A lab‐on‐a‐chip model of glaucoma

Abstract Aims We developed a glaucoma‐on‐a‐chip model to evaluate the viability of retinal ganglion cells (RGCs) against high pressure and the potential effect of neuroprotection. Methods A three‐layered chip consisting of interconnecting microchannels and culture wells was designed and fabricated from poly‐methyl methacrylate sheets. The bottom surface of the wells was modified by air plasma and coated with different membranes to provide a suitable extracellular microenvironment. RGCs were purified from postnatal Wistar rats by magnetic assisted cell sorting up to 70% and characterized by flow cytometry and immunocytochemistry. The cultured RGCs were exposed to normal (15 mmHg) or elevated pressure (33 mmHg) for 6, 12, 24, 36, and 48 hr, with and without adding brain‐derived neurotrophic factor (BDNF) or a novel BDNF mimetic (RNYK). Results Multiple inlet ports allow culture media and gas into the wells under elevated hydrostatic pressure. PDL/laminin formed the best supporting membrane. RGC survival rates were 85%, 78%, 70%, 67%, and 61% under normal pressure versus 40%, 22%, 18%, 12%, and 10% under high pressure at 6, 12, 24, 36, and 48 hr, respectively. BDNF and RNYK separately reduced RGC death rates about twofold under both normal and elevated pressures. Conclusion This model recapitulated the effects of elevated pressure over relatively short time periods and demonstrated the neuroprotective effects of BDNF and RNYK.


| INTRODUC TI ON
Glaucoma is a collective term used to define a group of neurodegenerative processes affecting the entire visual pathway best distinguished by progressive, irreversible destruction, and death of retinal ganglion cells (RGCs). The disease spectrum is estimated to affect more than 100 million people worldwide by the year of 2040 (Tham et al., 2014). The primary risk factor for progression and development of glaucoma is elevated intraocular pressure (IOP). IOP is regulated by the balance between aqueous humor secretion into the anterior chamber on one hand and its drainage via the trabecular meshwork (conventional outflow) or the uveoscleral outflow pathways on the other hand (Dawson, Ubels, & Edelhauser, 2011). The average value for normal IOP in most population-based human studies is 14-17 mmHg with 95% confidence intervals of 10-21 mmHg (Rieck, 2013).
Several experimental in vivo glaucoma models have been developed and used for study on glaucoma mechanisms. These include laser photocoagulation of the Perilimbal region (Gaasterland & Kupfer, 1974), red blood cell or microbeads injection into the anterior chamber (Quigley & Addicks, 1980), Episcleral vein obstruction (Fedorchak et al., 2014), and Episcleral vein saline injection (Morrison et al., 1997;Vecino & Sharma, 2011). Without sequential treatments, the duration of IOP elevation is transient in these models. Precise control over IOP elevation is difficult and certain problems may occur; these include intraocular inflammation, irreversible mydriasis, IOP variability, hyphema, reduced visibility of optic disks, corneal opacity, and scleral burns (Johnson & Tomarev, 2010;Rudzinski & Saragovi, 2005). While in vivo animal models are indispensable to determine what events occur in live organisms, these strategies typically involve poorly defined and uncontrollable factors and the results can be difficult to understand at cellular and molecular levels (Ishikawa, Yoshitomi, Zorumski, & Izumi, 2015). Ishikawa et al. designed an ex vivo hydrostatic pressure model which demonstrated better retention of neuron-neuron and neuron-glial interactions in dissected eye cups (Ishikawa, Yoshitomi, Covey, Zorumski, & Izumi, 2016Ishikawa, Yoshitomi, Zorumski, & Izumi, 2010. This model excludes the effects of ischemia and allowed studying of the direct effects of pressure on the retina. Eyecups were sunken to the bottom of a glass cylinders containing different liquid heights. Although ex vivo systems produce reliable results, there are several limitations. For example, the incubation period is limited by the duration of tissue viability because survival factors originating from axonal transport or the bloodstream are eliminated in ex vivo models. To overcome these limitations, in vitro models have been developed using cultured cells in supporting environments to clarify glaucoma mechanisms and appear as valuable tools to evaluate the response of individual cell populations against noxious conditions and novel treatments (Dai, Zhang, Zheng, & Wang, 2018;Maurya, Agarwal, & Ghosh, 2016). These models have studied optic nerve head astrocytes, RGCs, and other types of retinal cells utilizing pressure loading systems (Agar, Li, Agarwal, Coroneo, & Hill, 2006;Lei et al., 2011;Ricard et al., 2000;Salvador-Silva et al., 2004;Wax, Tezel, Kobayashi, & Hernandez, 2000;Yang et al., 2004). Elevation of pressure constitutes the gold standard to model ocular hypertension in vitro and has gained increasing attention in recent years. Elevated hydrostatic pressure (EHP) systems provide remarkable information on cell apoptosis, elastin synthesis, cell migration, and production of cell adhesion molecules. EHPs have provided new insights into the molecular and cellular mechanisms of glaucoma.
Recently, lab-on-a-chip (LOC) technology has been developed. This approach entails a simpler and smaller analysis platform, resulting in better temporal and spatial control of local cellular microenvironments, passive and active cell handling, less consumption of reagents, faster test results, economizing study logistics, and energy savings (Balijepalli & Sivaramakrishan, 2017). With the goal of transferring costly laboratory equipment onto small, user-friendly, easily replicable chips, LOC technology has dramatically altered many fields such as medicine, biochemistry, and biotechnology.
The aim of the present study was to design and establish a glaucoma-on-a-chip (GOC) model consisting of an EHP system coupled to a microculture system using purified primary rat RGCs. To achieve this aim, we designed pressure chips in which pressure could be modulated with the added benefits of higher speed, greater precision and finer control than previous EHP systems. We studied RGC survival under normal hydrostatic pressures (NHP) versus EHP and also compared survival of these cells when treated with a neuroprotective growth factor (brain-derived neurotrophic factor, BDNF) or a mimicry peptide, named RNYK, as a putative agonist of neurotrophic tyrosine kinase receptor type2 (NTRK2) . BDNF is a well-known neuroprotectant that induces signaling through the high-affinity neurotrophin receptor (NTRK2) for neuronal survival (Kaplan & Miller, 2000). However, there are several clinical concerns about its therapeutic applications for neurodegenerative diseases. Premature BDNF can bind to the low-affinity neurotrophin receptor (p75NTR) that paradoxically mediates neuronal apoptosis (Carter, Kaltschmidt, Kaltschmidt, & Offenhauser, 1996;Roux & Barker, 2002). Therefore, at high doses and in long-term delivery, BDNF may not increase cell survival or reverse neurodegeneration. Recently, we have demonstrated that RNYK as a novel BDNF mimetic can be an alternative to circumvent these problems (Nafian, Rasaee, Yazdani, Daftarian, et al., 2018;. RNYK has been selected from a phage-displayed random peptide library with high-affinity binding to NTRK2 and neurotrophic activities. RNYK represses neuronal apoptosis in a NTRK2-specific manner, while having minimal lethal interactions with p75NTR. This activity has been confirmed with equal efficacy to or even better than BDNF.

| Design of a three-layered chip
In order to induce a high pressure environment to simulate glaucomatous conditions, several three-layered EHP models were designed where u is the velocity vector, p is the pressure of injection, ρ is the density of media, and µ is the dynamic viscosity. The no-slip boundary condition was assumed at the walls. Constant velocity was specified at the inlet. The outlet was assumed to be at atmospheric pressure.
After the flow reached a steady-state condition, gas was assumed to be transported only by diffusion so that gas distribution was estimated using Fick's law (FL). Gas was released at the inlet at a specific concentration (C 0 ), which was monitored after a certain sampling time period. The transient two-dimensional mass transport equation is as follows: where C is the gas concentration in the bulk, and D is the diffusion coefficient.
A novel low-cost fabrication method was established to produce the three-layered chip entirely from poly-methyl methacrylate (PMMA, Cho Chen Acrylic Co. Ltd). PMMA is the most common member of the acrylic family. The optical clarity and ability to resist environmental stress make PMMA ideal for replacing glass in light transmission applications. PMMA biocompatibility has been established through years in medicine and ophthalmology (Frazer, Byron, Osborne, & West, 2005).

| Fabrication and testing of the threelayered chip
A laser system (fiber laser, PNC laser Co. Ltd) was used to cut polymerized PMMA sheets of about 10 mm thickness for layer B and 2 mm for layers A and C. The laser processing system was equipped with a programming system (NC system), an automatic programming combining computer and CAD technology. When using the laser for fabricating channels and wells on PMMA sheet, the laser power and scanning speed were considered as main factors. Several tests were performed to explore the correct correlation between laser power and scanning speed. A 25 W-laser power and a 300 mm/s laser scanning speed were Schematic of a designed chip in AutoCAD software. (a) Threedimensional design of layers a (1), b (2), and c (3) containing 12 wells (4), feeding ports (5), and microchannels (6), 2 main channels (7) and 1 single gas inlet (8). Layer (a) consisted of one feeding port per well and one main gas inlet. Layer B involved 12 hexagonal wells and microchannels. Layer (c) provided a surface area for cell cultures. Layers were fabricated (b) and assembled (c) to establish a complete chip selected to produce main channels and sub-channels on layer B. In addition, 20 W and 45 W-laser power were applied to cut PMMA sheets 10 and 2 mm in thickness respectively, at 100 mm/s laser scanning speed to make the wells and ports in layers A and B.
A thermal-solvent method was used to attach the three layers of any single chip. This method is based on dissolving PMMA in a thin layer of a solvent (isopropyl alcohol) between two PMMA substrates at 75°C for 10 min (Bamshad, Nikfarjam, & Khaleghi, 2016).
Adhesion of PMMA substrates involved the following steps. First, the layers cut by the laser were rinsed in deionized water for 5 min to remove tiny debris and dried at 40°C for 10 min in an oven. Next, the solvent was added between layers B and C, which were aligned and fixed using paper clamps to avoid entrapment of air bubbles between the binding surfaces. The attached sheets were heated and finally treated with air plasma. Radiofrequency discharged plasma was generated by a dinner plasma generator from Dinner CO. at 13.6 MHz frequency and 20 W power for 10 min. The base pressure of the plasma reactor was increased to 0.35 mmHg after gas feeding.
After attaching layers B and C, layer A was attached using the same method described above to complete the three-layered chip.

| Purification of RGCs and primary culture in the chip wells
Forty-eight hours before the experiment, chip surface was modified using plasma jet and exposed to UV light for 1 hr to assess successful sterilization. Then, 500 µl of PDL solution, 10 µg/ml, was added to each well for 1 hr. The wells were washed three times with sterile water and completely dried. The day before the experiment, 500 µl of laminin solution (5 µg/ml, Sigma-Aldrich) prepared in Neurobasal medium was added to each well and incubated overnight at 37°C (8% CO 2 ). The laminin/Neurobasal solution was replaced with 500 µl of "retinal ganglion cell media" which was prepared as fol-

| Flow cytometry and immunocytochemistry of RGCs
In different isolation steps, cells were stained with mouse anti-

| Measurement of cell viability
Retinal ganglion cells were seeded at an initial density of 10 4 cells per cm 2 in two sets of 12-well chips prior to the pressure experiment. RGCs were exposed to high hydrostatic pressures in a time-  These experiments were repeated for at least three times to check the reproducibility of biologic effects. All measured variables were described as mean ± SD.

| Statistical analysis
Cell viability in BDNF/RNYK treatment groups was compared with that of untreated controls under NHP and EHP. The GraphPad Prism 7.0 statistical software was employed (RRID: SCR_002798).

F I G U R E 3
Phase-contrast images of SH-SY5Y cells. (a) Cells phenotypically changed from N-type to S-type by ATRA treatment. (b) Relatively greater number of differentiated cells oriented and aligned on different membranes compared to the naked surface as a negative control. PDL/ laminin provided a physiologically optimal environment for neuronal adhesion and expansion. Scale bars: 50 µm Significance was determined at p < .05 level using two-way ANOVA with Holm-Sidak's multiple comparison test.

| Testing of the three-layered chip
The three-layered chip was designed to allow inflow of cell culture media from feeding inlets in layer A into the wells in layer B adjacent to layer C which acts the culture surface at the bottom of the wells ( Figure 1). We fabricated stand-alone feeding inlets for each well to let treat culture cells within the wells in an individual manner. We initially devised two feeding ports per well in layer A (the results not shown). The drawback to this design was inappropriate surface tension leading to an uneven distribution of the culture media. We then shifted to a single feeding port at one extreme of the well and placed a gas inlet on the other extreme to allow uniform liquid distribution.
Locating the feeding port at the extreme end of the well reduced shearing stress and avoided cellular detachment and necrosis in the mid-region of the well.
Based on simulation results, a hexagonal (divergent-convergent) well performed better than a rectangular well. If the cells are seeded into a rectangular well, they mainly localize in the central zone due to maximum velocity there influenced by boundary layers related to sidewalls (Figure 2a, red curve).
Whereas, velocity profile became more even in a divergent-convergent input due to the reduction cross-sectional area and mitigation of boundary layer effects (Figure 2b, blue curve), resulting in a homogeneous flow velocity and cell distribution across the well at the feeding and seeding times, respectively. We designed 12 wells in each chip for simultaneously RGCs treatment with different compounds per well without any mixing of culture media or cells from one well to another via the gas channels during feeding process. The gas channels necessarily had to be interconnected to provide the same level of pressure and gas conditions to all wells within a chip. A single gas inlet was located for each chip which led to two separate main channels with six branching sub-channels (Figure 2c).
The bottom surface of the wells (layer C) was treated by air plasma to chemical modification and functionalization. This surface treatment increases surface roughness, amount of oxygen atoms, and functional hydrophilic groups (and as a result the number of polar interactions), especially the electron-donor parameters. These modifications enhance cellular adhesion which is fundamental for cell growth, migration, and differentiation.

| Extracellular microenvironment model on modified PMMA surface
All-trans-retinoic acid was added to the culture media to differentiate SH-SY5Y neuroblastoma cells into more mature, neuron-like cells. ATRA is a powerful growth inhibitor but promotes normal cellular differentiation. This low-cost procedure is easily performed as

| Characterization of RGCs
The steps of RGC isolation by MACS technique are summarized in

| Measurement of cell viability
One set of the chips was connected to a gas tank to apply EHP at 33 mmHg, while another chip was exposed to NHP at 15 mmHg. As summarized in Figure 7a, the EHP system consisted of a compressed 8% CO 2 /92% air gas tank, which was placed just outside the incubator and adjusted with both regulators for normal and ultra-low delivery pressure. Polyurethane hoses (4/8″ ID) were used to feed the gas mixture into the pressure chip through a port at the back of the incubator. A digital pressure gauge (Ashcroft, DG25; 0.5% of 0-15 psi span accuracy, ±0.1 mmHg sensitivity; relative to atmospheric pressure) was used to monitor the input pressure value. Pneumatic fittings were placed into the inlet/outlet circuits to provide tight control over gas flow. During loading, the pressure was also monitored by a sensor in the chip and kept at the 33 mm Hg with the accuracy of ±0.1 mmHg (data not shown).
Untreated RGCs decreased in a time-dependent manner when elevated hydrostatic pressure was applied in the chip (Figure 7b1).  (Figure 7b2). RNYK provided slightly more stability than F I G U R E 5 CD90.1 immunostaining of isolated rat RGCs by MACS. (a) After 3-day normal culture, RGCs continued to express the specific marker CD90.1 (green). Nuclear staining was performed using DAPI (blue). (b) The two-color merged image was produced by overlaying the original CD90.1 and DAPI images, which indicates colocalization of the axonal connections between multiple RGCs against their nucleus of cell bodies. Scale bars: 100 µm BDNF over time. 93,89,87,and 86 percent viability under NHP compared with 67,66,63,60,and 58 percent under EHP at 6, 12, 24, 36, and 48 hr (p < .0001), respectively (Figure 7b3). BDNF and RNYK treatments induced separately an approximate twofold decrease in RGC death rates as compared to untreated cells under both NHP and EHP condition.
In the present study, we designed a glaucoma-on-a-chip system that provides continuous hydrostatic pressure with ±0.  (Nickells, 2012). Our results demonstrated that BDNF can maintain neuronal viability compared with untreated group under NHP. RNYK also indicated neuroprotective activities at an optimum concentration of 5 ng/ml compared with BDNF at 50 ng/ml. This observation is in agreement with our previous studies showing that RNYK could prevent neuronal degeneration of ATRA-treated SH-SY5Y with equal efficacy to or even better than BDNF (Nafian, Rasaee, Yazdani, Daftarian, et al., 2018;.
The negative impact of neurotrophin deprivation was compared with the insult of high pressure. In "untreated RGCs under EHP," the increase in pressure further added degenerative changes by a mechanical stress and strain in parallel to the deprivation of neurotrophic signals. Compared to normal pressure, BDNF and RNYK showed significant roles under EHP by moderating the detrimental effect of elevated pressure. They separately induced an approximate twofold decrease in RGC death rates as compared to untreated cells.
In light of the findings obtained, the identified peptidomimetic compound is significant as a promising neuroprotective agent for glaucoma treatment. Also, small size of RNYK might be an advantage for drug design and synthesis in future. It is clear that in vitro models may never substitute animal studies, but they are important tools in preclinical studies. Glaucoma-on-a-chip offers the advantages of allowing controlled experimental conditions, preliminary targeting of a specific cell type or pathway involved in glaucomatous damage.
More studies are needed to develop models to study RGC neurodegeneration and neuroprotection by putative agents such as neurotrophins, peptides, and other small molecules prior to assessment in animal models.

ACK N OWLED G M ENT
We are thankful to Dr Gharavi and Dr Tafreshi for technical support.

CO N FLI C T O F I NTE R E S T
The authors declare that no conflict of interest exists.

AUTH O R CO NTR I B UTI O N S
Fatemeh Nafian and Babak Kamali Doust Azad have made substantial contributions to conception, design, and acquisitions of data.
Mohammad Javad Rasaee, Narsis Daftarian, and Shahin Yazdani contributed to the analysis of data. Shahin Yazdani has been involved in writing the manuscript and given final approval of the version to be published. All authors have participated sufficiently in the work to take public responsibility for the content and agreed to be accountable for all aspects of work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

PE E R R E V I E W
The peer review history for this article is available at https://publo ns.com/publo n/10.1002/brb3.1799.

DATA AVA I L A B I L I T Y S TAT E M E N T
The data that support the findings of this study are available from the corresponding author upon reasonable request.