Avoiding the Pitfalls of siRNA Delivery to the Retinal Pigment Epithelium with Physiologically Relevant Cell Models

Inflammation is involved in the pathogenesis of several age-related ocular diseases, such as macular degeneration (AMD), diabetic retinopathy, and glaucoma. The delivery of anti-inflammatory siRNA to the retinal pigment epithelium (RPE) may become a promising therapeutic option for the treatment of inflammation, if the efficient delivery of siRNA to target cells is accomplished. Unfortunately, so far, the siRNA delivery system selection performed in dividing RPE cells in vitro has been a poor predictor of the in vivo efficacy. Our study evaluates the silencing efficiency of polyplexes, lipoplexes, and lipidoid-siRNA complexes in dividing RPE cells as well as in physiologically relevant RPE cell models. We find that RPE cell differentiation alters their endocytic activity and causes a decrease in the uptake of siRNA complexes. In addition, we determine that melanosomal sequestration is another significant and previously unexplored barrier to gene silencing in pigmented cells. In summary, this study highlights the importance of choosing a physiologically relevant RPE cell model for the selection of siRNA delivery systems. Such cell models are expected to enable the identification of carriers with a high probability of success in vivo, and thus propel the development of siRNA therapeutics for ocular disease.


Introduction
Age-related macular degeneration (AMD), diabetic retinopathy, and glaucoma are common age-related eye conditions and major causes of visual impairment and blindness. According to the World Health Organization, the number of people affected by AMD and glaucoma will increase from 195.6 million in 2020 to 243.4 million in 2030 due to a variety of factors such as ageing population, The effect of cell differentiation, including the role of melanin, on the siRNA knockdown efficiency is assessed.

Polyplexes
Cationic micelles were prepared by dissolving the amphiphilic polypeptide block copolymer PBE 30 -b-PK 30 [32] in nuclease-free water at a stock concentration of 1 mg/mL. Then, the siRNA and the micellar solution were both diluted in 10 mM HEPES (pH 7.2), mixed together, and incubated at room temperature for at least 20 min to form polyplexes. The used charge ratio (N/P) was 8/1. The charge ratio was the molar ratio of positive/negative charges originating from the amines of the carrier and the phosphate of the siRNA strand, respectively (N/P).

Lipoplexes
Cationic liposomes composed of DOTAP and DOPE were prepared by the thin lipid hydration method described by Ruponen et al. [33]. Shortly, the lipids were mixed at a molar ratio of 1:1 (3.2 mM) in chloroform. The organic solvent was evaporated to obtain a lipid film, which was hydrated in nuclease-free water at +30 • C for 30 min. After 2 h at room temperature, the liposome solution was sonicated for 5 min at +30 • C and extruded several times through a 100 nm polycarbonate membrane. The cationic liposomes were stored at +4 • C and used within three months of preparation.

Lipidoid-siRNA Complexes
Lipid-like molecules (lipidoid) were synthesized by a ring-opening reaction between 1,2-epoxytetradecane (ETD) and the amines as previously described [35]. Briefly, spermidine and ETD at a molar ratio of 1:4 were added into a 2 mL glass vial, then the mixture was stirred at 90 • C and protected from light for 2 days. The product was purified by silica gel column chromatography with a gradient elution from CH 2 Cl 2 to 75:22:3 CH 2 Cl 2 /MeOH/NH 4 OH. The transparent, pale yellow, oily product was stored at −20 • C. DSPE-PEG/lipidoid/cholesterol nanoparticles were prepared via the nanoprecipitation method and purified by centrifugation [36]. Briefly, the synthesized lipidoid, cholesterol, and DSPE-PEG were dissolved in 100% ethanol at concentrations of 100, 25 and 100 mg/mL, respectively, and combined at a weight ratio of 2:2:1. The mixture was then injected quickly into acetate buffer (200 mM, pH 5.4) while stirring to form empty nanoparticles. The DSPE-PEG/lipidoid/cholesterol-siRNA complexes (briefly, lipidoid-siRNA complexes) were formed by mixing 10 pmol of siRNA (for a 50 nM siRNA dose) with the empty nanoparticles in nuclease-free water at an RNA to nanoparticle ratio of 1 µg to 7.5 µL. After gentle mixing with a pipette, the solutions were incubated for 20 min at room temperature. The lipidoid-siRNA complexes were used immediately.

Dynamic Light Scattering (DLS) and Nanoparticle Tracking Analysis (NTA)
The hydrodynamic diameter of the siRNA complexes was determined by DLS, using a Zetasizer APS (Malvern Panalytical, Spectris plc, Egham, Surrey, UK) with a nominal 5 mV He-Ne laser operating at a 633 nm wavelength. The refractive index of the lipid-based carriers and polymer-based carriers was 1.33 and 1.450, respectively, and the viscosity was 0.8872 cP at 25 • C. For each sample, three separate measurements were conducted with 20 runs each. The size of the lipidoid and commercial complexes was analyzed by NTA using a NanoSight LM10 system (Malvern Panalytical, Spectris plc, Egham, Surrey, UK) fitted with a high-sensitivity cCMOS camera and a 405 nm laser. Each sample was appropriately diluted in PBS before the measurement. Videos of 60 s were recorded and analyzed with the NTA software (version 3.1, build 3.1.45). The concentration ranges of the particles were between 10 8 and 10 9 particles per mL. Three measurements per sample were taken.

Human Retinal Pigment Epithelial Cells (ARPE-19)
ARPE-19 cells from ATCC (Manassas, VA, USA) were cultured in a growth medium composed of Dulbecco's modified Eagle's medium/Nutrient Mixture F12, supplemented with 10% fetal bovine serum (FBS), 2 mM of L-glutamine, 50 U/mL of streptomycin, and 50 U/mL of penicillin. The cells were maintained at +37 • C in 7% CO 2 , sub-cultured (1:3 or 1:5) once a week, and the medium was renewed once a week. The ARPE-19 cells were differentiated, as previously described, on laminin-coated Costar Transwell inserts (pore size 0.4 µm, Corning, NY, USA) at a seeding density of 160,000 cells/cm 2 [34]. The growth medium was renewed twice a week. The composition of the growth medium was the same as above, except the FBS concentration was 1%. The cells were grown on inserts for 3-4 weeks. The tight junction formation was evaluated by a transepithelial electrical resistance measurement (TEER).

Primary Porcine RPE (pRPE)
Fresh porcine eyes were obtained from a local slaughterhouse (Jensens Slagtehus, Hadsten, Denmark). The porcine eyes were kept on ice during transport and handled immediately upon arrival to the laboratory. The extraocular tissues were removed with curved iris scissors. The anterior part of the eye, the lens and the vitreous, were gently removed. The eye cup was placed in a well plate and filled with PBS. After 10 min, the PBS was removed, together with the neural retina. The eye cup was then filled with 0.25% trypsin-EDTA and placed in the incubator at +37 • C for 30 min. After this time, the detached RPE cells were collected. The RPE cells of 20 eyes were collected in 50 mL tubes containing 1 part trypsin-EDTA with the cells and 2 parts cell culture medium (low glucose DMEM supplemented with pyruvate, 10% FBS, 2 mM GlutaMAX, and 1% penicillin-streptomycin (10,000 U/mL)). The tubes were centrifuged three times at 1000 rpm for 2 min, and each time the supernatant was replaced by fresh culture medium. Then, the cells were placed in a T25 tissue culture flask and the medium was replaced twice a week until the cell monolayer reached confluency. The pRPE cells were then seeded on collagen IV-coated (11.2 µg collagen/insert) Transwell inserts (12 well) at a density of 160,000 cells/cm 2 and cultured for 4 weeks. The cell culture medium had the same composition as described above, except it contained only 1% FBS. The tight junction formation was evaluated by a transepithelial electrical resistance measurement (TEER).

Transepithelial Electrical Resistance Measurement (TEER)
The TEER of the cell monolayers was measured with a Millicell ERS-2 Volt-Ohm Meter equipped with a silver/silver chloride electrode (EMD Millipore Corporation, Billerica, MA, USA). Briefly, the STX01 electrode was sanitized with 70% ethanol for 15 min immediately before use, then the electrode was used to carry out the measurements. To obtain the actual monolayer resistance, the resistance reading across a blank insert (insert without cells) was subtracted from the resistance reading across the monolayer (insert with cells). The unit area resistance (Ω cm 2 ) was calculated by multiplying the resistance reading (Ω) with the effective surface area of the filter membrane (cm 2 ).

Cytotoxicity
The cytotoxicity of the polyplexes, lipoplexes, and L2000/siRNA complexes was determined by an MTT assay. The ARPE-19 cells were seeded on a 24-well plate at a density of 100,000 cells/well in 500 µL of supplemented growth medium. The next day, the cells were washed with PBS (1x, pH 7.2). The formulations with 50 nM of siRNA were incubated with the cells in Opti-MEM for 5 h. Then, the cells were washed with PBS and incubated for 4 h with 400 µL of 0.5 mg/mL thiazolyl blue tetrazolium bromide (MTT, Sigma-Aldrich, St. Louis, MO, USA). After incubation, formazan crystals of the living cells were solubilized by adding 100 µL of 10% sodium dodecyl sulphate −0.1 M hydrochloric acid.
On the following day, the formazan amount was quantified by measuring the absorbance at 570 nm (Varioskan Flash, Thermo Fisher Scientific, Waltham, MA, USA). The cell viability was calculated as a percentage of the non-treated cells.
The proliferation of the cells transfected with the lipidoid and DF4 complexes was measured by an alamarBlue assay following the instructions given by the manufacturer. This assay was used with differentiated cells, because it did not require cell lysis. Briefly, an amount equal to 10% of cell culture volume (50 µL) of alamarBlue reagent was added apically and the cells were returned to the incubator for one hour. Thereafter, 100 µL of cell culture medium was transferred to a black well plate, and the fluorescence was measured at λ ex 560 nm and λ em 590 nm with a plate reader (Varioskan Flash, Thermo Fisher Scientific, Waltham, MA, USA). The cell proliferation was calculated relative to the non-treated cells. Dividing cells: The ARPE-19 cells were seeded on a 24-well plate at a density of 100,000 cells/well in 500 µL of supplemented growth medium. Next day, the cells were transfected with 50 nM of IL-6 siRNA in Opti-MEM for 5 h. After transfection, the cells were washed with PBS, and 1 mL of growth medium containing 10 µg/mL of lipopolysaccharide (LPS, Escherichia coli O55:B5 Sigma-Aldrich, St. Louis, MO, USA) was added to the cells. The LPS was used to induce IL-6 secretion [40]. Samples of 500 µL were collected from the supernatant 72 h after the polyplex and lipoplex removal and replaced with the equal volume of LPS medium. The amount of IL-6 was analyzed from the collected medium samples using a human IL-6 ELISA kit according to the instructions provided by the manufacturer (Gen-Probe, Diaclone SAS, Besançon, France). The relative IL-6 protein secretion was calculated by comparing the exposed cells to the LPS-treated cells without any exposure (100%).
Differentiated cells: The ARPE-19 cells cultured for 3-4 weeks were transfected apically with 100 nM of IL-6 siRNA in Opti-MEM for 5 h. After incubation, the medium was removed and the cells were washed with PBS. Growth medium supplemented with 10 µg/mL of LPS was added to the apical and basolateral sides of the cells. Medium samples were collected at 2, 5, and 9 days. The removed samples were replaced with an equal volume of the LPS medium. The samples were analyzed with the human ELISA kit described above.

Glyceraldehyde-3-Phosphate Dehydrogenase (GAPDH) and Hypoxanthine Phosphoribosyltransferase 1 (HPRT1) MRNA Knockdown
The differentiated cells were transfected apically with 10, 50, or 100 nM of GAPDH siRNA (ARPE-19 and hESC-RPE) or HPRT1 siRNA (pRPE) overnight in full medium, including serum. Two days after transfection, the total RNA was isolated using the NucleoZOL reagent (Macherey-Nagel, Düren, Germany). The total RNA concentration was determined with a DeNovix DS-11 spectrophotometer (DeNovix Inc., Wilmington, DE, USA). The cDNA was synthesized using the RevertAid First strand cDNA synthesis kit (Thermo Fisher Scientific, Waltham, MA, USA). The quantification of the GAPDH mRNA, HPRT1 mRNA, and β-actin mRNA was performed with the LightCycler 480 SYBR Green I Master (Roche, Basel, Switzerland). The primer concentration was 250 nM, and 4 ng of cDNA was used in the final 10 µL reaction volume. All the quantitative PCRs were performed on a LightCycler ® 480 Instrument (Roche, Basel, Switzerland). The sequence-specific amplification of cDNAs was verified by performing a melting-point analysis. For the differentiated ARPE-19 and hESC-RPE cells, the GAPDH mRNA expression was normalized to the endogenous β-actin, whereas for the pRPE cells, the HPRT1 mRNA expression was normalized to the endogenous GAPDH.

Cell Uptake
The cells were transfected as in paragraph 2.7 using either FAM-IL-6 siRNA or Alexa Fluor 488 dsDNA (21-mer). After incubation, the cells were washed three times with Hanks' Balanced Salt Solution (HBSS) (Thermo Fisher Scientific, Waltham, MA, USA) supplemented with 10 mM of HEPES (pH 7.4). Then, the cells were detached with trypsin, re-suspended in the HBSS-HEPES buffer, and centrifuged at 1200 rpm for 10 min. The cells were analyzed by flow cytometry (LSR II, Becton, Dickinson and Company, Franklin Lakes, NJ, USA and Gallios, Beckman Coulter Life Sciences, Indianapolis, IN, USA). The data collection and analysis were controlled by the FACSDiva software (BD Biosciences) and Kaluza software (Beckman Coulter Life Sciences), where 10,000 cell events were collected for each sample. Non-treated and naked FAM-IL-6 siRNA-treated cells were used for setting the baseline and for calculating the relative cell association of the siRNA formulations.

Intracellular Distribution
The intracellular distribution of Atto565 dsDNA (21-mer) was observed with confocal microscopy. The hESC-RPE cells were transfected as in paragraph 2.7. The lysosomes were labeled with the CellLight Lysosomes-GFP reagent (Thermo Fisher Scientific, Waltham, MA, USA) following the instructions given by the manufacturer. Briefly, 24 h after transfection, the CellLight reagent (16 µL) was added and incubated with the cells for ≥ 16 h. Thereafter, the cells were fixed with 4% paraformaldehyde for 10 min at room temperature and mounted in Vectashield mounting medium (Vector Laboratories Inc., Burlingame, CA, USA). The melanosomes were labeled by immunocytochemistry as the following. The cells were fixed with 4% paraformaldehyde for 10 min at room temperature, followed by permeabilization with 0.5% saponin in PBS for 10 min and blocking with 1% bovine serum albumin (BSA), 22.52 mg/mL glycine in PBST (PBS +0.1% Tween 20) for 30 min. The cells were then incubated with the primary polyclonal antibody Rab27a (1:200, Sicgen, Cantanhede, Portugal) in 1% BSA in PBST in a humidified chamber for 1 h at room temperature, followed by incubation with the secondary antibody (donkey anti-goat Alexa Fluor 488, 1:500, Thermo Fisher Scientific, Waltham, MA, USA) in 1% BSA in PBST for 1 h at room temperature. The cell nuclei were stained with the DAPI included in the Vectashield mounting medium. Images were captured on an Olympus IX-83 fluorescent microscope with the Andor confocal spinning unit and Andor iXon Ultra 897 camera, Olympus Upsalo W, 60x/1.20 NA water objective lens, using the Olympus cellSens software (Olympus). Multichannel images were processed using the Fiji software (ImageJ 1.52s).

Statistical Analysis
A one-way ANOVA followed by Tukey's multiple comparisons test was performed using GraphPad Prism version 8.4.2 for macOS (GraphPad Software, San Diego, CA, USA). The significance level of the test was set to 0.05.

Particle Size and Polydispersity Index (PdI)
The particle size and polydispersity index (PdI) of the siRNA/carrier complexes, evaluated by DLS and NTA, are summarized in Table 1. The PBE 30 -b-PK 30 polyplexes were around 50 nm and had a narrow size distribution, whereas the DOTAP/DOPE/PS lipoplexes were about 200 nm in diameter and displayed a higher polydispersity. The lipidoid-siRNA complexes (lipidoid) were approximately 150 nm in diameter and had a monodisperse distribution. Comparable results were obtained when their sizes were measured with NTA. The commercial complexes were measured by NTA, since the phenol red present in the nanoparticle preparation medium disturbed the DLS measurement. The commercial complexes sizes were 160-280 nm and polydisperse. n/a n/a 281 ± 160 Metafectene PRO n/a n/a 160 ± 88 Lipofectamine 2000 n/a n/a n/a

Cytotoxicity
The in vitro toxicity of all the siRNA/carrier complexes was low, and the RPE cells maintained an over 80% viability after treatment (Figure 1).

Cytotoxicity
The in vitro toxicity of all the siRNA/carrier complexes was low, and the RPE cells maintained an over 80% viability after treatment (Figure 1).

Human ARPE-19
Dividing ARPE-19 cells: The IL-6 protein knockdown was about 70% (P < 0.0001) with the lipoplexes (Figure 2), whereas no knockdown was measured with the polyplexes, even after treatment with chloroquine, a commonly used lysosomotropic agent that promotes endosomal escape and enhances transfection [41] (Supplementary Figure S1). L2000 resulted in a 60% (P = 0.0003) decrease in protein secretion. There was no statistically significant difference between the knockdown efficacy of the lipoplexes and L2000 (P = 0.4663). The negative control siRNA did not promote silencing.

Human ARPE-19
Dividing ARPE-19 cells: The IL-6 protein knockdown was about 70% (p < 0.0001) with the lipoplexes (Figure 2), whereas no knockdown was measured with the polyplexes, even after treatment with chloroquine, a commonly used lysosomotropic agent that promotes endosomal escape and enhances transfection [41] (Supplementary Figure S1). L2000 resulted in a 60% (p = 0.0003) decrease in protein secretion. There was no statistically significant difference between the knockdown efficacy of the lipoplexes and L2000 (p = 0.4663). The negative control siRNA did not promote silencing.

Cytotoxicity
The in vitro toxicity of all the siRNA/carrier complexes was low, and the RPE cells maintained an over 80% viability after treatment ( Figure 1).

Human ARPE-19
Dividing ARPE-19 cells: The IL-6 protein knockdown was about 70% (P < 0.0001) with the lipoplexes (Figure 2), whereas no knockdown was measured with the polyplexes, even after treatment with chloroquine, a commonly used lysosomotropic agent that promotes endosomal escape and enhances transfection [41] (Supplementary Figure S1). L2000 resulted in a 60% (P = 0.0003) decrease in protein secretion. There was no statistically significant difference between the knockdown efficacy of the lipoplexes and L2000 (P = 0.4663). The negative control siRNA did not promote silencing.  Differentiated ARPE-19 cells: The IL-6 protein knockdown was highest with the lipoplexes, at about 60% (p < 0.0001) 5 days after transfection (Figure 3), whereas the L2000 resulted in a 40% (p = 0.0207) decrease in protein secretion at the same time point. The knockdown efficacy between the lipoplexes and L2000 at day 5 was not statistically significant (p = 0.5235). Interestingly, the silencing effect of lipoplexes was delayed compared to L2000, and prolonged to at least 9 days after transfection (p < 0.0001). The polyplexes were not tested with the differentiated ARPE-19 cells, since no knockdown was detected in the dividing cells. The negative control siRNA did not promote silencing. The siRNA dose was doubled (100 nM) in the differentiated cells, because no knockdown could be measured with the same dose used in the dividing cells (50 nM). The TEER of the ARPE-19 cells was 39 ± 5 Ω cm 2 . Figure 3 illustrates the decrease in IL-6 secretion on the apical side of the monolayer; the basolateral IL-6 secretion was also measured with similar results (data not shown).
Pharmaceutics 2020, 12, x FOR PEER REVIEW 9 of 17 0.0207) decrease in protein secretion at the same time point. The knockdown efficacy between the lipoplexes and L2000 at day 5 was not statistically significant (P = 0.5235). Interestingly, the silencing effect of lipoplexes was delayed compared to L2000, and prolonged to at least 9 days after transfection (P < 0.0001). The polyplexes were not tested with the differentiated ARPE-19 cells, since no knockdown was detected in the dividing cells. The negative control siRNA did not promote silencing. The siRNA dose was doubled (100 nM) in the differentiated cells, because no knockdown could be measured with the same dose used in the dividing cells (50 nM). The TEER of the ARPE-19 cells was 39 ± 5 Ω cm 2 . Figure 3 illustrates the decrease in IL-6 secretion on the apical side of the monolayer; the basolateral IL-6 secretion was also measured with similar results (data not shown). The lipidoid-siRNA complexes-mediated GAPDH mRNA silencing in differentiated ARPE-19 cells (Figure 4) was achieved with the same 50 nM dose used in the dividing cells. After transfection, the GAPDH gene expression was decreased by 60% (P = 0.0019). The DF4-mediated silencing was approximately 90% (P < 0.0001), however there was no statistical significance between the lipidoids and DF4 (P = 0.1530). The lipidoids retained a 64% gene silencing efficacy even at a low 10 nM siRNA dose (data not shown). The negative control siRNA did not promote silencing. The TEER of the ARPE-19 cells in this experiment was 78 ± 13 Ω cm 2 .  The lipidoid-siRNA complexes-mediated GAPDH mRNA silencing in differentiated ARPE-19 cells (Figure 4) was achieved with the same 50 nM dose used in the dividing cells. After transfection, the GAPDH gene expression was decreased by 60% (p = 0.0019). The DF4-mediated silencing was approximately 90% (p < 0.0001), however there was no statistical significance between the lipidoids and DF4 (p = 0.1530). The lipidoids retained a 64% gene silencing efficacy even at a low 10 nM siRNA dose (data not shown). The negative control siRNA did not promote silencing. The TEER of the ARPE-19 cells in this experiment was 78 ± 13 Ω cm 2 .
Pharmaceutics 2020, 12, x FOR PEER REVIEW 9 of 17 0.0207) decrease in protein secretion at the same time point. The knockdown efficacy between the lipoplexes and L2000 at day 5 was not statistically significant (P = 0.5235). Interestingly, the silencing effect of lipoplexes was delayed compared to L2000, and prolonged to at least 9 days after transfection (P < 0.0001). The polyplexes were not tested with the differentiated ARPE-19 cells, since no knockdown was detected in the dividing cells. The negative control siRNA did not promote silencing. The siRNA dose was doubled (100 nM) in the differentiated cells, because no knockdown could be measured with the same dose used in the dividing cells (50 nM). The TEER of the ARPE-19 cells was 39 ± 5 Ω cm 2 . Figure 3 illustrates the decrease in IL-6 secretion on the apical side of the monolayer; the basolateral IL-6 secretion was also measured with similar results (data not shown). The lipidoid-siRNA complexes-mediated GAPDH mRNA silencing in differentiated ARPE-19 cells (Figure 4) was achieved with the same 50 nM dose used in the dividing cells. After transfection, the GAPDH gene expression was decreased by 60% (P = 0.0019). The DF4-mediated silencing was approximately 90% (P < 0.0001), however there was no statistical significance between the lipidoids and DF4 (P = 0.1530). The lipidoids retained a 64% gene silencing efficacy even at a low 10 nM siRNA dose (data not shown). The negative control siRNA did not promote silencing. The TEER of the ARPE-19 cells in this experiment was 78 ± 13 Ω cm 2 .

Primary Porcine RPE (pRPE)
A siRNA dose of 50 nM was sufficient to silence the housekeeping gene HPRT1 in differentiated pRPE cells with lipidoids ( Figure 5). After transfection, the remaining HPRT1 gene expression was 45% (p = 0.0294). The DF4 achieved a 50% knockdown (p = 0.0482). The negative control siRNA did not promote silencing. The TEER of the pRPE cells was 853 ± 39 Ω cm 2 .

Primary Porcine RPE (pRPE)
A siRNA dose of 50 nM was sufficient to silence the housekeeping gene HPRT1 in differentiated pRPE cells with lipidoids ( Figure 5). After transfection, the remaining HPRT1 gene expression was 45% (P = 0.0294). The DF4 achieved a 50% knockdown (P = 0.0482). The negative control siRNA did not promote silencing. The TEER of the pRPE cells was 853 ± 39 Ω cm 2 .

Human Embryonic Stem Cell-derived RPE (hESC-RPE)
A siRNA dose of 50 nM in lipidoids silenced the GAPDH by 50% (P = 0.0004) in 2-week matured hESC-RPE cells (Figure 6a). The best carrier was PRO, with a knockdown efficiency of 75% (P < 0.0001), while the efficiency of the other commercial carrier DF4 was 55% (P < 0.0001). The negative control siRNA did not promote statistically significant silencing for PRO and lipidoids. The TEER of the 2-week matured hESC-RPE cells was 18 ± 5 Ω cm 2 .
Despite the promising results obtained with the 2-week matured hESC-RPE cells, no statistically significant knockdown was achieved in the 4-week matured cells (Figure 6b). The negative control siRNA did not promote silencing. The TEER of the 4-week matured hESC-RPE cells was 435 ± 47 Ω cm 2 . The 2-week matured cells were not pigmented, whereas pigmentation is clearly visible in the 4week matured cells.

Human Embryonic Stem Cell-derived RPE (hESC-RPE)
A siRNA dose of 50 nM in lipidoids silenced the GAPDH by 50% (p = 0.0004) in 2-week matured hESC-RPE cells (Figure 6a). The best carrier was PRO, with a knockdown efficiency of 75% (p < 0.0001), while the efficiency of the other commercial carrier DF4 was 55% (p < 0.0001). The negative control siRNA did not promote statistically significant silencing for PRO and lipidoids. The TEER of the 2-week matured hESC-RPE cells was 18 ± 5 Ω cm 2 .

Human Embryonic Stem Cell-derived RPE (hESC-RPE)
A siRNA dose of 50 nM in lipidoids silenced the GAPDH by 50% (P = 0.0004) in 2-week matured hESC-RPE cells (Figure 6a). The best carrier was PRO, with a knockdown efficiency of 75% (P < 0.0001), while the efficiency of the other commercial carrier DF4 was 55% (P < 0.0001). The negative control siRNA did not promote statistically significant silencing for PRO and lipidoids. The TEER of the 2-week matured hESC-RPE cells was 18 ± 5 Ω cm 2 .
Despite the promising results obtained with the 2-week matured hESC-RPE cells, no statistically significant knockdown was achieved in the 4-week matured cells (Figure 6b). The negative control siRNA did not promote silencing. The TEER of the 4-week matured hESC-RPE cells was 435 ± 47 Ω cm 2 . The 2-week matured cells were not pigmented, whereas pigmentation is clearly visible in the 4week matured cells.  Despite the promising results obtained with the 2-week matured hESC-RPE cells, no statistically significant knockdown was achieved in the 4-week matured cells (Figure 6b). The negative control siRNA did not promote silencing. The TEER of the 4-week matured hESC-RPE cells was 435 ± 47 Ω cm 2 . The 2-week matured cells were not pigmented, whereas pigmentation is clearly visible in the 4-week matured cells.

Cell Uptake
Cell uptake studies were performed in order to evaluate the effect of cell differentiation on the endocytosis of siRNA complexes. The carriers tested in the dividing ARPE-19 cells (polyplexes, lipoplexes, and commercial carriers) all had a 100% cell association (measured as the % of positive cells, Figure 7a). In the differentiated ARPE-19 and 4-week matured hESC-RPE cells instead, the lipidoids had the highest % of positive cells (>90% in hESC-RPE, Figure 7b), while the lipoplexes had 20% positive ARPE-19 cells. The L2000 and DF4 had 20% (ARPE-19) and 40% (hESC-RPE) positive cells, respectively.

Cell Uptake
Cell uptake studies were performed in order to evaluate the effect of cell differentiation on the endocytosis of siRNA complexes. The carriers tested in the dividing ARPE-19 cells (polyplexes, lipoplexes, and commercial carriers) all had a 100% cell association (measured as the % of positive cells, Figure 7a). In the differentiated ARPE-19 and 4-week matured hESC-RPE cells instead, the lipidoids had the highest % of positive cells (>90% in hESC-RPE, Figure 7b), while the lipoplexes had 20% positive ARPE-19 cells. The L2000 and DF4 had 20% (ARPE-19) and 40% (hESC-RPE) positive cells, respectively.

Intracellular Distribution in hESC-RPE
The intracellular distribution of the fluorescently labelled dsDNA (used as a non-biologically active mimic for siRNA) was studied in order to explain the striking knockdown differences observed in the hESC-RPE that were matured for 2 and 4 weeks (Figure 6a,b).

2-Week Matured hESC-RPE
The intracellular distribution of the fluorescently labelled dsDNA (used as a non-biologically active mimic for siRNA, in red) was dependent on the type of carrier. For DF4 (Figure 8a) and especially for PRO (Figure 8b), cytosolic staining (arrowheads), which may indicate endosomal release, was observed. With these carriers, the labelled dsDNA also localized inside large aggregates (arrows). In the case of lipidoids (Figure 8c), the dsDNA was found inside discrete, small vesicles, which did not colocalize with lysosomes (in green). The Z-stack images indicate that for all the tested carriers, dsDNA was found inside the cells.

Intracellular Distribution in hESC-RPE
The intracellular distribution of the fluorescently labelled dsDNA (used as a non-biologically active mimic for siRNA) was studied in order to explain the striking knockdown differences observed in the hESC-RPE that were matured for 2 and 4 weeks (Figure 6a,b).

2-Week Matured hESC-RPE
The intracellular distribution of the fluorescently labelled dsDNA (used as a non-biologically active mimic for siRNA, in red) was dependent on the type of carrier. For DF4 (Figure 8a) and especially for PRO (Figure 8b), cytosolic staining (arrowheads), which may indicate endosomal release, was observed. With these carriers, the labelled dsDNA also localized inside large aggregates (arrows). In the case of lipidoids (Figure 8c), the dsDNA was found inside discrete, small vesicles, which did not colocalize with lysosomes (in green). The Z-stack images indicate that for all the tested carriers, dsDNA was found inside the cells. Pharmaceutics 2020, 12, x FOR PEER REVIEW 12 of 17

Discussion
The pathogenesis of several age-related ocular disorders is characterized by the presence of continuous, low-grade inflammation. Silencing pro-inflammatory genes by siRNA is an attractive therapeutic option to reduce inflammation and restore homeostasis in the retina. Despite the undeniable potential of siRNA therapeutics, a key obstacle still hampering their widespread use is

Discussion
The pathogenesis of several age-related ocular disorders is characterized by the presence of continuous, low-grade inflammation. Silencing pro-inflammatory genes by siRNA is an attractive therapeutic option to reduce inflammation and restore homeostasis in the retina. Despite the undeniable potential of siRNA therapeutics, a key obstacle still hampering their widespread use is

Discussion
The pathogenesis of several age-related ocular disorders is characterized by the presence of continuous, low-grade inflammation. Silencing pro-inflammatory genes by siRNA is an attractive therapeutic option to reduce inflammation and restore homeostasis in the retina. Despite the undeniable potential of siRNA therapeutics, a key obstacle still hampering their widespread use is the safe and efficient delivery to target cells. Gene silencing in the RPE requires a carrier that can successfully deliver the siRNA to a quiescent, terminally differentiated cell monolayer. siRNA therapeutics must first of all gain access into cells, a process which occurs via endocytosis and whose mechanisms have been widely studied. It is worth keeping in mind, however, that most of our current knowledge on endocytic mechanisms is derived from studies of proliferating cells, and that the mechanisms prevailing in non-dividing cells might be quite different [42]. Several studies have shown, for example, that cell differentiation causes a reduction in the uptake of different cargoes in intestinal [43][44][45], kidney [46], and airway [47] epithelial cells. While we could not find any published literature on the effect of cell differentiation on the endocytic activity of RPE cells, we also observed a decreased uptake of siRNA complexes in differentiated ARPE-19 cells compared to dividing cells. While lipidoid-siRNA complexes did retain a > 90% cell association in 4-week matured hESC-RPE cells, a confocal microscope analysis revealed that they were not internalized inside cells, but instead they localized apically at the cell surface. Thus, our results indicate that, similarly to other epithelial tissues [43][44][45][46][47], differentiation alters the endocytic activity of RPE cells and causes a decrease in the uptake of siRNA complexes.
RPE cell differentiation also lessened the efficacy of gene silencing. While clearly a less efficient cell uptake may partly explain the decreased knockdown, an additional important barrier for efficient siRNA therapy was revealed to be the sequestration of dsDNA (used as siRNA mimic) by melanosomes. Melanosomes do not normally occur in ARPE-19 cells [48] and 2-week matured hESC-RPE (Figure 6a), whereas the 4-week matured hESC-RPE were heavily pigmented (Figure 6b). Some drugs have long been known to be stored inside melanosomes via melanin binding. Indeed, the first reports describing the accumulation of drugs inside ocular melanin were published more than 60 years ago [49][50][51], and several thousands of compounds, mostly small molecular weight drugs, have been investigated for melanin binding in vitro and in vivo since [52,53]. Very few studies, however, describe the melanin binding of oligonucleotides and nanoparticles. Pitkänen and colleagues [54] found that oligonucleotides did not bind to isolated and synthetic melanin, whereas Geng and coworkers [55] showed that melanin from bacteria interacted with DNA. Despite these conflicting results, the entrapment of siRNA complexes inside melanosomes may not require melanin binding. Schraermeyer and colleagues, for example, observed the accumulation of gold nanoparticles (10-20 nm) used to label rod outer segments [56] and latex particles (100 nm) [57] in the narrow space between melanin and the melanosome membrane in the RPE of bovine RPE-choroid tissue explants. Melanosomes, whose membranes and contents are largely derived from endosomes, are members of the lysosome-related organelle (LRO) family [58]. Previous work has found that, while the vast majority of siRNA lipoplexes that enter cells end up in endolysosomes, only a few percent of siRNA reach the cytosol, either by direct fusion between the lipoplexes and the plasma membrane [59] or via endosomal escape [60,61]. Accordingly, only a fraction of the siRNA dose that localizes in the cytosol is responsible for the gene silencing effect. In our study, we observed the significant cytoplasmic distribution of siRNA in 2-week matured hESC-RPE (Figure 8b). In addition, none of the different complexes co-localized with lysosomes. Upon cell pigmentation at 4 weeks, however, the cytoplasmic staining disappeared, and, in the case of Metafectene PRO, the siRNA localized inside melanosomes instead. This also corresponded to a loss in gene silencing. It is possible that if, upon delivery to the melanosomes, the siRNA is still incorporated into lipoplexes, then the cationic lipoplexes may bind to the polyanionic melanin [53]. This binding may prevent the interaction of lipoplexes with the melanosomal membrane, and thus preclude the release of siRNA into the cytosol.
In summary, our study highlights the importance of picking a physiologically relevant RPE cell model for the selection of siRNA delivery systems. While it is tempting to evaluate their efficacy only in dividing, non-pigmented RPE cells, there is a real risk that these systems will not work in vivo. Given their easy culture, low costs, and quick growth, dividing ARPE-19 cells are still valuable in the early selection rounds and when large preliminary screens are performed. Nevertheless, advanced selections should be carried out with RPE models that better mimic the outer blood-retina barrier-that is, differentiated, pigmented RPE cells. This simple but important choice may lead to a more stringent selection of delivery systems and increase their chance of success in vivo.