Tissue-engineered anterior segment eye cultures maintain intraocular pressure within a normal range

Grant information: ​National Eye Institute K08EY022737 (NAL); Initiative to Cure Glaucoma of the Eye and Ear Foundation of Pittsburgh (NAL); Wiegand Fellowship of the Eye and Ear Foundation of Pittsburgh (YD); P30-EY08098 (NAL); Department grant by Research to Prevent Blindness (NAL); an unrestricted fellowship grant from the Xiangya Hospital of Central South University (CW). Glaucoma is a blinding disease largely caused by increased resistance to drainage of fluid from the eye’s anterior chamber, resulting in elevated intraocular pressure (IOP). A major site of fluid outflow regulation and pathology is the trabecular meshwork (TM) at the entrance of the eye’s drainage system. We aimed to characterize the structural and functional properties of a newly developed tissue-engineered anterior segment eye culture model. We hypothesized that the repopulation of a decellularized TM with non-native TM cells could restore aspects of normal


Introduction
Glaucoma is a progressive optic neuropathy that affects over 70 million people worldwide and can cause irreversible blindness 1 . Clear fluid in the eye's anterior chamber, called aqueous humor, is produced within the confines of the globe at 2 to 3µL/min 2,3 . In healthy eyes, aqueous humor outflow is in equilibrium with production but in primary open angle glaucoma, an increased outflow resistance elevates the intraocular pressure (IOP) 4,5 . Intentionally elevating IOP in primate studies leads to glaucoma, while lowering it prevents it 6 . IOP remains the only clinically relevant factor that can be altered to reduce glaucomatous retinal ganglion cell death and vision loss 7,8 . As a primary site of outflow pathogenesis, the trabecular meshwork (TM) is a target of great therapeutic interest. The TM is a dynamic, multilayer, filter-like structure that responds to environmental signals like mechanical strain and shear forces with cytoskeleton and extracellular matrix (ECM) changes to maintain a normal IOP 9-13 . Simple in vitro TM cultures lack a pressure gradient and flow, defining features of glaucoma. In vitro TM perfusion models that use thin, layered scaffolds have been constructed to address these shortcomings 14,15 . However, these models can still not replicate the complex 3D structure of ECM beams and fibers of native TM. Their translational impact is often limited by cell physiology that is different compared to a complex organotypic substrate which strongly resembles the normal in vivo physiology [16][17][18][19] . Ex vivo perfused anterior segment models from a range of species have been used to examine and manipulate the TM function [20][21][22][23][24][25][26] because there is a scarcity of whole human donor eyes for research. In contrast, TM cells from corneal rims used for corneal transplantation, TM-removal surgeries, and TM cell lines, are more readily available 27,28 and could lend themselves to scalable studies when seeded into decellularized anterior segment scaffolds from pig eyes. As done in other tissues in research [29][30][31][32][33][34] and medical applications [35][36][37][38] , removal of cells from ECM can provide a bioartificial scaffold for recellularization with cells of choice. We had previously described a freeze-thaw protocol for this purpose 39 , but in pilot experiments detected occasional nuclear debris when we developed a protocol for recellularization. The surfactant sodium dodecyl sulfate (SDS) has been used to remove cellular material in various ECM-rich tissues, while hallmark TM ECM components like collagen, elastin, and laminin, were mostly well conserved 29,[40][41][42] .
We hypothesized that decellularized porcine anterior segments 39 can be repopulated with non-native TM cells to maintain IOP within a physiologic range. We seeded the eyes with porcine TM cells to eliminate species differences in this feasibility study and to obtain fresh cells with high viability. The porcine anterior segment scaffolds described here are accessible, storable, biocompatible, free of porcine cells, and can be seeded with transplanted cells.

Scaffold production
After cycles of freeze-thaw and perfusion of surfactant (FT+PS, Fig. 1a, 1a'), no cells or nuclear debris could be detected by DAPI staining (Fig 1a''). The ECM structure was well-preserved. After FT and agitation in surfactant (FT+AS, (Fig. 1b, 1b')), some remnants of nuclear debris could be made out by DAPI staining in the TM most distal from the anterior chamber ( Fig. 1b''). FT+AS samples on days 1, 2, and 5 days looked similar. FT alone (Fig. 1c) destroyed cells but could show displaced nuclear material in the mid and distal TM (Fig. 1c' and c''). Untreated controls (Fig. 1d) had a normal TM cellularity (Fig. 1d') and an even DAPI staining pattern. Here, DAPI was limited to nuclei within cells instead of the diffuse staining in FT+AS (b'') and FT (c''). Mean DAPI fluorescence in FT+PS was much lower than in controls (p = 0.010). Despite a different histological appearance as described, the mean DAPI fluorescence intensity was relatively similar in FT+AS (b''), FT (c'') and controls (d'', FT+AS vs controls: p=0.575, FT vs controls: p=0.387). Because FT+PS had a well-preserved ECM but were devoid of cells or nuclear debris, we used FT+PS scaffolds for all subsequent experiments.
IOP remained within a normal range throughout the perfusion-decellularization process without evidence of any abnormal physical stress on the ECM (Fig. 2, n = 15). After 24 hours of washing, scaffold IOP remained stable (within a 0.6 mmHg range) through the remainder of the process. Washed scaffolds appeared grossly normal. 24 scaffolds were produced. Eight IOP recordings were lost due to hardware error. One eye was contaminated and not included in the IOP analysis.

Scaffold repopulation and IOP
We transduced robust CrFK cells with an eGFP expressing lentiviral FIV vector, enriched them by fluorescence-activated cell sorting (FACS), and seeded them onto scaffolds to establish protocols for repopulation. The fluorescent signal of eGFP-expressing CrFK cells could be visualized at all time-points (24, 48, and 144 hours) and localized in the TM region. Occasional fluorescent cells could be seen on the corneal endothelium at 24 hours but not at subsequent time-points. Fluorescence was at times seen where the ciliary body had been attached. The TM was significantly brighter than other regions by an average of 90.1 ± 7.5% (p <0.001, Fig. 3).
The depth of cell infiltration into the TM was measured and compared to control at each time-point (supplemental material 1). The average depth of cells in the TM increased over time and peaked day four (Fig. 4, Table 1). CrFK cell depth at 24 and 48 hours was significantly lower than control (p < 0.05) while the depth of TM cells at 96 and CrFK cells at 144 hours was not significantly different (p = 0.19, 0.06). At 24 hours post-seeding, some cells could be seen histologically on the corneal endothelium, matching occasional cells fluorescing in Fig. 3. Cells migrated in the direction of outflow over time. Additionally, cellularity increased with time ( Table 1, n= 3458 nuclei measured total). After 96 hours in culture, many cells had large nuclei with euchromatin, indicating active transcription during infiltration. No cells were seen in the region of the angular aqueous plexus, a region similar to Schlemm's canal and proximal collector channels in primates 43 .
A stable IOP baseline was achieved after 24 hours (Fig. 5). No significant difference was found between decellularized (D) and scaffolds reseeded (RS) with TM cells at baseline (D= 11.8±0.5 mmHg, RS= 12.4±0.5 mmHg, p= 0.40). There was no difference between D and RS after seeding group RS (D= 9.2±0.4 mmHg, RS= 8.5±0.2 mmHg, p= 0.07). In contrast, scaffolds seeded with CrFK cells had a higher IOP that was quite unstable as indicated by a large standard deviation as a measure of the amount of variation (27.0±17.3 mmHg (avg±SD)) when compared to TM cells (8.5±2.7 mmHg (avg±SD)). At 48 hours post TM reseeding in RS or sham procedure in D, the infusion rate was doubled from 3 to 6 µL/min to challenge the TM's IOP maintenance response. One normal control eye developed a leak and had to be removed, reducing the count to n=7. RS maintained IOP within a normal range that was not different from normal control eyes (p > 0.05, n=8), while D became hypertensive (RS= 13.7±0.4 mmHg, n=8, D= 35.2±2.2 mmHg, p < 0.0001, n=8, Fig. 5). The IOPs of RS and D after challenge were significantly different from their respective IOPs in the reseed/sham phase (both p < 0.001, average increases of 8.05 and 25.04mmHg, respectively). The IOP of normal control eyes after the challenge was also significantly higher than before the challenge (p < 0.001). The IOP of challenged RS was not different from baseline (p = 0.06). Neither RS nor normal control eyes showed a homeostatic IOP decline 11 .

Discussion
In this study, we developed a tissue-engineered anterior segment model that has an outflow function with similarities to standard ex vivo and in vivo models. After an infusion rate challenge, anterior segments with repopulated TM maintained a physiological IOP while anterior segments with decellularized TM nearly tripled. Our results were similar to those observed in standard human anterior segment perfusion models with infusion rate doubling 11 and after IOP increase 44 .
TM cell numbers decrease gradually with age 45,46 . Both increased age and reduced TM cellularity are a major risk factor for the development of glaucoma. On the other hand, partially removing TM cells can also reduce IOP at least temporarily 47 . It has been previously shown that the presence of denuded TM beams can eventually fuse, associated with TM collapse, reduction of intertrabecular spaces for outflow, and pathologically elevated IOP 48-53 . Abu-Hassan et al. 23 recently showed that the transplantation of human TM cells or TM-like iPSCs into the anterior segment perfusion model can restore outflow after killing about 1/3rd of TM cells with saponin. In our study, all resident cells were removed to guarantee that any observed effect could only be caused by the transplanted cells but not by residual ones. We also washed any debris off scaffolds and avoided saponin to allow for a healthy TM cell function. While SDS + Triton X-100-mediated decellularization was chosen for this model in part due to demonstrated conservation of ECM components such as glycosaminoglycans (GAGs), collagen, elastin, and laminin 40,42,54,55 , hampered TM cell adhesion may reflect mild GAG loss. We chose porcine TM cells for reseeding because they can be rapidly generated and have a high viability.
Even though there was no Schlemm's canal, this model allowed us to isolate the effects of TM cells on outflow physiology in the TM region. The TM has a multitude of mechanisms at its disposal to adjust outflow that includes cytoskeletal as well as extracellular matrix changes. Important mediators include PGF2 , TGF , IL-1 / , TNF , nitric oxide, adenosine, and Rho kinase 9-13,56 among others. We focused on establishing TM ablation and repopulation and did not investigate the mechanisms of IOP regulation in this model. It is striking that only TM cells but not CrFK cells, a spontaneously immortalized kidney epithelial cell line, could maintain IOP. The increased IOP after perfusion challenge in decellularized scaffolds is likely caused by collapsing intertrabecular spaces and spaces within the angular aqueous plexus. In contrast to TM cells, CrFK cells were unable to maintain IOP within a physiological range. Instead, it was high and quite unstable as evidenced by a large standard deviation, a measure of the amount of variation. As a spontaneously immortalized epithelial kidney cell line, CrFK cells have a different migration speed and threshold for contact inhibition compared to primary TM cells. The porcine TM is nearly three times as thick as human TM. TM compression as a result of increased IOP can lead to narrowing and rarefaction of outflow pathways with a declining facility as established in bovine 57 and human eyes 58 . In contrast, when porcine eyes are subjected to ab interno trabeculectomy, outflow does increase, however 59-61 .
In the current study, fluorescent cells infiltrated the TM to a depth that was comparable to resident cells of normal controls. In pilot studies not presented above, we seeded 1 million cells, a cell number closer to that of a young, healthy TM 46 but this resulted in a low tissue-engineered TM cellularity. This is consistent with cell loss percentages in other tissue-engineered organs 29,62 . It is possible that seeded TM cells require sufficient cell-cell contact early in the process of meshwork infiltration as they cannot survive or proliferate otherwise. Cells settled extensively along and into the TM in most quadrants but not all. This supports the notion that flow through a comprehensively ablated anterior segment is still segmental. Similar to the eGFP expressing CrFK cells in this study, outflow can be visualized with tracers that do not readily pass through the TM but typically are 10 to 100 times smaller 59 or with fluorescent dyes that have a fast TM passage time and allow for precise computation of focal outflow 59,61,63 . Determining changes of segmental outflow more precisely with fluorescent dye canalograms will be in future studies that use this model.
Limitations of this study are the lack of additional canalograms to test for changing segmental flow patterns, confirmation of the predominant mechanism by which TM cells keep IOP within a normal range, and the use of porcine instead of human eyes that have a Schlemm's canal. Cells of the TM are a heterogeneous population of uveal, corneoscleral, juxtacanalicular, SC-like endothelial cells, and stem cells. For this reason, future characterization of TM cells in these scaffolds needs to include an analysis of gene expression and location to determine the similarities and differences between TE-cultures and normal anterior segments.
IOP in scaffolds with reseeded TM cells was not different from normal control eyes and remained within a normal range but this was not present in sham or CrFK seeded eyes. A homeostatic response with an IOP decline to pre-challenge levels was absent in both reseeded and normal control eyes. Our observation is different from prior publications that used gravity-driven perfusion with a fixed IOP. In those, an increasing facility was noted and termed "TM homeostasis" 11,64 to contrast "washout" [65][66][67][68][69][70] . Perhaps an infusion rate change is more difficult to compensate than an increase in pressure in gravity-driven systems.
Lastly, losing seeded cells is a common problem. For instance, Ott et al. report that 46% of cells seeded into bioartificial heart culture were lost already within 20 minutes 29 . Improving efficiency, for example, through the use of polymer gels will be necessary 71 .
TE-ex vivo cultures provide a new tool in the aqueous humor outflow research toolkit that is easy to generate and scale. In vitro and 2D perfusion cultures are harder to produce and/or lack biofidelic cues from a physiological multilayered, 3D environment, a hallmark of this tissue. Not only is generating the scaffolds straightforward, the ability to store them at -80℃ and install them rapidly adds flexibility. The expansion rates of primary cells harvested clinically during ab interno glaucoma surgery can vary considerably but stored scaffolds can be deployed instantly and at a scale that cannot be matched by human donor eyes.
In conclusion, we developed a fully decellularized porcine anterior segment scaffold and reseeded it with non-native meshwork cells. Tissue-engineered ex vivo cultures demonstrated localization of transplanted cells to the TM region, cell infiltration of ECM, and IOP maintenance ability after infusion rate challenge, each hallmarks of ex vivo culture structure and function. These scaffolds may allow testing of patient-derived iPSCs, human TM cell lines, or expanded human primary TM cells with a higher throughput and reduced need for human donor eyes.

Methods
A total of 49 anterior segments were prepared as previously reported 20,56,72 . Briefly, porcine eyes were acquired within 4 hours of sacrifice, decontaminated with povidone-iodine, and hemisected along the equator in an aseptic biosafety cabinet. The posterior segment, lens, and iris were carefully removed.

Scaffold production
Freeze-thaw -treated 39,73 anterior segments (FT) were sealed in an airtight container and cycled between -80 °C and room temperature two times to lyse all native cells 39,73 . Freeze-thawed scaffolds were perfused with culture media (DMEM supplemented with 1% FBS and 1X antibiotic/antimycotic) for 5 days (n=4). A time-course assay was conducted to determine the minimum time to produce decellularized scaffolds via immersion/agitation -mediated decellularization (IA). Freeze-thawed anterior segments were placed in SDS solution (0.01% wt/vol in PBS + anti-anti) for 1 day (n=3), 2 days (n=3), and 5 days (n=3) . Following SDS incubation, solutions were changed to 0.1% TritonX-100 for 24 hours and then a perfusion culture media wash for 48 hours. Samples were affixed to a vertical stage rotated at 20 RPM during each incubation. Fluid exchanges were performed every 24 hours.
For perfusion -mediated decellularization (P), we modified an existing matrix production protocol used in bioartificial heart construction 29 . Freeze-thawed segments were maintained via constant-rate perfusion 21 at 6µL/min with SDS solution for 24 hours, TritonX-100 solution for 24 hours, and perfusion culture media for 48 hours (n= 24). Untreated control samples were maintained for 5 days with perfusion culture media (n= 4). All scaffolds were stored at -80℃ and up to two months before reseeding. To avoid any confounding effects from prior experiments, none were reused.

Cell culture
Porcine TM culture was performed as done previously 20,74 . The TM was dissected away from anterior segments under an ophthalmic operating microscope (Stativ S4, Carl Zeiss, Oberkochen, Germany) and cut into 0.5mm 3 segments. Tissue pieces were cultured in T25 flasks containing OptiMEM (31985-070, Gibco, Life Technologies, Grand Island, NY, USA), supplemented with 5% FBS and antibiotic/antimycotic (15240062, Thermo Fisher Scientific, Waltham, MA, USA). The method used here to obtain cells from the porcine angular plexus naturally provides a non-homogenous cell population. Cells were passaged at 80% confluence and used for experiments at passages 2-4.
For initial scaffold biocompatibility testing and real-time tracking of seeded cells, we used a versatile CrFK cell-line (CRFK ATCC CCL-94) transduced with eGFP-expressing feline immunodeficiency viral (FIV) vector GINSIN 21,47 . Cells were transduced at a multiplicity of infection (MOI) of 5 transducing units (TU) per cell and enriched by fluorescence-activated cell sorting (FACS) after 7 days with a high fluorescence gate cut-off that eliminated all non-fluorescent cells. GINSIN transduced CrFK cells were maintained in 2.5% FBS DMEM with antibiotic and antimycotic and passaged at 80% confluence.

Live-cell tracking
Three million GINSIN transduced CrFK cells were seeded into scaffolds to determine biocompatibility and cell localization in real-time. A 20G cannula (BD PrecisionGlide 305176) attached to a 1mL syringe (Norm-Ject Tuberkulin Luer 4010-200V0) was connected to a 3 cm length of perfusion tubing. The inlet and outlet of each perfusion dish were disconnected from the perfusion system and reconnected to the 1mL syringes. A 200µL bolus of culture media was removed from anterior chambers through the outlet. A 200µL bolus of 3 million GINSIN-CrFK cells was slowly introduced into the anterior chamber. Cultures were gravity perfused at 15mmHg for 20 minutes. Dishes were positioned with corneas facing downward and perfusion culture was restarted.
TM was visualized from the underside of transparent perfusion dishes with an epifluorescence-equipped dissecting microscope (Olympus SZX16 with GFP filter cube and DP80 monochrome/color camera; Olympus Corp., Center Valley, PA, USA) and imaged at 24, 48, and 144 hours post-seeding. Mean fluorescent intensity of images was measured in four concentric regions (central cornea (1), peripheral cornea (2), TM (3), and sclera/ciliary body remnant (4), Fig. 3, bottom right) in Fiji 75 and compared. A uniform threshold was applied to fluorescence images to aid visualization in Fig. 3, right panels.

IOP maintenance challenge
Decellularized scaffolds were perfused for 24 hours with TM culture media before seeding to achieve stable baseline IOPs. Three million porcine TM cells were seeded as described above in RS (n=8) while D (n=8) received a sham bolus of cell-free TM culture media. Normal anterior segment cultures were used as controls (n=8). After 48 hours, the infusion rate was doubled from 3µL/min to 6µL/min to challenge the TM's IOP maintenance response.

Histology
Samples were fixed in 4% PFA for 48 hours, embedded in paraffin, and sectioned at 6 µm thickness. Sections were stained with hematoxylin and eosin (H&E) for morphological analysis and DAPI for DNA content. A specimen was analyzed from the center of both the superior and inferior portion of each anterior segment.

Statistical analysis
Fluorescence intensities of DAPI-labeled scaffold samples were compared to control. Live-cell eGFP fluorescence of the TM region was compared to all other locations. Cell depth at each time-point was compared to control. IOP recordings were sampled into 2-hour blocks, and D was compared to RS at each experimental phase (baseline, cell seeding, and challenge). Periods of 8 hours after media refilling were excluded from the analysis to allow for IOP stabilization. Statistical comparisons were conducted with a student's t-test in Python 3.6. Data are expressed as mean ± SEM unless otherwise noted. P-values < 0.05 were considered statistically significant.      After 48 hours, the infusion rate was doubled from 3 µL/min to 6 µL/min. Reseeded (blue) eyes maintained IOP within a normal range while IOP in decellularized scaffolds was significantly elevated (p < 0.0001, n= 8 for each). Error fields indicate SEM, ns= not significant, *p < 0.0001.