Introduction

Lying between the outer retina and the choroidal vessels, the retinal pigment epithelium (RPE) fulfils many crucial tasks for visual function such as phagocytosis, 11-cis retinal recycling, ion and water transports, and growth factor secretion1. Deficiencies in RPE functions give rise to various diseases such as Leber congenital amaurosis2 (LCA) or retinitis pigmentosa3, most of which result in drastic visual impairments or blindness. The generation of RPE in vitro is particularly valuable to understand pathophysiological mechanisms through disease modelling and to help answer the major challenge of developing therapies.

Numerous protocols to differentiate embryonic or induced-pluripotent stem cells (ESCs or iPSCs, respectively) have been published4,5,6, providing researchers with a renewable source of RPE. Thanks to the protocols developed, the therapeutic potential of stem cell-derived RPE cells has been explored by transplantation in RPE degenerative diseases, first in animal models5,7,8, and more recently in human patients suffering from age-related macular degeneration9,10,11,12 (AMD) or Stargardt’s disease13.

In addition, iPSC-derived RPE (iRPE) has been used as an in vitro model for Best disease14 and choroideremia15. It has proven to be an interesting alternative model for some diseases such as AMD16, where no animal recapitulates all the cellular features of human AMD17.

Worthy of note, more and more genetic diseases are tackled with a gene therapy approach and iRPE has also allowed to test the efficacy of adeno-associated virus (AAV)- or CRISPR-mediated gene correction15,16,18.

Moreover, in many diseases that are candidates for gene augmentation therapy, it is not clear how many vector copies are necessary to re-establish a physiological level of gene expression. For instance, AAV-mediated gene augmentation therapy on RPE65-deficient dogs showed a significant improvement in visual function that appeared on electroretinogram readings, but this success could not be reproduced on LCA2 patients with similar doses19,20,21,22. Therefore, studying the vector dose-response relationship in human iRPE may help to better estimate the vector dose required to attain optimal rescue in human in vivo.

In our study, we evaluated the suitability of iRPE cells as an in vitro model for lentiviral gene therapy, first through an extensive characterization of RPE features, followed by testing different lentiviral constructs on iRPE cells. We subjected healthy iRPE to an RPE65 gene augmentation therapy previously tested on RPE65 deficiency mouse models23,24 and healthy non-human primates25 to study RPE65 gene and protein expression.

Results

From hiPSCs to RPE

Differentiation protocol

We first tested 3 hiPSC lines for their potency to generate iRPE: two of them produced 10 to 50 pigmented foci during the differentiation protocol, whereas the one described by Singh and colleagues14 produced 50 to 200 foci. We thus decided to focus the iRPE characterization on this hiPSC line. The differentiation protocol was adapted from Singh et al.14, recapitulating its main steps with modifications in time points (plating at day 1) and minor changes in the media (addition of embryoid body media – EBM and retinal differentiation media cum vitamin A - RDMcA) (Fig. 1a). Human iPSCs were lifted and placed in free-floating condition to form embryoid body-like aggregates (Fig. 1b) that were plated one day later to expand (Fig. 1c) in Neural Induction Medium (NIM) which contains a high concentration of insulin potentially mimicking IGF1 action26,27. The medium was switched 9 days later to a Retinal Differentiation Medium (RDM) in order to commit cells to the retinal fate possibly by the stimulation of the retinoic acid28 and thyroid hormone29.

Figure 1
figure 1

RPE differentiation protocol timeline. (a) Schematic depiction of the differentiation protocol highlighting the time course and media used. The protocol starts with hiPSCs in feeder-free condition that are lifted in clumps using collagenase to form (b) embryoid body-like aggregates (EB) in floating condition in an embryoid body media (EBM). EBs are plated on day 1 on Matrigel Reduced Factor (MRF) in a neural induction media (NIM) and (c) expand. The cells are switched to a retinal differentiation medium with vitamin A (RDMcA) from day 10 to 13, then without vitamin A (RDMsA). (d,e) Pigmented foci (PFs) appear around day 18 to 24 and are collected on day 30. Optic vesicle-like structures can also be observed (e, arrowheads). (f,g) Plated PFs develop pigmented patches (PPs) that are collected, dissociated and seeded to obtain (h) the pure hiPSC-derived RPE (iRPE) cell monolayer.

Full size image

In our hands, a short induction in RDMcA resulted in quicker pigmentation that may originate from the retinoic acid (RA). The RA produced from the vitamin A in the RDMcA may help to create the optic vesicle presumptive fields that later form the retina and RPE. However, after 3 days, the medium is changed to a vitamin A-free RDM (RDMsA) to prevent the neural retina fate30 that may arise detrimentally to RPE differentiation due to the trans-differentiation ability of RPE during development31.

Pigmented foci (Fig. 1d,e) appeared around day 18-24 and were manually collected around day 30 to be re-plated as a purification and enrichment step. The pigmented foci developed large pigmented patches reaching up to several millimetres of diameter (Fig. 1e), composed of iRPE cells harbouring the typical RPE cobblestone morphology (Fig. 1f). After 20 to 30 days of expansion, the pigmented patches were dissected, dissociated and plated. This step is considered as passage 1 (P1). All assays performed were on P3 iRPE matured 6 weeks on transwell (Fig. 1h) to allow time for the cells to re-acquire RPE characteristics lost during the passages (as previously observed in ESC-derived RPE32,33).

Expression of RPE gene markers

In order to confirm the RPE identity of the iRPE, we investigated the RPE gene markers’ expression compared to iPSCs and cultured foetal RPE (fRPE). The RT-qPCR results (Fig. 2a) showed a marked and significant upregulation of RPE65, MITF-H (RPE expressed isoform H), PMEL, DCT and BEST1 gene expression in iRPE compared to iPSCs while there was no significant difference for KLF4, EZR, MERTK and MITF-M (melanocyte specific isoform M). A significant downregulation for the pluripotency markers POU5F1-A and NANOG was observed. Nonetheless, the iRPE was still positive for these markers (<30 Cq).

Figure 2
figure 2

Expression of RPE markers at mRNA and protein levels. (a) Quantitative PCR investigating RPE marker expression in fRPE (n = 3) and iRPE (n = 6) normalized to iPS (n = 3). Bars represent mean ± SD. Kruskal-Wallis tests were performed (iPS vs. iRPE and iPS vs. fRPE for each gene) followed by Dunn’s multiple comparisons test. Significance threshold: *<0.05; **<0.01. (b,e) Z-stack Maximum Intensity Projection of iRPE cells labelled with anti-BEST1 and -RPE65 antibodies, respectively. (c,d) iRPE cells immunolabelled for OTX2. (f–i) Z-stacks acquired over the whole iRPE cell monolayers to investigate RPE marker localizations. Orthogonal view confirmed the preferential basolateral or apical distributions of BEST1 (f) and ZO-1 proteins (g), respectively. RPE65 and CRALBP localizations were widespread in iRPE cells (h,i). (j) Violin plots corresponding to images (f–i) showing the distribution of the labelled proteins and DAPI staining throughout the Z-stacks. Images are representative of 1-2 observations per marker in at least 3 iRPE lines.

Full size image

Surprisingly, a downregulation of LRAT expression in iRPE compared to iPSCs was observed whereas an increase similar to RPE65 was expected, as both are involved in the visual cycle34. Nonetheless LRAT mRNA expression can be considered positive in iRPE (<30 Cq).

Despite this unexpected result, the iRPE showed an expression pattern close to fRPE with a strong significant correlation (R = 0.902) for the 10 genes investigated (Supp. Fig. 1b). Note that LRAT, EZR and MERTK in fRPE have similar expression levels in comparison to iRPE.

At the protein level, the RPE-specific proteins RPE65, RLBP1 (also known as CRALBP), BEST1, TJP1 (also known as ZO-1) and OTX2 were all expressed although with some heterogeneity (Fig. 1b–i). NANOG was not expressed. However, a few POU5F1-A positive cells were observed (Supp. Fig. 2). Quantification of OTX2 positive cells (Supp. Fig. 2) showed that more than 94% of cells express OTX2 confirming a homogeneous RPE identity. Moreover, the polarization of some critical proteins was confirmed by confocal microscopy, particularly the apical distribution of ZO-1 and the basolateral presence of BEST1 as assessed by orthogonal view (Fig. 1f–i). In order to better visualize the protein distribution in the cell in the Z-axis, the mean grey value of each Z-stack of Fig. 1f–i was measured to present the Z-profile of both the nucleus (labelled by DAPI) and the protein of interest (Fig. 1j).

As a comparison, fRPE showed similar staining and protein localization (Supp. Fig. 3). However, after cell passages, the cells lose progressively their identity (Supp. Fig. 7). The iRPE coming from replicated differentiation experiment displayed consistent results (Supp. Fig. 4).

Cell ultrastructure analysis reveals polarized distributions of certain organelles

Vertical sections of iRPE cells were imaged by electron microscopy to reveal their ultrastructure (Fig. 3a). The presence of tight junctions between iRPE cells (Fig. 3b,c), apical microvilli (Fig. 3d), as well as a basal nucleus and melanosomes at different stages of maturation35 (Fig. 3e) confirmed their RPE identity.

Figure 3
figure 3

iRPE cells present characteristic ultrastructure features of human adult RPE (a) Typical iRPE cell ultrastructure with a large nucleus in the basal part of the cell and with dense structures in the apical region containing microvilli. (c–e) Inserts show these iRPE cell features with high magnification. (b,c) Tight junctions (arrows) between two iRPE cells. (d) Multiple apical microvilli (AMv) are present at the cell surface. (e) Melanosomes (arrowheads) at different maturation stages were observed. Nu, Nucleus; Ph, Phagosome; M, Mitochondria.

Full size image

Differentiated iRPE form an epithelium with high cohesion

As part of the blood retinal barrier1, RPE cells need to form a functional epithelium that can regulate ion, molecule and protein transport. By measuring the transepithelial resistance, we sought to investigate the epithelium integrity. The iRPE monolayer showed an increasing transepithelial resistance in parallel with the number of days in culture (Fig. 4a) with a mean transepithelial resistance of around 300 Ohm*cm2 at day 42 (when characterization is performed), which is in line with previous publications on iRPE14,18,36,37,38 and above the functional minimum required as stated by others39,40. Foetal RPE displayed a limited transepithelial resistance reaching approximately 150 Ohm*cm2 at day 42, confirming results obtained by Takahashi’s lab with fRPE from the same source38. However, fRPE from a different source reached a higher transepithelial resistance14,39, suggesting heterogeneity in fRPE lines that may stem from the age of the donor, isolation process or culture conditions.

Figure 4
figure 4

After maturation, iRPE cells present typical biological functions of RPE cells. (a) TER was measured on iRPE and fRPE monolayers at 14, 28 and 42 days in culture on transwell. Bars represents mean ± SD. (b) PEDF and VEGF secretion levels were measured by ELISA after 42 days in culture for both apical (blue) and basal (grey) compartments of n = 5 iRPE lines. Comparison between compartments was done by paired Wilcoxon match-pairs signed-rank test. Bars represent mean ± SD. (c–f) FITC-conjugated POS were fed to iRPE and fRPE cells for 16 h and the RHODOPSIN protein was immunolabelled without permeabilization (red labelling). As a result, the internalized POS fluoresce in green only, and microvilli-bound POS fluoresce in green and red. p: p-value. P-value significance threshold for all tests: 0.05.

Full size image

VEGF is preferentially secreted at the basal iRPE side

Factor secretion is another important feature of RPE cells1. Many proteins have been shown to be secreted by the RPE such as the vascular endothelial or pigment epithelium growth factor (VEGF and PEDF, respectively). To determine the secretive ability of the iRPE, the levels of both VEGF and PEDF were measured by ELISA (Fig. 4b) in the medium of the apical and basal compartments of transwells 24 hr after changing the medium. VEGF was preferentially secreted towards the basal compartments of the transwell, which corresponds to the choroid in vivo, as expected from previous publications on stem cell-derived RPE14,38,41,42,43 or adult human RPE44. On the other hand, although PEDF secretion level was comparable to previous reports14,36,38,43, it was not observed in a polarized fashion.

iRPE cells phagocytize bovine photoreceptor outer segments

In vivo, the RPE phagocytizes debris coming from the daily partial renewal of photoreceptor outer segments (POS)45. The internalization mechanism of the POS has been identified and it has been shown that the integrins involved in POS internalization are highly conserved between bovine and human46, allowing to test this particular pathway and not only observe passive uptake of the POS. To assess the iRPE phagocytic capacity, we fed fRPE and iRPE with 5 million bovine POS over 16 hr and performed the same treatment with fRPE in parallel as a comparison point. Both fRPE and iRPE were able to phagocytize bovine POS (Fig. 4c–f), although POS uptake was heterogeneous.

Lentiviral-mediated transduction of iRPE cells

Lentiviral vectors (LV) are suitable vectors to efficiently deliver a transgene in the RPE47,48 and to carry long construct containing either long transgene or complex regulatory sequences49,50. We thus tested LV in the human iRPE to evaluate the bona fide RPE identity of the iRPE as well as confirm the tropism of the lentiviral vector for the human RPE as observed in vivo in rodent24,47. We thus transduced iRPE and fRPE with 6 different lentiviral vectors (Fig. 5a) each bearing a construct to express GFP driven by different promoters: either RPE-specific (RPE65 short sequence: R0.8)20, ubiquitous (EFS) or specific for glial, photoreceptor or neuronal cells (GFAP, RHO, SYN1 and rat Eno2 respectively). To ensure that the lentiviral vectors transduced the cells, we determined by qPCR the lentiviral genome copy number relative to the iRPE or fRPE ACTB gene copies. All the vectors successfully transduced the cell, as revealed by the positive qPCR for the lentiviral genome amplification. To assess the activity of the transduced promoters, the percentage of GFP+ cells was linearly regressed to the lentiviral genome relative copy number. The ubiquitous EFS and specific R0.8 promoters showed a substantial prediction of the percentage of GFP+ cells by the lentiviral genome relative copy number (Fig. 5b,c) whereas none of the nonspecific promoters for RPE did, as assessed by the R2 coefficient of determination. In addition, only the EFS and R0.8 promoters presented significantly positive slopes of the best-fit line equations.

Figure 5
figure 5

GFP reporter gene expressions by lentiviral vector transduction reveals iRPE cells specificity. (a) Six different lentiviral vector constructs were produced bearing a GFP under ubiquitous (EFS), RPE specific (R0.8 = partial RPE65 promoter) or retinal but RPE aspecific (RHODOPSIN), glial (GFAP) or neural (SYNAPSIN and rat NSE) promoter. iRPE and fRPE cells were transduced using these vectors and the proportion of GFP-positive iRPE and fRPE cells was determined by FACS. In addition, the relative lentiviral genome copy number was quantified by qPCR. (b,c) EFS and R0.8 promoters showed a significant positive correlation between the percentage of GFP-positive cells (iRPE and fRPE together) and the relative lentiviral genome copy number whereas none of the aspecific promoters for RPE (d–g) did. R2: coefficient of determination of the linear regression. p: p-value from the F-test under the assumption of a null slope. P-value significance threshold: 0.05.

Full size image

RPE65 gene augmentation by lentiviral gene therapy

Finally, we investigated the response of the iRPE to a RPE65 gene augmentation treatment that has shown rod and cone function restoration in mouse models of Rpe65 deficiency23,24 and is also manageable to use in non-human primate retina25. We transduced iRPE cells with a multiplicity of infection (MOI) of 1 or 5 and measured RPE65 mRNA level and the lentiviral genome relative copy number by qPCR five days later (Fig. 6a). The RPE65 mRNA level was regressed against the lentiviral genome relative copy number (Fig. 6b), revealing a substantial prediction of the mRNA level by the lentiviral genome relative copy number. The expression of RPE65 mRNA was significantly upregulated after the gene augmentation treatment for both MOI 1 and 5 compared to untreated control (35.00 ± 21.27 and 137.34 ± 95.22 times more RPE65 mRNA compared to control, respectively; mean ± SD; Fig. 6c). The RPE65 protein level was also investigated by fluorescence immunoblotting densitometry (Fig. 6d,e) for MOI 5-transduced iRPE and showed an upregulation that was not significant (1.55 ± 0.63-fold of Ctrl, mean ± SD).

Figure 6
figure 6

Lentivirus-mediated RPE65 gene augmentation expression. (a) Schematic of iRPE cells transduced with a lentivirus expressing RPE65 cDNA under the control of the R0.8 partial RPE65 promoter. Lentiviral genome relative copy number, relative RPE65 mRNA and protein expressions were quantified five days later. (b) The linear regression between the relative lentiviral genome copy number and RPE65 mRNA relative expression showed a positive correlation. Dashed lines represent the 95% confidence interval of the regression curve. R2: coefficient of determination of the linear regression. p: p-value from the F-test under the assumption of a null slope. P-value significance threshold: 0.05. (c) Comparison of normalized RPE65 mRNA expressions between untreated control (Ctrl) and iRPE cells transduced with R0.8-RPE65 lentivirus at multiplicity of infection (MOI) 1 or 5. p: p-value of Dunn’s multiple comparisons test after Kruskal-Wallis test. P-value significance threshold: 0.05. Data are expressed as mean ± SD. (d) Representative near infra-red imaging of RPE65 and α-Actin proteins on immunoblot. (e) Western blot densitometry analysis of RPE65 protein level normalized to α-Actin between untreated (Ctrl) and R0.8-RPE65 lentivirus-transduced iRPE cells (MOI 5). Mean ± SD. Unpaired t-test p-value significance threshold: 0.05.

Full size image

The same experiment was repeated with two other RPE lines, namely fRPE and iCell RPE line, a stem cell-derived RPE line commercially available. Instead of 5 days, RPE cells were collected 14 days post-transduction to leave more time for the transgene to be expressed (transduction at day 28, collection at day 42).

Both iCell RPE and our iRPE line transduced with the RPE65 gene augmentation therapy presented a significant upregulation of RPE65 mRNA expression compared to their Ctrl (71.58 ± 32.63 and 7.94 ± 1.51-fold upregulation respectively; mean ± SD, Fig. 7a). fRPE showed a high but non-significant upregulation of RPE65 mRNA expression (280.73 ± 274.09-fold of Ctrl) probably due to its very low expression in the non-transduced Ctrl fRPE cells. While fRPE cells were at P3 in characterization experiments, they were at P5 in this experiment and their characterization at the time of transduction (Supp. Fig. 7) suggests a certain loss of RPE identity at this higher passage. Although all RPE markers tested were expressed, their expression was substantially lower than for our iRPE or the iCell RPE (Supp. Fig. 7a). In addition, immunohistochemistry (Supp. Fig. 7b) showed an absence of BEST1 protein, partial OTX2 expression and actin stress fibres that have been linked to RPE identity loss51.

Figure 7
figure 7

Lentivirus-mediated RPE65 gene augmentation expression in several RPE models, but poorly affects RPE65 protein level. iRPE, iCell RPE and fRPE cells were transduced with a lentivirus expressing RPE65 cDNA under the control of the R0.8 partial RPE65 promoter. Relative RPE65 mRNA and protein expressions were quantified  fourteen days later. (a) Comparison of normalized RPE65 mRNA expressions between untreated controls (Ctrl) and iRPE, iCell RPE or fRPE cells transduced with R0.8-RPE65 lentivirus at MOI 5. Dunn’s multiple comparisons test after Kruskal-Wallis test. P-value significance threshold: 0.05. Data are expressed as mean ± SD. (b) Comparison of RPE65 protein expression between untreated controls (Ctrl) and iRPE, iCell RPE or fRPE cells transduced with R0.8-RPE65 lentivirus at MOI 5 using unpaired t-tests. P-value significance threshold: 0.05. Data are expressed as mean ± SD. (c) Near infra-red imaging of RPE65 and GAPDH proteins on immunoblot. (d) Western blot densitometry analysis of RPE65 protein level normalized to GAPDH between untreated (Ctrl) and R0.8-RPE65 lentivirus-transduced iRPE, iCell RPE and fRPE cells (MOI 5). Mean ± SD. Unpaired t-test p-value significance threshold: 0.05.

Full size image

At the protein level, only the iRPE presented an upregulation (modest) of RPE65 by ELISA (1.17 ± 0.03-fold of Ctrl, 114.5 ± 6.11 vs 133.9 ± 3.56 pg/µg of total protein, Fig. 7b) and Western blotting (2.55 ± 0.66-fold of Ctrl Fig. 7c,d) whereas no upregulation was observed for iCell RPE (ELISA: 1.01 ± 0.06-fold of Ctrl, 118.7 ± 6.87 vs 120.5 ± 8.11 pg/µg of total protein; WB: 2.26 ± 1.67-fold of Ctrl) and fRPE (ELISA: 1.04 ± 0.05-fold of Ctrl, 99.9 ± 5.98 vs 103.9 ± 5.6 pg/µg of total protein). Worthy of note, RPE65 protein could not be quantified in fRPE due to a weak signal out of the range of the standard curve, which is consistent with the characterization results discussed previously. Although the outcomes were consistent between ELISA and WB, no correlation in RPE65 quantification was found between the two methods (R2:0.167, p-value= 0.116, n = 16).

Discussion

In this study, we modified a protocol to differentiate hiPSC into RPE and performed an extensive characterization of the resulting iRPE to validate its RPE identity. The majority of RPE marker genes investigated showed a significant upregulation after differentiation, whereas MERTK and EZR, two genes involved in the phagocytosis of POS52,53 did not. Several publications confirmed MERTK expression in ESC-derived RPE or iRPE by conventional PCR, while only a few investigated the MERTK expression level by qPCR. However, none compared iRPE and iPSC MERTK expression14,54. Only Liao and colleagues55 showed a conserved MERTK level after differentiation by microarray transcriptome analysis, suggesting that MERTK mRNA expression may play a role in pluripotent cells. Similarly, most publications investigated EZR protein expression by immunohistochemical staining, and none by RT-qPCR. Transcriptomic and proteomic stem cell data from repositories (respectively StemMapper56, and PepTracker®57) confirmed EZR expression in stem cells suggesting also a role of this gene in the pluripotent cells function. Although little differences of gene expression for MERTK and EZR were observed, the mRNAs levels of these genes were substantial revealing, along with the other markers, a good differentiation of the hiPSC into RPE.

In vivo, the RPE is responsible for recycling the 11-cis retinal chromophore through the visual cycle, in which RPE65 and LRAT proteins are involved. While RPE65 expression was upregulated, LRAT exhibited a downregulation in iRPE compared to iPSCs. To our knowledge, no publication investigated LRAT mRNA level by qPCR in iRPE compared to iPSCs, however conventional PCR results from most studies suggest a weak LRAT expression in iRPE and ESC-derived RPE5,15,58,59. In a previous work using RT-PCR, LRAT expression was not detected in the iPSC whereas it was upregulated during RPE differentiation60. This different result in comparison to the present work may be related to the sensitivity of the method used and/or stem cell line used to derive RPE. Transcriptomic and proteomic data respectively from StemMapper and PepTracker® confirm LRAT expression in stem cells. When examining the relative expression in iPSCs, it is interesting to note that LRAT is more expressed than KLF4 (Supp. Fig. 1a), a result that needs confirmation in other iPSCs lines. Thus, similarly to MERTK and EZR, LRAT upregulation in iRPE compared to iPSCs may not be indicative of RPE functions. A functional assay investigating the chromophore recycling capability of iRPE as performed by Maeda and colleagues59 could confirm this hypothesis. Moreover, previous reports have shown that the LRAT protein is 25’000-fold more active61 than the RPE65 one, supporting the hypothesis that LRAT does not need to be highly expressed to perform its function, in contrast to RPE65. Although LRAT expression is positive, but weak, one cannot exclude that our culture conditions are not optimal to promote the full maturation of the iRPE. Carr and colleagues62 have shown that iRPE in contact with retinal explant have an increased microvilli length, suggesting an adaptation of iRPE to the culture environment.

Regarding the expression of MITF-M, a melanocyte-specific isoform of MITF, an upregulation in iRPE compared to iPSCs was observed. Previous publications showed no expression of MITF-M in human immortalized RPE cell lines63,64,65. However, our results suggest that MITF-M expression in iRPE and fRPE is possible, although at a very low level, confirming results obtained by Maruotti and colleagues42. In comparison, the MITF-H isoform that is well expressed in RPE42,65 is approximately 300 times more expressed than MITF-M in our iRPE (Supp. Fig. 1a).

Lying at the interface between the choroid and the retina, the RPE helps to maintain the homeostasis of both tissues and is therefore a highly polarized epithelium. Immunohistochemical analyses showed a polarized distribution of some RPE marker proteins, as well as a polarized ultrastructure by electron microscopy. The iRPE displayed a tendency for a polarized secretion of VEGF but not PEDF, unlike previous reports have shown14,36,38,43.The difference in PEDF secretion polarization with other works on iRPE and ESC-derived RPE may originate from the culture conditions. Indeed, Kamao and colleagues38 cultured their iRPE on type I-A collagen that contains a PEDF binding site66,67 which could potentially buffer PEDF and limit its secretion to the basal compartment resulting in a polarized secretion of PEDF towards the apical compartment. Our iRPE were cultured on MRF that does not contain type I but type IV collagen to which PEDF does not bind68. The polarized secretion of PEDF by the RPE in vivo may also be the result of PEDF buffering by the Bruch’s membrane, composed of collagens type I and III69,70, among many other proteins. However, other publications reported a polarized secretion of iRPE on substrate without type I collagen14,43, suggesting that the RPE inherently secretes PEDF in a polarized manner. Interestingly, Singh and colleagues14 observed a polarized secretion with the same culture medium used in this study suggesting that the differentiation process may be responsible for this discrepancy. Improvement of the differentiation protocol by the addition of small molecules or proteins, such as Activin A, may help to promote a better differentiation and polarization4. Moreover, supplementation of the culture medium with ROCK inhibitor has shown significant results on F-actin stress fibre morphology of RPE cells51 and could be beneficial to iRPE polarization.

Overall, the iRPE displayed many features of bona fide RPE. Its RPE identity was further confirmed by the transduction of lentiviral GFP reporter constructs bearing RPE specific promoters or promoters specific for other cell types. As the differentiation protocol we used promotes a neural and then retinal fate, we tested promoters of lineages that could have arisen during differentiation, namely glia (GFAP), retina (RHO), and neuronal (SYN1, rat Eno2) specific promoters. Notably, only the specific RPE65 and ubiquitous EFS promoters drove GFP expression and no leakiness from aspecific promoters for RPE was observed.

Finally, we transduced iRPE cells differentiated from healthy iPSCs with an RPE65 gene augmentation approach previously tested on RPE65 deficiency mouse models23,24 and healthy non-human primates25. Interestingly, the RPE65 mRNA level could be substantially increased despite the iRPE already expressing RPE65 and the use of a partial RPE65 promoter. However, no marked increase of the protein level was observed. This experiment was repeated with two other RPE models, namely fRPE and iCell RPE cells (commercially available stem cell-derived RPE), and with a longer transgene expression period (5 vs 14 days). A significant upregulation of RPE65 mRNA expression was observed in iRPE and iCell RPE transduced cells, although the magnitude of the increase was lower in iRPE than in previous results for the same MOI (~10-fold vs ~130-fold). At the protein level, a significant but modest increase in RPE65 was observed by Western blotting and ELISA for the iRPE (in spite of the lower RPE65 mRNA upregulation), but none for the iCell RPE and fRPE. The longer transgene expression period might have helped to reach a significant RPE65 protein upregulation in iRPE although the magnitude in the increase remains modest, confirming a discrepancy between RPE65 mRNA and protein level increase. In total, three different batches of lentivirus were used (two for Fig. 6 and one for Fig. 7) which may explain the heterogeneity observed in RPE65 mRNA level expression post-transduction in iRPE. In addition, heterogeneity between iRPE batches may also be involved, although their characterization showed consistent results.

The difference of magnitude between the mRNA and protein upregulation in all 3 RPE models suggests a potential post-transcriptional regulation mechanism that has yet to be elucidated. Although inhibitory translational regulation of the RPE65 mRNA by miRNA has been described71,72, this regulation was mediated by the 3’UTR that is absent from the lentiviral construct containing only the RPE65 coding sequence. To our knowledge, no other publication reported post-transcriptional regulation mechanisms of RPE65. This mechanism of regulation may be of great interest for gene augmentation therapy since it suggests no potential toxic effects due to high gene doses. Nonetheless, such observations need to be confirmed in RPE deficient for RPE65 or expressing mutated forms of the protein.

Several clinical trials have been conducted to treat LCA2 caused by RPE65 deficiency73,74,75, one of which led to the first AAV-based gene therapy to reach the market76. Although AAV-mediated RPE65 gene augmentation therapy showed substantial vision improvements in the patients, none proved to prevent the progression of the photoreceptor degeneration observed77, contrasting with LCA2 animal model results23,78. This discrepancy may originate from the different levels of RPE65 protein needed to perform physiological functions between animals and human beings22, which suggests that a higher concentration of RPE65 is necessary to provide an optimal rescue in human. Bainbridge and colleagues74 observed a limiting dose of 1012 vector genomes inducing no toxicity, therefore precluding from using higher amounts to achieve higher RPE65 protein expression, and creating the need for a more potent gene therapy. Georgiadis and colleagues79 developed an AAV-based optimized RPE65 (OPTIRPE65) gene augmentation therapy 300 times more potent than their original work and were able to reach a plateau of rescue with only 109 vector genomes in the Rpe65 deficiency mouse model. Interestingly, when they tested their OPTIRPE65 on wild-type mice they observed physiological expression of the RPE65 transgene, meaning a doubling of the overall cDNA Rpe65 expression. Although comparison with our in vitro results is tricky due to differences in vector doses, vector biodistribution, model (endogenous RPE vs iPSC-derived) and methodology (absolute vs relative mRNA quantification), the significant increase in RPE65 mRNA level (up to 135-fold at MOI 5) suggests a high potential of the LV-based gene therapy. Further studies on the RPE65-deficient iRPE model are needed to confirm that this strategy is able to rescue RPE65 mRNA and protein to a physiological level, and which vector dose is needed to attain such concentration.

Lentiviral vectors have long been used in gene therapy, but their use has mostly been restricted to ex-vivo gene therapy in human80. Studies of lentiviral integration revealed that more than 60% of the insertions occur in transcription regions, but rarely in promoter sequences81,82, reducing the risk of tumorigenesis as the disruption of a promoter sequence generally impacts more genes than the disruption of a transcription region. Moreover, RPE is a post-mitotic tissue and adenomas of the RPE are seldom while adenocarcinomas are even rarer83, suggesting the RPE is a safer target for LV integration than most tissues. In addition to safety studies in non-human primate25, studying the safety of lentiviral-mediated gene therapy in in vitro models such as iRPE derived from RPE65 deficient patient or engineered by CRISPR/Cas9 to knockdown the RPE65 gene is an important preclinical step to elucidate untoward effect occurring in this cell type before considering its translation to clinic.

Conclusions and future directions

In conclusion, we modified an RPE differentiation protocol that produces iRPE bearing close features to bona fide RPE. The efficiency of this differentiation protocol should still be validated in other hiPSC lines. The iRPE revealed to be a suitable in vitro model for lentiviral gene therapy tests after transduction of several different constructs, confirming the lentivirus tropism for RPE previously observed in vivo. RPE65 gene augmentation therapy in iRPE cells resulted in high RPE65 mRNA expression corroborating previous results obtained in animal models23,24 but only elicited a modest RPE65 protein increase. Future directions include the production of an RPE65-deficient iRPE model to further test the lentiviral vector gene therapy and demonstrate its safety and efficacy in human cells.

Methods

RPE differentiation protocol and iRPE culture

A detailed protocol is available in Supplementary Methods (SM). Briefly, hiPSCs described in Singh et al.14 were cultured in feeder-free conditions in either mTeSR-1 or E8 media kit (StemCell) on Matrigel Embryonic Stem Cell grade (Corning). hiPSC clumps (between P18 and P30) were lifted using type IV collagenase (LifeTechnologies) and cultured overnight in free floating conditions in Embryoid Body Medium (EBM, see SM for composition). Resulting Embryoid Body-like aggregates (EBs) were plated on Matrigel Reduced Factor (MRF, Corning) in a neural induction medium (NIM also described in SM) changed every 3 days. On day 10, medium was switched to a Retinal Differentiation Medium with vitamin A (RDMcA). On day 13, medium was switched to a version without vitamin A (RDMsA). Pigmented foci (PF) typically appeared between days 18 and 24 and matured until collection at day 30. PF were plated on MRF as a purification and expansion step. Resulting pigmented patches were collected after 20 to 30 days, dissociated and plated on MRF at a density of 0.5 to 1·105 cells·cm−2 (P1 stage) to obtain pure RPE cell culture. NIM, RDMcA and RDMsA are adapted from6. Passage 3 iRPE cells at 42 days in culture on MRF-coated transwell (CLS3460, Sigma-Aldrich) were used for all assays of this publication unless otherwise stated. Images from Fig. 1(b–e) depicting the differentiation process were taken with a DMS300 microscope from Leica.

fRPE and iCell RPE culture

Human foetal retinal pigment epithelium (fRPE) cells were acquired from Lonza (Ref. 00194987, lot number 0000465423) and cultured according to the manufacturer’s protocol in RtEGM®Bulletkit® (Ref. 00195409, Lonza). Human iCell induced pluripotent stem cell-derived RPE (iCell RPE) cells were acquired from StemCell (Line 01279, Ref. C1047, lot103261) and cultured in RDMsA. For any assay performed on fRPE or iCell RPE, the cells were used at P3 and grown on MRF-coated transwells (CLS3460, Sigma-Aldrich) for 42 days in RDMsA (unless stated otherwise). All tests were performed with the same fRPE or iCell RPE line and, for fRPE, several maturation tests were done independently.

RNA extraction and analysis of gene expression

RNA isolation was performed using TRI Reagent® (T9424, Sigma-Aldrich) according to the manufacturer’s protocol. Retro-transcription PCR was performed using the High Capacity cDNA Reverse Transcription Kit from Applied Biosystems (Ref. 4368814) according to manufacturer’s protocol with a Biometra Tadvanced thermocycler from Analytik Jena AG. For each RNA sample, a negative control lacking the retrotranscriptase in the retrotranscription step was used in qPCR. Gene expression was assessed by qPCR using the LightCycler®96 and the FastStart Essential DNA Green Master mix (Ref. 06402712001) from Roche. Experimental design and gene expression quantification are explained in SM. ATP5F1B, SRP72 and GPI reference genes were chosen after having reviewed literature42,84 and tested their stability in iPS, iRPE, fRPE and iCell RPE. Primer sequences, primer efficiencies, qPCR programmes and additional information on qPCR data processing are available in SM.

Immunohistochemistry and imaging

iRPE, fRPE and iCell RPE cells were cultured on 12 mm diameter transwells, (CLS3460, Sigma-Aldrich) coated with MRF (at a 1:30 dilution) in RDMsA for 42 days (unless stated otherwise) before processing for immunohistochemistry (see SM for detailed protocol). iPS on coverslips were processed for immunohistochemistry at around 60% confluence. Briefly, cells were fixed in 4% paraformaldehyde for 10-20 min at room temperature (RT), rinsed with PBS and blocked for 3-5 h in 5% foetal bovine serum + 5% normal goat serum + 0.1% Triton X-100 in PBS 1×. Primary antibody was incubated overnight at 4 °C in blocking solution (see SM for antibody references and dilutions). Primary antibody was washed with PBS 1X before secondary antibody incubation (see SM) 1 h at RT in PBS 1×. After wash with PBS 1×, the transwell membrane was cut and mounted in Mowiol. Z-stacks were acquired using a LSM700 confocal microscope from Zeiss at the EPFL BioImaging & Optics Core Facility (EPFL-BIOP). Other images were acquired with a BX60 microscope from Olympus equipped with a DP72 camera or a Leica DM6B microscope equipped with a DFC9000GT camera.

Electron microscopy

iRPE cells cultured on 6.5 mm transwells (CLS3470, Sigma-Aldrich) for 45 days were washed with PBS 1×, fixed in 2.5% glutaraldehyde for 90 min and left in PBS 1X until processing and imaging by the Electron Microscopy Facility of the University of Lausanne (EMF-UNIL). Details of processing and acquisition are available in SM.

Transepithelial resistance measurements

Transepithelial resistance measurements were performed with an EVOM2 Epithelial Voltohmmeter (World Precision Instrument) on iRPE and fRPE cells grown on 12 mm diameter transwells. One transwell with only MRF coating was used as control. Transepithelial resistance was determined as the mean of three iRPE/fRPE transwell resistance measurements minus the control resistance times the transwell surface (1.12 cm2).

POS isolation and phagocytosis assay

Photoreceptor outer segments (POS) isolation and phagocytosis assay protocol were adapted from46 (see SM for detailed protocol). Briefly, fresh bovine eyes were collected from a local slaughterhouse and dissected in dim red light. Retinas were collected in a pre-chilled homogenization Tris buffer-based solution and ultracentrifugation was performed in a sucrose gradient (113’000 g during 48 min at 4 °C). The resulting pink band - characteristic of rhodopsin - was collected and washed three times. POS concentration was determined using a Neubauer chamber and aliquots of 5·106 POS were stored at -80 °C until use. Thawed POS were incubated with an FITC solution to obtain FITC-conjugated POS and fed to iRPE or fRPE cells for 16 hours (5·106 POS/transwell). iRPE or fRPE cells were washed to remove the non-internalized and unbound POS, immunolabelled for rhodopsin without the permeabilization step, and fixed in 4% paraformaldehyde before mounting on microscope slides for imaging.

PEDF and VEGF quantifications

Pigment epithelium-derived factor (PEDF) and vascular endothelial growth factor (VEGF) concentrations in medium were quantified using ELISA kits (RD191114200R from Biovendor, Brno, Czech Republic and RAB0507 from Sigma-Aldrich, respectively). iRPE cells cultured on MRF-coated transwells for 42 days were rinsed with HBSS twice and fresh RDMsA was added. Twenty-four hours later, the RDMsA was collected from the apical and basal chambers and stored at -80 °C. ELISAs were performed maximum one week after sample collection. For each iRPE cell line tested, triplicate samples (i.e. iRPE cell wells) were quantified in ELISA technical duplicates (i.e. ELISA wells). Mean of triplicate samples was calculated and Wilcoxon match-pairs signed-rank tests comparing contents of basal and apical chambers for VEGF and PEDF were performed using GraphPad Prism 6 software.

Lentiviral vector production and titration

Lentiviral vector and titration protocols have already been described in23, however detailed protocols are available in SM. Briefly, HEK293T cells were transiently transfected with a second generation lentiviral packaging system, culture supernatants were collected two days later and underwent two rounds of ultracentrifugation (70’000 g, 90 min, 4 °C) to wash and pellet the lentivirus that was resuspended, aliquoted and stored at -80 °C until use.

Lentiviral transduction of iRPE, iCell RPE and fRPE cells with GFP constructs driven by different promoters

The 6 different lentiviral constructs were originally designed in the following publications: hLox-Ro.8-GFP-WPRE23, hLox-EFS-GFP-WPRE85, hLox-GFAP-GFP-WPRE (Unpublished – details in SM), hLox-Rhodopsin-GFP-WPRE85, SIN-ratNSE-AcGFPnuc-WPRE86, and SIN-Synapsin-GFP-WPRE (Unpublished – details in SM). iRPE and fRPE cells were cultured on transwells for 42 days and transduced with one of the aforementioned lentiviral vectors at a multiplicity of infection (MOI) of 1. Five days later, iRPE and fRPE cells were collected, fixed and one part of the cells were analyzed at the Flow Cytometry Facility of the University of Lausanne (on a BD FACSCalibur® cell analyser, BD Biosciences) to determine the proportion of GFP-positive cells while genomic DNA (gDNA) was extracted from the other part to quantify lentiviral genome relative copy number by qPCR (details available in SM).

Lentiviral-mediated RPE65 augmentation therapy

The Good Manufacturing Practices (GMP)-compliant construct pCCL-R0.8-RPE65-WPRE425 was used to produce lentivirus delivering the RPE65 protein cDNA under the control of the partial RPE65 promoter R0.8. RPE cells were transduced with this lentivirus at targeted MOI of 1 and 5. Five (Fig. 6) or 14 (Fig. 7) days later, RPE cells were collected, RNA and gDNA extracted using TRI reagent® and protein isolated using a custom lysis buffer (details available in SM). RPE65 mRNA expression and lentiviral genome relative copy number were determined by qPCR (details available in SM).

RPE65 protein level was determined by an RPE65 ELISA kit according to the manufacturer’s protocol (XPEH2516, XpressBio, see SM for details) or Western blotting. For Western blotting, 40 to 50 µg total protein lysates were run on a 10% acrylamide gel, transferred to a PVDF membrane, the membrane was then blocked in Odyssey® TBS Blocking Buffer (P/N 927-50100, Licor), incubated overnight at RT with RPE65 and α-Actin or GAPDH primary antibodies, washed in TBS containing 0.1% Tween-20, incubated with near infra-red (680 nm and 800 nm emitting) conjugated fluorophore secondary antibodies and washed again. Blot imaging was performed with an Azure c600 imaging system and densitometry analysis was performed with the Image Studio Lite software from Licor. Detailed Western blotting procedure is available in SM as well as original blots in Supp. Fig. 6.

Statistical analysis

All statistical analyses were performed and graphs realized with Graphpad Prism 6 software. The number of biological replicates is stated on figures and in their caption, as well as statistical tests used and the p-value significance level. For the characterization experiments, one biological replicate corresponds to one iRPE line differentiated from hiPSCs. For RPE65 gene augmentation therapy experiments, one biological replicate corresponds to one transwell.