarthrodial joints, mechanical stimulation comprises either intermittent compressions (standing position), or shear forces when the two bones linked by the articulation are sliding with respect to each other during movement. As a result, chondrocytes experience both compressive and bulk shear forces during movement, and both modalities induce different biological responses [58][59][60]. Our original mechanical actuation unit was designed to recapitulate both these compressive and more complex multi-axial stimulations, since the three actuation chambers are individually addressable, and both positive and negative pressures can be applied on-demand in any of the chamber with excellent temporal control. To illustrate this capability, different pressure sequences were applied in the three chambers. As depicted on Fig. 6a, a (Fig. 6b & c).
For a zone of 1220 µm x 1952 µm in the chamber (marked in Fig. 6c), vectors were plotted to represent the average direction and amplitude of the microbead displacement in this zone (Fig. 6d). Following this, heat maps were established for both the normal (in %) and bulk shear strain (in millirad) to represent the amplitude of cell deformation under compression and shear ( Fig. 6e and f, respectively). Although compression remained the dominant mechanical stimulation modality, microbeads also moved along the membrane, with amplitudes amounting to 55% of their normal displacement. Compression occurred next to the central actuation chamber, and relaxation next to the left one. Normal strain was reduced in this configuration compared to a homogeneous compression modality, which can be explained by the partial deflection of the membrane in this scenario, when only one or two chambers were simultaneously pressurized. Bulk shear strain was concomitantly examined, as detailed in Supplementary information S3. The average value of the shear strain amplitude was 9.8±2.9 mrad, which is in good agreement with previously reported values for shear strain on cartilage [58][59][60][61], as summarized in Supplementary information S6. Interestingly, by further varying the actuation sequence, a great variety of stimulation patterns could be created in our device to mechanically stimulate cell-laden hydrogel matrices.

Conclusion
We reported here a monolithic platform comprising a vertical PDMS membrane actuated by a series of three independent yet connected chambers, to create various mechanical stimulation patterns on engineered tissues, while providing continuous optical access in the entire culture chamber. Thorough characterization of the platform allowed identifying optimal parameters to maximize the membrane deflection under uniform compression, and to mimic both healthy and hyper-physiological mechanical stimulation, as found in articular cartilage. Chondrocytes In future work, we will examine the influence of the multi-modal stimulation on chondrocyte behavior, in particular on their ability to create their own microenvironment and in terms of gene and protein expression patterns after retrieval of the engineered tissue from the platform after prolonged exposure to physiological compression. Moreover, it is well-known that the exact stiffness of the hydrogel material has an impact on the cell behavior and differentiation, as well as on the matrix formation [62,63]. Other hydrogel matrices or combination of materials could be tested in our platform in combination with mechanical stimulation, to identify an optimal matrix composition to engineer cartilage tissues. Noteworthy, the herein reported easy-to-produce and cutting-edge platform is not only of great interest to study cartilage, as demonstrated here, but it also offers unprecedented options to apply unconventional mechanical stimulation in other micro-physiological organ models. wrote the manuscript and all authors approved its content.

Conflict of Interest
There are no conflicts of interest to declare.

Acknowledgements
We acknowledge financial support for this project from the ReumaNetherlands grant LLP-25.