Abstract
Purpose.:
Retinal pigment epithelium (RPE) transplantation is a promising strategy for the treatment of dry age-related macular degeneration (AMD). However, previous attempts at subretinal RPE cell transplantation have experienced limited success due to poor adhesion, organization, and function on aged or diseased Bruch's membrane. Instead, cell-based strategies may benefit from a synthetic scaffold that mimics the functions of healthy Bruch's membrane to promote the formation of a functional RPE monolayer while maintaining metabolite exchange between the vasculature and outer retina.
Methods.:
This study evaluated the behavior of human RPE on nanopatterned porous poly(ε-caprolactone) (PCL) film as a potential scaffold for therapeutic transplantation. Fetal human RPE (fhRPE) was cultured on porous PCL, nonporous PCL, or Costar porous polyester transwells for up to 8 weeks and assessed using light microscopy, fluorescent microscopy, transepithelial resistance, quantitative PCR, ELISAs, and phagocytosis assays.
Results.:
fhRPE on porous PCL displayed improved markers of maturity and function compared with both porous polyester transwells and nonporous PCL, including pigmentation, increased cell density, superior barrier function, up-regulation of RPE-specific genes, and polarized growth factor secretion.
Conclusions.:
This study indicates that porous PCL is an attractive scaffold for RPE transplantation. In addition to being biocompatible with the subretinal space, porous PCL also allows for trans-scaffold metabolite transport and significantly improves RPE cell behavior compared to nonporous PCL or porous polyester transwells.
Age-related macular degeneration (AMD) is the leading cause of blindness among the elderly.
1 AMD is a multifactorial, progressive disease of the central retina that can be divided into two distinct categories, wet (choroidal neovascularization) and dry (geographic atrophy). One of the first pathological hallmarks of dry AMD is the loss of functional retinal pigment epithelium (RPE), which normally supports the survival and function of both photoreceptors and vasculature.
2,3 However in AMD, RPE anomalies lead to irregular behavior and atrophy that contribute to downstream photoreceptor degeneration. As a result, therapeutic RPE transplantation represents an attractive strategy for halting the progression of retinal degeneration by restoring normal RPE function prior to permanent vision loss.
4 Unfortunately, past attempts to transplant RPE have failed because allografted cells were introduced without addressing other pathological changes to the host microenvironment such as Bruch's membrane (BrM) thickening, cross-linking, and drusen accumulation.
5–9
Interestingly, some of the structural alterations to BrM have even been detected in dry AMD prior to RPE atrophy, suggesting that extracellular matrix changes occur soon after disease onset and likely contribute to downstream degeneration.
9,10 In vitro and in vivo studies have provided further evidence for the impact of age- and disease-related changes to BrM on RPE function. Healthy RPE seeded onto aged or diseased BrM in vitro show a dramatic decline in cell behavior, gene expression, and function.
11–15 In vivo experiments that injected RPE in suspension have yielded similarly poor results with minimal cell attachment, survival, and organization on the animal's native BrM.
15–17
Therefore, effective cell-based strategies will likely require the co-implantation of a BrM substitute and mature, well-organized RPE to enhance cell survival and function.
18,19 Although extracellular matrix components could be assembled to form this type of support, synthetic polymers may be more attractive due to reproducibility, ease of modification, and nonimmunogenicity.
20–23
Poly(lactic acid) (PLA) and poly(lactic-co-glycolic acid) (PLGA) were two of the first materials proposed for supporting RPE delivery
18,24 ; however, despite their classification as generally biocompatible,
25 both materials have been shown to cause widespread retinal degeneration when implanted into the subretinal space.
19,26 This lack of tissue-specific biocompatibility is likely due to rapid degradation of these polymers (typically 1–18 months
27 ), which leads to a buildup of acidic degradation products in the small subretinal space.
28 These products lower the microenvironmental pH leading to a major inflammatory response and associated retinal cell death. As a result, recent subretinal implant studies have focused on nondegradable or more slowly degrading polymers including poly(methyl methacrylate),
19,29 parylene, poly(ethylene glycol),
30 poly(ε-caprolactone) (PCL),
31 and poly(glycerol sebacate).
32
This study aimed to develop an RPE scaffold constructed from PCL, a material that has demonstrated high compatibility with the subretinal space of mice and pigs.
26,31 Although PCL, like PLA and PLGA, degrades via hydrolytic cleavage of ester bonds,
33 this occurs more slowly (typically over longer than 3 years
22,34 ) and produces degradation products that are comparatively less acidic.
27 These differences help to minimize local alterations in pH and any associated inflammatory response.
25 A microfabrication-based approach was used to pattern PCL films with pores that support metabolite transport. This method resulted in PCL scaffolds with full-thickness pores that were small (<1 μm), well controlled, and highly reproducible—characteristics that have been identified as beneficial for ocular tissue engineering scaffolds.
16,35 These porous PCL scaffolds were tested in vitro to determine the effect of cell culture substrate properties on RPE maturation, gene expression, and function.
Scaffolds were created using the pore casting process our group has previously described.
36 Briefly, photolithography and deep reactive ion etching were used to produce a silicon mold with sub-micron cylindrical features protruding from its top surface. The mold was then covered with a PCL solution and spun at high speed resulting in a solid PCL film. The film was then peeled from the mold to yield a thin scaffold with full-thickness cylindrical pores.
Scaffold morphology including pore size and total porosity was determined using scanning electron microscopy (SEM). Prior to imaging, scaffolds were coated with approximately 20 nm of gold/palladium using a Cressington 108 sputter coater (Cressington Scientific Instruments, Watford, UK) for 25 seconds at 20 mA with a distance of 3 cm. Images were then collected using a Zeiss Supra 55VP Field Emission SEM (Carl Zeiss, Oberkochen, Germany) at 5.0 kV and subsequently analyzed in Adobe Photoshop (Adobe Systems, San Jose, CA). Porous PCL scaffold thickness was determined using an Alpha Step 500 Profiler (KLA-Tencor, Milpitas, CA) and qualitatively confirmed by comparing the visible height of the silicon cylinders with and without an adherent PCL film.
Fetal human RPE cells (fhRPE) were purchased from Lonza Biologics (cat. no. 00194987; Allendale, NJ), expanded, and plated at passage 5 at high density (3 × 105 cells/cm2) on porous PCL (700-nm pores), nonporous PCL, and Costar polyester transwells (400-nm track-etched pores; Corning, NY). Nonporous PCL and polyester transwells were used as controls for material and porosity, respectively. PCL samples were mounted on transwell supports devoid of their original membrane. All substrates were coated with 300 μL of 10 μg/mL laminin from Engelbreth-Holm-Swarm murine sarcoma basement membrane (cat. no. L2020; Sigma-Aldrich, St. Louis, MO) in PBS for 2 hours prior to cell culture. An ELISA confirmed that statistically similar amounts of laminin adsorbed to the surface of all substrates (data not shown), suggesting that changes in cell behavior are a direct result of culture substrate rather than an indirect function of laminin adsorption. After seeding, cells were maintained at 37°C, 5% CO2 in serum-free, growth factor-free RtEBM media (Lonza, cat. no. 0019540) supplemented with 1% (v/v) GlutaMAX (Invitrogen, Carlsbad, CA) and 1% (v/v) penicillin-streptomycin (Lonza). Media were changed twice per week for up to 8 weeks.
All antibodies and reagents used in this study were purchased from Invitrogen and Sigma-Aldrich, respectively, unless otherwise noted. Cells were fixed for 10 minutes in 4% paraformaldehyde and permeabilized with 0.1% Triton X-100 in PBS for 5 minutes. After washing, samples were incubated for 2 hours in a blocking buffer containing of 3% (v/v) goat serum and 2.5% (w/v) bovine serum albumin in PBS. Samples were incubated overnight at 4°C in blocking buffer also containing 2.5 μg/mL rabbit-raised antibody for zona occludin 1 (ZO-1), a tight junction-associated protein. The following day samples were incubated in blocking buffer containing 6.7 μg/mL AlexaFluor 488 goat anti-rabbit and 10 μL/mL 4′,6-diamidino-2-phenylindole (DAPI) for 1 hour and then imaged using an Axioskop MOT 2 (Carl Zeiss Meditec, Dublin, CA). Cell size was evaluated by manually tracing cell borders defined by ZO-1 staining in Adobe Photoshop.
Transepithelial electrical resistance (TER) was used to quantitatively assess fhRPE tight junction formation and scaffold coverage. TER was measured using an EVOM2 Voltohmmeter (World Precision Instruments, Inc., Sarasota, FL) as previously described.
36 Nonporous PCL film was not tested because the resistance of the substrate alone exceeded the reliable range of the instrument.
fhRPE RNA was isolated using RNA-bee solution (IsoText Diagnostics, Friendswood, TX) according to the manufacturer's protocol. One microgram of RNA was reverse-transcribed using Superscript III (Invitrogen). Real-time PCR was performed using FastStart SYBR Green Master mix with the LightCycler 480 Real-Time PCR System (Roche Applied Science, Indianapolis, IN). The primer sequences used are listed in
Table 1. Relative gene expression was determined using the delta-delta Ct method after normalizing sample loading with the housekeeping genes
GAPDH and
HPRT1.
Cell culture media conditioned for 72 hours was collected from the apical and basal chambers at 1 or 4 weeks and stored at −80°C until further use. Concentration of secreted vascular endothelial growth factor (VEGF) and pigment epithelium derived factor (PEDF) were measured using human ELISAs kits (R&D Systems, Minneapolis, MN, and BioProducts MD, Middletown, MD). A custom 19-analyte multiplex ELISA (Milliplex Map; EMD Millipore, Billerica, MA) for human cytokines and chemokines was used to measure fhRPE secretion of soluble epidermal growth factor (EGF), eotaxin, fibroblast growth factor (FGF)-2, CX3CL1 (fractalkine), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophage colony-stimulating factor (GM-CSF), CXCL1 (GRO), interleukin (IL)-1α, IL-1β, IL-1RA, IL-6, IL-8, IL-15, interferon γ–protein 10 (IP-10), platelet-derived growth factor (PDGF)-AA, PDGF-AB/BB, CCL5 (RANTES), transforming growth factor (TGF)-α, and tumor necrosis factor (TNF)-α. All conditioned media samples were run in technical duplicate.
Bovine eyes were obtained from Research 87 (Boylston, MA) within 6 hours of enucleation. Photoreceptor outer segments (POS) were then isolated based on a method previously reported.
37 Thirty million POS (100 per RPE cell) were loaded into the apical media of cells cultured for 4 weeks. After 16 hours of co-incubation, samples were washed extensively with PBS to remove POS not bound to RPE. Cells were then fixed with 4% paraformaldehyde, permeabilized with 0.1% Triton X-100, and immunostained with 2 μg/mL mouse anti-rhodopsin primary antibody (EMD Millipore) overnight at 4°C. The next day cells were incubated with 10 μL/mL DAPI and 2 μg/mL AlexaFluor 594 rabbit anti-goat secondary antibody for 1 hour, washed, mounted onto glass slides, and imaged using an Axioskop MOT 2. The relative number of POS bound/internalized per cell was quantified by dividing the number of red fluorescent pixels above a threshold by the number of DAPI-stained nuclei in that area. The size of bound and phagocytosed POS in these images was also assessed using the “threshold” and “analyze particle” functions in ImageJ (National Institutes of Health, Bethesda, MD).
All values are expressed as the mean ± standard deviation unless otherwise noted with statistical significance determined using an unpaired two-tailed Student's t-test. Error bars on graphs represent standard error of the mean. TER measurements were taken for 18, 10, and 2 biological replicates per condition for 1, 4, and 8 weeks, respectively. ELISAs for VEGF and PEDF were performed on cultured media collected from six experimental replicates while multiplex analysis used five replicates. All other cell culture experiments were completed in biological triplicate.
SEM was used to quantify the average pore diameter and total porosity of the substrates used in this study. Track-etched polyester transwells, the current standard RPE culture substrate in the field, contained pores close to the manufacturer's quoted size of 400 nm distributed across the surface of the membrane. Although many of these pores were circular, there was also a subset of irregular pores that appeared to have been formed by multiple overlapping circular pores (
Fig. 1A, arrows), which is known to occur with a certain probability due to the track etching process.
38 Porous PCL scaffolds contained significantly larger pores (
P < 0.05), which were spaced evenly in a grid and covered a larger percentage of the surface (
P < 0.001). Pore size and total porosity of the porous PCL were higher than the polyester transwell control, but on the same order of magnitude (
Fig. 1B). Aside from the pores, the surfaces of all three substrates were smooth and unremarkable.
Bright field microscopy revealed little-to-no fhRPE pigmentation after 1 and 4 weeks of culture. However, after 8 weeks fhRPE cells displayed varying degrees of pigmentation depending on their culture substrate. Cells on porous polyester transwells displayed the least amount of pigmentation with few dark cells (
Fig. 2A). A slightly higher proportion of fhRPE on nonporous PCL appeared pigmented (
Fig. 2B), while the most pigmentation was observed on porous PCL (
Fig. 2C). In all conditions pigmented fhRPE appeared to be localized in foci interspersed amongst nonpigmented cells. fhRPE shape and size were determined using immunostaining. ZO-1 and DAPI staining indicated that cells on all substrates generally assumed the hexagonal morphology that is characteristic of RPE, with nuclei located toward their lateral edge (
Figs. 2D–F). Cell density was similar on polyester transwells and nonporous PCL, but significantly higher (
P < 0.001) on porous PCL (
Fig. 2G).
Epithelial barrier formation was assessed using tight junction staining (ZO-1) and TER measurements. ZO-1 staining of fhRPE cultured for 8 weeks on polyester transwells was generally diffuse, though faint localization was observed at cell borders (
Fig. 2D). Cells on nonporous PCL displayed moderate ZO-1 localization at cell–cell borders but again showed a considerable amount of diffuse staining (
Fig. 2E). In comparison, ZO-1 staining of fhRPE on porous PCL was more intense and continuous at the cell–cell border indicating the formation of mature tight junctions (
Fig. 2F). This improvement in tight junction formation was quantitatively confirmed by TER measurement.
Figure 2H illustrates the progression of cell-derived electrical resistance over the course of 8 weeks. At all time points other than day 0, fhRPE on porous PCL exhibited a significantly higher resistance than cells on polyester transwells (
P < 0.01); however, the statistics for time points greater than 28 days were not counted due to the limited number of biological duplicates used (
n = 2). The resistance of fhRPE on transwells increased slowly, whereas the resistance of cells on porous PCL increased rapidly and continued to increase through 8 weeks of culture at which point RPE cultured on polyester and porous PCL achieved TER values of 31 ± 2 and 182 ± 4 Ω·cm
2 , respectively.
mRNA analysis of fhRPE cultured for 4 weeks on each substrate revealed several differentially expressed genes associated with RPE differentiation, homeostasis, and function. While a majority of the genes tested did not display major differences between culture substrates (
Table 2), many genes critically involved in RPE visual or neurotrophic functions such as
RPE65,
RLBP1,
BEST1, and
SERPINF1 (
PEDF) were strongly up-regulated in cells cultured on porous PCL compared to polyester transwells (
Fig. 3). The antioxidant enzyme SOD2 was also down-regulated approximately 50% on porous PCL compared with polyester transwells (
Fig. 3A).
Secretion of the major fhRPE-produced growth factors, VEGF and PEDF, was characterized over time by ELISA. As expected, a significant increase in the secretion of both cytokines was observed during RPE maturation in all conditions (
Figs. 4A,
4B). In addition, both VEGF and PEDF secretion were highest on porous PCL, mimicking the results obtained at the mRNA level (
Fig. 3) and suggesting that the production of these growth factors are largely controlled at the transcriptional level. The cell culture substrate used also affected the polarization of growth factor secretion. fhRPE cultured on polyester transwells obtained higher apical media concentrations of VEGF and higher basal media concentrations of PEDF relative to the adjacent chamber (
Table 3). Alternately, cells on porous PCL assumed a secretion profile more characteristic of in vivo RPE, with higher levels of VEGF in the basal media and higher levels of PEDF in the apical media.
39,40
A multiplex ELISA was used to determine the production of inflammatory cytokines by fhRPE on each substrate in order to better evaluate the clinical potential of porous PCL scaffolds for RPE transplantation. Most of the cytokines analyzed were not detected in conditioned media, indicating minimal production (
Table 4). The analytes that were present above the minimum threshold for detection were EGF, CX3CL1, CXCL1, IL-6, and PDGF-AA, all of which were produced in similar amounts by fhRPE on polyester transwells and porous PCL.
Finally, we evaluated the effect of culture substrate on RPE phagocytic function by quantifying the number and size of bovine POS bound to or internalized by fhRPE at steady-state (
Fig. 5). Phagocytosis expressed as rhodopsin-labeled fluorescence per cell (analogous to POS bound/internalized per cell) revealed no statistically significant difference between fhRPE cultured for 4 weeks on any of the substrates (
Fig. 5D). However, POS fragments were significantly smaller and more numerous on porous PCL compared to polyester transwells (
P < 0.01,
Fig. 5E). For all conditions the average POS size on each substrate was within the range of individual rod (0.8–4.9 μm
2) and cone (1.1–9.6 μm
2) POS.
41,42
This study demonstrates that porous PCL scaffolds enhance fhRPE morphology, cell density, barrier formation, gene expression, and protein secretion compared to cells on nonporous PCL or porous polyester. Pigmentation has been widely used as an overt marker of RPE maturity
43 and also serves a functional role by absorbing light to minimize scatter for improved vision.
44 Therefore, increased pigmentation observed in fhRPE on porous PCL serves as an indication that these cells are more mature and functional, at least in terms of light absorption.
Immunohistochemistry revealed that fhRPE achieved a monolayer of cells with characteristic hexagonal morphology on all three substrates; however, tight junction staining was most intense at the cell–cell junction on porous PCL. Intense ZO-1 staining, like that seen on porous PCL, has been associated with the native macular RPE, while low-intensity staining, like that observed on porous polyester transwells, is characteristic of peripheral RPE.
45 fhRPE on porous PCL achieved an average cell density of 3317 ± 171 cells/mm
2, which was between the average cell density previously reported for the midperiphery (3002 ± 460 cell/mm
2) and fovea (4220–4980 cells/mm
2).
46,47 Comparatively, fhRPE on porous polyester transwells and nonporous PCL were less densely packed than on porous PCL and similar to the cell density found at the periphery (1600 ± 411 cells/mm
2).
46 As a result, RPE cultured on porous PCL are likely more suitable for submacular RPE replacement therapies that require densely packed, high-functioning cells similar to native central RPE.
fhRPE on porous PCL also displayed more complete and uniform tight junction staining indicating the formation of a well-developed barrier. This qualitative observation was reinforced by TER measurements, which demonstrated that RPE on porous PCL provided more resistance than cells on porous polyester transwells at every time point beyond day 0. Because barrier formation is required for many homeostatic functions including fluid transport, maintenance of the blood–retinal barrier, and polarized growth factor secretion,
44 this improvement on porous PCL may be especially critical for success in RPE-based transplantation therapies.
Gene expression analysis provided further indications of improved RPE maturity and function on porous PCL. Many genes associated with RPE differentiation, maturation, or function such as
EZR,
BEST1,
RPE65, and
RLBP1 were significantly up-regulated in cells cultured on porous PCL. Interestingly the enzyme SOD2, which acts as an antioxidant in the RPE,
48 was down-regulated on porous PCL suggesting that these cells may be under less oxidative stress than cells on either control material. Both VEGF and PEDF secretion levels were similar to what has been previously reported for fhRPE.
49 While changes in VEGF expression were rather minor and unlikely to result in major physiological differences, PEDF expression was highly up-regulated on porous PCL compared to polyester transwells. Because PEDF serves two functions as both a neurotrophic and anti-angiogenic factor,
3,50 up-regulation on porous PCL may indicate a superior ability to both support photoreceptors and maintain neuroretinal avascularity in cell-based AMD therapies.
The production of inflammatory cytokines by fhRPE on all substrates was very low compared to VEGF and PEDF. The inflammatory cytokines CX3CL1, CXCL1, and IL-6 were all produced in such minimal amounts relative to serum levels that they would be negligible in vivo.
51–53 fhRPE on all three substrates secreted PDGF-AA at a similar rate which was approximately 5-fold higher than what has been reported for quiescent fhRPE, but much lower than the level produced by fhRPE after TGF-β stimulation.
54 In general, this analysis suggests that culture on porous PCL is unlikely to significantly affect intrinsic inflammatory cytokine secretion by RPE.
Although the total steady-state fluorescence from bound and internalized POS was similar on all substrates, fluorescent fragment size was significantly decreased on porous PCL compared to polyester transwells. After POS discs are shed by photoreceptors and internalized by RPE they enter the phagolysosome where they are degraded into their component proteins, peptides, and lipids. Therefore, the presence of smaller, more numerous fluorescent fragments in fhRPE suggests that POS on porous PCL are further along in the process of degradation due to more rapid processing. This type of improvement in POS degradation on porous PCL would be especially beneficial from a translational perspective because inadequate processing has been cited as a possible cause of material accumulation in BrM eventually resulting in RPE atrophy.
14
Overall, the results of this study indicate that porous PCL enhances fhRPE maturation compared to both nonporous PCL and polyester transwells. Improvements in RPE phenotype and function were observed for all analysis methods used except phagocytosis, which displayed minor, but statistically insignificant improvements on porous PCL. Interestingly, it appears that both substrate composition and porosity have an effect on fhRPE behavior. fhRPE on porous PCL exhibited improved markers of maturation and function compared with cells on porous polyester, suggesting that PCL is a superior RPE substrate for in vitro culture. However, RPE on nonporous PCL did not experience the same improvements as cells on porous PCL, suggesting that substrate porosity also has a beneficial effect on behavior. BrM porosity and permeability decrease with age or disease, suggesting that impaired basal RPE flux may play a role in RPE dysfunction.
55,56 Although reduced flux in vivo is likely due to BrM thickening, cross-linking, or high lipid content, nonporous artificial RPE substrates are likely to have the same effect in vitro, leading to reduced cell function. Instead, substrates that more closely approximate the porosity of healthy native BrM, such as porous PCL, are likely to achieve appropriate basal RPE flux and elicit improved cell behavior or function.
As a result, porous PCL, which benefits from both material composition and porosity, is an attractive scaffold for RPE maturation and delivery in cell-based dry AMD therapies. In addition to its subretinal biocompatibility
26,31 and appropriate mechanical properties for surgical handling, we have now shown that porous PCL also improves RPE cell maturation, which may be critical in reestablishing normal tissue function. Future studies will aim to deliver an established monolayer of RPE on porous PCL into the subretinal space of animals with nonfunctional or atrophic RPE to determine the in vivo function of transplanted RPE.
The authors thank Jinmei Wang for POS isolation.
Supported by the National Institutes of Health (NIH) through the NIH Director's New Innovator Award Program, 1-DP2-OD006649 (MSG), and award number 5-T32-EB006359-05 from the National Institute of Biomedical Imaging and Bioengineering (KM). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health or National Institute of Biomedical Imaging and Bioengineering.
Disclosure: K.J. McHugh, The Charles Stark Draper Laboratory, Inc. (E); S.L. Tao, The Charles Stark Draper Laboratory, Inc. (E), CooperVision, Inc. (E), P; M. Saint-Geniez, P