Restoration of Mesenchymal RPE by Transcription Factor-Mediated Reprogramming

Abstract

Purpose: Transforming growth factor β–mediated epithelial-to-mesenchymal transition (EMT) is a major component of the wound healing response and a negative determinant of retinal pigment epithelium (RPE) differentiation. We have shown previously that inhibition of TGFβ signaling restored the capacity of mesenchymal RPE to differentiate; however, the potential lessens with extensive passaging. We investigated TGFβ-independent mechanisms that regulate RPE differentiation following repetitive passaging.

Methods: Retinal pigment epithelium–EMT was induced by repetitive passaging of fetal RPE at subconfluence. Suppression of EMT was achieved by the addition of A-83-01, a TGFβ receptor kinase inhibitor. Transcriptomic analysis was used to identify potential TGFβ independent processes that prevent differentiation after extensive passage. Downregulated transcription factors were identified and transduced into highly passaged RPE to restore cell differentiation. Restoration was evaluated according to morphology, expression of RPE/mesenchymal markers, transcriptomic analysis, cell doubling time, and senescence-associated β-galactosidase (SA-β-gal) activity.

Results: A-83-01–treated RPE failed to differentiate after 7 passages (P7). This failure was concomitant with downregulation of RPE genes, misregulation of cell cycle genes, a decline in proliferative potential, and cell senescence. Exogenous expression of MYCN and OTX2 in conjunction with A-83-01 restored P7-RPE differentiation to a status similar to minimally passaged RPE. Moreover, the treatment allowed cells to maintain their differentiation capacity after extended passaging.

Conclusions: Retinal pigment epithelium subjected to chronic wound stimulus undergoes TGFβ-mediated EMT, loss of expression of signature RPE genes, and senescence. Targeting these aspects with a TGFβ receptor kinase inhibitor, a RPE transcription factor, and a cell cycle regulator restores the capacity of highly passaged RPE cells to regenerate and differentiate.

Dysregulated wound responses are a component of proliferative vitreoretinopathy (PVR) and neovascular age-related macular degeneration (nAMD). In PVR, this results in the formation of epiretinal or subretinal cellular membranes that can cause secondary retinal detachments and vision loss. In nAMD, this manifests itself as a vascularized fibrotic subretinal lesion that impedes RPE and neural retinal function. While multiple cell types and processes are involved in both disorders, common features are RPE cell proliferation and a chronic RPE-to-mesenchymal transition.^1–3 

While significant effort has been directed at identifying factors that promote proliferative RPE fibrosis, it is important to recognize that much of the behavior observed in PVR and nAMD is inherent to an intrinsic RPE wound repair program that is independent of exogenous factors. As evident by the behavior of primary RPE cells in culture, RPE cells have a limited capacity for productive wound repair. When differentiated RPE monolayers are dissociated and reseeded at moderate densities, they enter the cell cycle and undergo a reversible epithelial-to-mesenchymal transition (EMT). Once a confluent monolayer is reestablished, the RPE cells regain their prototypical phenotype. However, if the period of time that the cells are in a subconfluent state is prolonged by low density plating or repeated passage, the RPE cells undergo a lasting switch to a mesenchymal state.⁴ We recently have shown that a dominant mechanism underlying this behavior is a persistent, self-reinforcing activation of the TGFβ signaling pathway that is independent of continued wound stimulus.⁵ Not only can inhibition of TGFβ signaling using receptor kinase inhibitors forestall the onset of passage-dependent loss of phenotype, it can restore the capacity to differentiate to passaged cells that previously underwent the mesenchymal switch. 

Despite the effectiveness of blocking TGFβ signaling to inhibit expression of markers of classical EMT in passaged RPE, there are limits with regard to supporting RPE differentiation. Even in the continual presence of TGFβ signaling inhibitors and sustained suppression of EMT marker gene expression, the cells eventually lose the capacity to attain a characteristic RPE phenotype. This suggests that additional mechanisms are at play that eventually act to prevent RPE differentiation after protracted periods of wound stimulus and mitosis. We reported investigations directed at providing insight into the non-TGFβ-mediated processes that contribute to the demise of the RPE phenotype after extended periods of proliferative wound response. 

Methods

Lentivirus Production

Human EYA2, LHX, MYCN, and OTX2 open reading frames (ORF) were obtained from GeneCopia (Rockville, MD, USA; see Supplementary Table S1). Human RAX ORF was a gift from Dennis Clegg. Open reading frames were cloned into pLenti7.3/V5-DEST (Life Technologies, Carlsbad, CA, USA), LeGO-iCer, or LeGO-iV2 vector⁶ (Addgene, Cambridge, MA, USA). Packaging plasmids: pMDL, pCMV_VSVG, and pRSV_REV (Addgene) were used to generate lentiviral particles in HEK293T cells. Lentiviral vectors were concentrated with PEG-it (System Biosciences, Mountain View, CA, USA) and titered according to expression levels of coexpressed fluorescent proteins in transduced ARPE-19 using flow cytometry. Expression of designated genes was verified by quantitative (q) PCR (see Supplementary Fig. S1; qPCR information in Supplementary Table S2). 

Cell Culture and Lentiviral Transduction

Human fetal RPE cells were prepared and passaged as described previously.⁵ In general, fetal RPE were plated at 4 × 10³ cells/cm² on laminin-coated (Life Technologies) tissue culture plates or microporous supports (Millicell-HA Culture Inserts; EMD Millipore, Inc., Billerica, MA, USA) using a base medium described by Maminishkis et al.⁷ containing 5% heat-inactivated fetal calf serum. For differentiation, fetal RPE cells were seeded at 8 × 10⁴ cells/cm² and cultured in basal medium or basal medium with 500 nM A-83-01 (Tocris Bioscience, Bristol, UK) for 32 days. For transduction, frozen stocks of passage 7 fetal RPE cells (P7-RPE) were seeded at 1 × 10⁴ cells/cm² in basal medium. On the next day, the cells were transduced with the desired lentiviral vectors (multiplicity of infection [MOI] = 0.1 − 0.2 for each vector). Transduced cells were cultured in basal medium or basal medium with 250 nM A-83-01 as indicated. At 24 hours after transduction, the viral containing medium was replaced with fresh medium. ARPE-19 cells were seeded at 8 × 10⁴ cells/cm² and grown in basal medium with 250 nM A-83-01 for 7 days. For RNA sequencing (RNA-Seq), immunocytochemistry (ICC), and qPCR, cells were harvested on day 32 to 40. All experiments were done in triplicate, using cells from three different donors with experimental conditions and controls done in parallel. 

RNA-Seq and Bioinformatics

The detailed RNA-Seq methods and data can be accessed via the Gene Expression Ominbus (GEO: GSE78740). In brief, RNA-Seq libraries were prepared from Poly(A)+ RNA and sequenced using Ion PGM or Ion Proton next-generation sequencers (Life Technologies). The resulting sequences were aligned to the human transcriptome and genome (hg38) using a two-stage alignment pipeline employing TopHat2⁸ and Torrent Mapping (TMAP) read aligners. The number of reads per protein coding mRNA was determined using Partek Genomics Suite (Partek, Inc., St. Louis, MO, USA) and the dataset was normalized using the trimmed mean of the M-values method.⁹ Genes with read counts per million (RPM) ≥ 1 in three or more samples were selected and differential expression and statistical analysis were performed using the classic implementation of edgeR.¹⁰ Cluster analysis was done on log2 transformed data centered on the midpoint of the range using AutoSOME.¹¹ Gene ontology enrichment analysis was achieved using DAVID Bioinformatics Resources 6.7 with default settings.¹² 

Retinal Pigment Epithelial Reprogramming, Enrichment, and Passage

Passage 7-RPE cells from three different donors were seeded at 1 × 10⁴ cells/cm². On day 2, cells were treated with 250 nM A-83-01 and transduced with lentiviral vectors that coexpress MYCN with Venus, a yellow fluorescent protein (MOI = 1) and OTX2 with Cerulean, a blue fluorescent protein (MOI = 0.5). After 7 to 14 days, cells were harvested by trypsin digestion for 10 to 15 minutes at 37°C. Cell suspensions were washed with PBS three times before cell sorting. Venus and Cerulean double-positive cells were enriched by fluorescence-activated cell sorting (FACS) using a FACS Aria II SORP (BD Biosciences, Franklin Lakes, NJ, USA). Nontransduced P7-RPE cells served as negative controls for setting the gates. The collected Venus⁺/Cerulean⁺ cells then were plated in basal medium supplemented with 250 nM A-83-01. After 32 to 40 days the cells were harvested by trypsin digestion for 35 to 50 minutes at 37°C. Every 5 to 10 minutes cells were collected by gentle trituration. The resulting cell suspension was mixed with an equal volume of trypsin neutralization solution and fresh trypsin was added to the remaining attached cells. After 50 minutes any remaining cells were dissociated by scraping and the combined cell suspensions were filtered through a 40-μm cell strainer. Cell suspensions were washed three times with PBS. Pigmented cells then were isolated by cell sorting based on an increase in the side scatter to forward scatter ratio (Supplementary Fig. S2). These cells were named MOA-RPE to indicate the reprogramming treatment and the passage number was set to 9 to reflect that they were generated from P7-RPE and had been passaged twice during the reprogramming/enrichment processes. For passaging, MOA-RPE were plated at 4 × 10³ cells/cm² using medium supplemented with 250 nM A-83-01. For differentiation, MOA-RPE were plated at 8 × 10⁴ cells/cm² in basal medium supplemented with 250 nM A-83-01 for 7 days. For RNA-Seq, ICC, and qPCR, cells were harvested 32 to 40 days after plating. 

Immunocytochemistry

Cells grown on microporous supports were fixed with 4% paraformaldehyde for 10 minutes. Specimens were incubated with 5% normal donkey serum for 1 hour at room temperature to block nonspecific binding. Primary antibodies: mouse anti-ACTA2, 1:200 (Thermo Fisher Scientific, Waltham, MA, USA); mouse anti-BEST1, 1:100 (Alomone Labs, Jerusalem, Israel); mouse anti-PMEL, 1:50 (Dako, Carpenteria, CA, USA); Alexa Fluor 594 mouse anti-TJP1, 1:100 (Life Technologies); and rabbit anti-LUM, 1:200 (Abcam, Cambridge, UK) were incubated at 4°C overnight. Secondary antibodies: Alexa Fluor 488 or 546 donkey anti-mouse or rabbit IgG, 1:850 (Life Technologies) were incubated at room temperature for 1 hour. Nuclei were stained using Hoechst 33342, 1:2000. 

Cell Doubling Time and Senescence-Associated B-Galactosidase (SA-β-gal) Staining

To measure doubling time, cells were seeded at 1 × 10⁴ cells/cm² in basal medium. Before reaching confluence the cells were harvested and the number of cells per culture were determined by automated cell counting. For SA-β-gal staining, cells were seeded at 1 × 10⁴ cells/cm² in basal medium. After 2 days, the cells were fixed with 4% paraformaldehyde and stained using a SA-β-gal staining kit (Cell Signaling Technology, Danvers, MA, USA) following the manufacturer's protocol. 

Results

Failure to Differentiate in Highly Passaged RPE Is Associated With Decreased Expression of RPE Signature RNAs

To identify potential non-TGFβ–mediated processes that contribute to the passage-dependent loss of RPE phenotype, we performed a comparative analysis of passage 4 (P4) and passage 7 (P7) RPE cultures treated with the TGFβ receptor kinase inhibitor, A-83-01. As highlighted in Figure 1A and as we have previously reported,⁵ RPE cells passaged 4 times in basal medium and subsequently treated with A-83-01 exhibited a substantial degree of pigmentation and epithelial morphology, relative to untreated P4-RPE cells. In contrast, after seven passages in basal medium there was minimal improvement of morphology following A-83-01 treatment. Consistent with the gross phenotype of these cultures, principle component analysis and fuzzy clustering of the transcriptome profiles of the cultures depicted in Figures 1B and 1C demonstrated a rank order of likeness of P0-RPE > A-83-01–treated P4-RPE > A-83-01–treated P7-RPE > P4-RPE. Furthermore, in contrast to P4 cells, where A-83-01 treatment leads to recovery of epithelial and reduction of mesenchymal phenotype–associated gene expression, A-83-01 treatment of P7 cells was significantly more efficacious in the suppression of the expression of genes associated with the mesenchymal state than in the restoration of expression of genes associated with differentiated state (Figs. 1D, 2). On a systemic level, A-83-01–treated P7 cells misexpressed more than 1000 genes in comparison to P0 cells (>4-fold change and false discovery rate [FDR] < 0.01, Supplementary Material S1). Ontology analysis of these genes indicated that there is an overrepresentation of genes involved in extracellular matrix formation (n = 72, P < 1 × 10⁻²⁵), angiogenesis (n = 40, P < 1 × 10⁻¹⁵), cell growth (n = 38, P < 1 × 10⁻³), ion transport (n = 78, P < 1 × 10⁻⁶), and vision (n = 19, P < 1 × 10⁻³). Of the 1010 genes misexpressed in A-83-01–treated P7 cells, 290 also were misexpressed in A-83-01–treated P4 cells (Supplementary Material S2). 

Figure 1

 

Extensive passage results in a TGFβ pathway independent loss of ability to differentiate and a preferential decrease in expression of RPE-associated genes. (A) Bright field micrographs of P0-RPE (P0), P4-RPE (P4), A-83-01–treated P4-RPE (P4+A), and A-83-01–treated P7-RPE (P7+A). (B) Principal components and (C) fuzzy cluster analysis of the transcriptome profiles reveal that P0-RPE and A-83-01–treated P4 RPE are most similar, whereas untreated P4-RPE are most different. A-83-01–treated P7-RPE are intermediate to P0-RPE and P4-RPE. For the fuzzy cluster analysis the length, thickness, and opacity of the lines (edges) connecting the samples (nodes) are inversely proportional to the degree of similarity. (D) Genes with significant differences between P0-RPE and P4-RPE were identified as epithelial genes (highly expressed in P0-RPE, n = 548) and mesenchymal genes (highly expressed in P4-RPE, n = 684). The extent that A-83-01 could return the expression levels of these genes to P0 values was determined for P4- and P7-RPE. Each dot represents a gene. The blue line indicates the median of each group.

Figure 2

 

Expression profile of selected mesenchymal genes, RPE genes, and RPE transcription factors. In comparison to P4-RPE, expression levels of mesenchymal genes (A) were downregulated in A-83-01–treated P4-RPE (P4+A) and P7-RPE (P7+A). Retinal pigment epithelial genes (B) and RPE transcription factors (C) were highly expressed in P0-RPE, but most of these genes were downregulated in P4-RPE. Expression levels of the RPE genes and transcription factors were lower in P7+A than P4+A cells. Expression levels of selected genes were determined by RNA-Seq and plotted relative to the maximum value for that gene. (n = 3, ±SE).

Restoration of P7-RPE Cells by Lentiviral-Mediated Expression of MYCN and OTX2

The observation that A-83-01 largely inhibits the TGFβ-mediated induction of mesenchymal gene expression yet is relatively ineffective in restoring RPE signature and RPE transcriptional regulator gene expression in highly passaged cells, suggests that repeated passage may result in a loss of cell-type programming. To test this hypothesis, we investigated whether it might be possible to restore the capacity to differentiate using transcription factor–mediated reprogramming. Passage 7-RPE cells were transduced with a pool of lentiviral vectors coding for the expression of the four RPE transcriptional regulators with the least degree of rescue by A-83-01 (EYA2, LHX2, OTX2, and RAX; see Fig. 2C). To block the TGFβ-mediated aspects of EMT, A-83-01 also was included in the cocktail. For the purpose of an efficacious screening, pigmentation was used as an indicator of potentially successful restoration of RPE functions. Pigment was chosen over other characteristic RPE markers or functions because determining pigmentation requires nothing more than simple observation. As shown in Figure 3A, transduction with the 4-factor pool either with or without A-83-01 treatment did not lead to substantial improvements in RPE differentiation based on a lack of pigmentation. 

Figure 3

 

Inhibition of TGFβ signaling in conjunction with exogenous MYCN and OTX2 expression restores the differentiation capacity of P7-RPE. (A) P7-RPE cells transduced with a pool of lentiviral constructs coding for four RPE transcription regulators: EYA2 (E), LHX2 (L), OTX2 (O), and RAX (R) with or without A-83-01 (+A) failed to pigment. (B) Addition of MYCN to the pool with A-83-01 treatments results in substantial pigmentation. Selective removal of individual constructs identifies MYCN and OTX2 as essential reprogramming factors. Pink areas are the sights of basal medium trapped under doming RPE. (C) Higher magnification images of P7 cells transduced with different gene combinations and treated with A-83-01 demonstrate that reprogramming with MYCN, OTX2, and A-83-01 results in the development of a prototypical RPE morphology. Either MYCN or OTX2 alone when paired with A-83-01 result in some pigmentation; however, it is less abundant and the cells do not have a normal RPE morphology.

This lack of success in restoring the capacity to differentiate using RPE transcriptional regulators alone led us to consider whether there is more to the loss of phenotype with extended passaging than just the loss of RPE identity. In addition to the induction of mesenchymal gene expression and loss of RPE gene expression with passage, there is a decrease in the expression of genes associated with mitosis that is accompanied by a decrease in the rate of cell division.⁵ Moreover, it has been shown that stimulating mitosis with FGF2 can enhance RPE differentiation and delay the passage-dependent loss of phenotype.^5,13,14 Based on these findings we surmised that it might be advantageous to include MYCN in the transcription factor cocktail to stimulate mitosis and as a possible reprogramming facilitator.¹⁵ The transcription factor MYCN was chosen over MYC, because MYCN expression level is higher than MYC during mouse retina development¹⁶ and MYCN expression level is 3.6-fold lower than MYC in A-83-01–treated P7-RPE cells (Supplementary Fig. S3). The addition of MYCN to the transcription factor pool, in conjunction with A-83-01, resulted in numerous patches of pigmentation in P7-RPE (Fig. 3B). To assess which of the four regulators of RPE transcription were essential for this result, a series of transductions was performed where one of the RPE transcription vectors was omitted from the cocktail. Of the four RPE transcription factors, only the absence of OTX2 led to substantial loss of capacity to acquire pigmentation. In addition, removal of A-83-01 from the medium also prevented the accumulation of pigment, thus demonstrating that the TGFβ-mediated induction of mesenchymal genes overrides RPE differentiation. To determine the minimal combination of reprogramming transcription factors needed to reestablish the capacity to pigment, P7-RPE were transduced with MYCN/OTX2, just MYCN, or just OTX2 and maintained in basal medium containing A-83-01. The best results were obtained with the MYCN/OTX2 combination. Minimal pigmentation was observed with just OTX2 and those cells that did pigment displayed an atypical morphology (Figs. 3B, 3C). A fair degree of pigmentation was seen with MYCN alone (Figs. 3B, 3C), but the efficacy was variable and RPE donor dependent (Supplementary Fig. S4). 

Using this information, we then developed a protocol for the efficient isolation of differentiation competent RPE cells lines from P7-RPE cells (Fig. 4A). In brief, P7-RPE cells were treated with 250 nM A-83-01 in basal medium and cotransduced with MYCN and OTX2 lentiviral vectors that coexpress either the yellow fluorescent protein, Venus, or the blue fluorescent protein, Cerulean. After 7 to 14 days, cells expressing both fluorescent proteins were isolated by fluorescence-activated cell sorting and allowed to mature for 32 to 40 additional days in the presence of 250 nM A-83-01. At this stage, the cultures contained pigmented and nonpigmented populations of cells. Based on the premise that pigmentation is an essential feature of RPE cells, it follows that the nonpigmented cells were not successfully reprogrammed. Therefore, a second round of enrichment was used where the pigmented cells were isolated by cell sorting based on a pigmentation-dependent increase in right angle light scattering (Supplementary Fig. S2). We termed these pigmentation competent cells MOA-RPE and they were designated passage number 9 as they were generated from P7-RPE and were passaged twice during the reprogramming/enrichment procedures. In a separate experiment it was determined that only 7 days of A-83-01 treatment were required to obtain the maximal number of pigmented reprogrammed cells (Supplementary Fig. S5). For further analyses, the pigmented sorted MOA-RPE were cultured with A-83-01 for 7 days, then allowed to mature for an additional 25 to 33 days in basal medium. As shown in Figure 4B and Supplementary Figure S6, this protocol effectively generated relatively homogenous cultures of pigmented cells from different donors, which can maintain a prototypical RPE morphology after passaging. Nevertheless, the reprogramming and enrichment procedures were not optimized to perfection; some mesenchymal cells remained. Unfortunately, these cells often die over time and prevent the characterization of the cultures using methods that require an intact monolayer; such as transepithelial electrical resistance measurement. Instead, immunocytochemistry and transcriptomic analyses were used to access the MOA-mediated restoration of RPE markers as presented below. In contrast, in cells transduced with MYCN alone and enriched by the same procedure, the percentage of pigmented cells decreased when the cultures were subsequently passaged. A-83-01–treated P7-RPE cells failed to differentiate following the enrichment procedure. 

Figure 4

 

Isolation of MOA-RPE cell lines. (A) A schematic illustrating the enrichment procedure. P7-RPE cells were plated and maintained in medium containing A-83-01. After one day, the cells were transduced with lentiviral vectors that express MYCN-IRES-Venus or OTX2-IRES-Cerulean. Seven to 14 days after transduction, Cerulean (blue fluorescent protein) and Venus (yellow fluorescent protein) double-positive cells were enriched by FACS. The collected cells then were cultured for 32 to 40 days to allow for differentiation and the accumulation of pigment and pigmented cells were isolated by cell sorting based on the light scattering property of melanosomes. (B) Enriched A-83-01–treated OTX2 and MYCN transduced cells (MOA-RPE) can maintain their differentiation capability and yield a homogenous culture. Most of the A-83-01–treated enriched MYCN transduced cells failed to differentiate after two passages. P7-RPE cells treated with A-83-01 alone failed to yield significant quantities of pigmented cells.

Reprogramming With MOA Restores RPE Morphology and Gene Expression

In addition to an evaluation based on pigmentation, the MOA-RPE cells were assessed by immunocytochemistry. Passage 0-RPE cells expressed high levels of RPE markers (TJP1, BEST1, and PMEL), but mesenchymal markers (LUM and ACTA2) were absent (Fig. 5). In contrast, P7-RPE cells, expressed high levels of LUM and ACTA2 with minimal to no RPE marker expression. Reprogramming with MYCN/OTX2/A-83-01 effectively suppressed expression of the mesenchymal markers, restored the expression and subcellular localization of the RPE markers, and allowed for the establishment of a prototypical cobblestone morphology. 

Figure 5

 

Reprogrammed RPE have a prototypical RPE phenotype. Cells were allowed to differentiate and the expression and localization of RPE and mesenchymal markers were evaluated by immunocytochemistry. Antibody labeling is indicated in green and nuclei were stained with Hoechst 33342. Without treatment with A-83-01, P7-RPE expressed low levels of RPE markers (TJP1, BEST1, and PMEL) and high levels of mesenchymal markers (LUM, ACTA2). Exogenously expressing MYCN and OTX2 with A-83-01 treatment effectively restored the expression and localization of RPE markers to that seen in P0-RPE and attenuated the expression of mesenchymal markers.

To achieve a more global assessment of the degree of reprogramming, MOA-RPE cells also were evaluated using RNA-Seq. Multidimensional scaling (MDS) analysis showed that P0-RPE, MOA-RPE, 32-day A-83-01–treated P4-RPE group together, thus demonstrating that their gene expression profiles are relatively similar (Fig. 6A). Distantly removed from the differentiation competent RPE cells were untreated P4-RPE cells and 7-day A-83-01–treated P7-RPE cells. Intermediate to these two groups was P7-RPE treated with A-83-01 alone for 32 days. Interestingly, the commonly used immortal RPE cell line, ARPE-19, was the most dissimilar of all the cells that were examined. The same results were obtained using differential expression analysis to quantify the number of differentially expressed genes among the cell cultures (Fig. 6B), with only 3.7% of the expressed genes in MOA-RPE cells being differentially expressed (>4-fold change and FDR < 0.01) relative to P0-RPE. 

Figure 6

 

Reprogramming restores the RPE transcriptome. (A) Multidimensional scaling plot of 32-day P0-RPE (P0), P4-RPE (P4), P4-RPE treated for 32 days with 250 nM A-83-01 (P4+A), P7-RPE treated for 7 days with 250 nM A-83-01 (P7+A7d), P7-RPE treated with 250 nM A-83-01 for 32 days (P7+A), MOA-RPE, and ARPE-19. For each sample group, cells from three different donors were assessed. For ARPE-19, experimental triplicates were used. (B) Percentage of differentially expressed genes (>4-fold changes and FDR < 0.01) for each sample group in comparison to P0-RPE. (C) Cluster analysis to visualize overall expression pattern of MOA-RPE, P0-RPE, P7-RPE plus 7 days of 250 nM A-83-01, and ARPE-19. Color scale indicates the log₂ transformed relative expression level centered on the midrange. (D) Gene ontology enrichment analysis of the top seven clusters.

Genes with similar expression pattern were clustered to provide a better visualization of the overall expression profile of MOA-RPE cells using P0-RPE, 7-day A-83-01–treated P7-RPE, and 7-day A-83-01–treated ARPE-19 cells as comparators (Fig. 6C; see Supplementary Material S3 for detailed clustering results; Supplementary Fig. S7 for specific genes discussed below). The top-eight clusters were evaluated using gene ontology enrichment analysis to identify functional relationships of the genes in each cluster (Fig. 6D). Clusters II and VI contain genes related to RPE function, including visual cycle (RPE65, RDH5, RLBP1), pigmentation (PMEL, TYR, TYRP1, MITF), ion and water transport (AQP11, BEST1, TRPM1, TRPM3), lipid metabolism (ABCA4, APOE, LPL), and growth factor-associated (SERPINF1, FGFR2) that were highly expressed in P0-RPE and MOA-RPE cells. Epithelial-to-mesenchymal transition-associated genes (including FN1, LUM, ITGA5, ITGA11, TGFB2, TGFBI, THBS1, VIM), that are highly expressed in 7-day A-83-01–treated P7-RPE but attenuated in P0-RPE and MOA-RPE cells, are predominantly in clusters III and V. Cluster VIII contains genes that were uniquely upregulated in MOA-RPE; however, there is no clear functional enrichment associated with this group. In addition to the genes identified in the cluster analysis, statistical analysis revealed that 171 genes differentially expressed in P4 cells, were not rescued in either A83-01–treated P4-RPE cells or MOA reprogrammed P7-RPE cells (Supplementary Material S4). Among the genes whose expression did not return to levels seen in P0-RPE, inhibitors of DNA-binding proteins (ID3 and ID4) may be the most interesting (see Discussion). Beyond enrichments for signal peptide (n = 60, P < 1 × 10⁻⁸) and plasma membrane (n = 53, P < 1 × 10⁻³), there were no significant ontology enrichments among the genes that were not restored by A-83-01 or MYCN and OTX2. 

MOA Reprogramming Enhances Proliferation and Expression of Cell Cycle-Associated Genes Expression

A progressive decrease in the rate of proliferation and dysregulation of cell cycle gene expression are other major changes associated with serial passage of primary RPE. In addition, the MYC family, which includes MYCN, is well-known for its involvement in normal cell cycle progression and oncogenic transformation. With this in mind, we investigated the effect of MOA treatment on the cell cycle. First, the effects on the rate of proliferation were assessed by determining the doubling time. In P10-MOA-RPE the cell doubling time was significantly reduced to a level similar to P0-RPE (Fig. 7A) that was maintained for at least 6 more passages (Fig. 7A). The expression levels of genes associated with the cell cycle were evaluated next. As shown in the Table, only 14% of the cell cycle genes (KEGG Pathway: hsa04110) were differentially expressed (Benjamini-Hochberg P < 0.1) in MOA-RPE relative to control P0-RPE. This is in stark contrast to P7-RPE and ARPE-19 cells treated with 250 nM A-83-01 for 7 days, where 45% and 80% of the cell cycle genes have altered expression. A total of 25% of cell cycle genes were dysregulated in P7-RPE maintained in the continual presence of A-83-01. 

Figure 7

 

Reprogramming rejuvenates the RPE growth potential and results in sustained differentiation potential. (A) Cell doubling time and (B) Percentage of SA-β-gal–positive cells of P0-RPE (P0), P7-RPE (P7), P7-RPE treated with 250 nM A-83-01 for 32 days (P7+A), P10–MOA-RPE, and P15-MOA-RPE (n = 3, ±SE, *P < 0.01). (C) Immunocytochemistry of RPE markers (TJP1, BEST1, and PMEL) and mesenchymal markers (LUM, ACTA2) on P15-MOA-RPE indicates that after extensive passaging, MOA-RPE retained full differentiation capability. Antibody labeling is indicated in green. Nuclei (blue) were stained with Hoechst 33342.

Table

 

KEGG Orthology Group Gene Expression

 

Cell growth, proliferation, and differentiation require ribosomes to newly synthesize numerous enzymes, transcription factors, and structural proteins. Previous studies have shown that MYCN can improve protein biosynthesis by directly increasing RNA expression of cytosolic ribosomal proteins (cRP) and translation factors.¹⁷ Similar to these previous reports, we observed that 40% of the expressed cRP genes were significantly upregulated in the MOA-RPE compared to P0-RPE. Less than 3% of the basic transcription or translation components examined were significantly different in MOA-RPE relative to P0-RPE (Table). On the other hand, in ARPE-19 and P7-RPE treated with A-83-01, expression of the transcription and translation components were substantially different than P0-RPE. These results suggest that MOA treatment may result in improved protein biosynthesis through increased cRP transcript expression and restoration of transcription and translation machinery. 

MOA-RPE Are Rejuvenated and Retain the Capacity to Differentiate After Passage

Decreases in the proliferative potential of a population as a whole can result from a portion of the cells reaching terminal senescence and generally is associated with the expression of SA-β-gal. An increase in the SA-β-gal activity has been reported in serially passaged human primary RPE and RPE in elderly rhesus monkey eyes.^18,19 Similar to previous reports, we observed a significant increase in the number of SA-β-gal–positive cells in P7-RPE, and this phenomenon could not be prevented or reversed by A-83-01 (Fig. 7B). Conversely, reprogramming with MOA led to a reduction of SA-β-gal–positive cells to levels similar to or below what is observed in P0-RPE that remained constant after six additional passages. Although the MOA-RPE cells behave as if they may be immortal, they do not appear to be transformed based on the lack of evidence of ongoing mitosis in mature cultures (data not shown). Moreover, MOA-RPE retain the capacity to attain a prototypical RPE morphology after repetitive passage (Fig 7C, Supplementary Fig. S8). 

Discussion

The failure to differentiate in highly passaged RPE is concomitant with (1) misregulation of TGFβ-mediated EMT, (2) global downregulation of RPE gene expression, and (3) a decrease in the rate of cell proliferation. In a prior study, we have shown that the self-reinforcing activation of the TGFβ pathway is a dominant process driving passage-dependent EMT and loss of RPE phenotype. We showed that the loss of developmental RPE programming or identity is another major contributor to loss of phenotype. Using transient TGFβ pathway inhibition in conjunction with exogenous expression of MYCN and OTX2 we were able to substantially restore the proliferative and differentiative potential to very highly passaged RPE. In these very highly passaged cells, inhibition of TGFβ signaling, MYCN, and OTX2 are all essential; omitting one of them jeopardizes the restoration. 

Possible Roles of OTX2 and MYCN in the Regulation of RPE Restoration

The transcription factor OTX2 is a master RPE regulator that can synergistically control RPE specification and differentiation with MITF.^20,21 In addition, OTX2 can interact with more than a thousand genomic loci in adult mouse RPE to maintain key RPE functions, such as melanogenesis, retinol metabolism, and ion/metal transport.²² Conditional knockout of OTX2 in adult mouse RPE not only disrupts expression of RPE genes, but also leads to photoreceptor degeneration and induces an inflammatory response in the retina.²² Evident by the importance of OTX2 in RPE development and differentiation, it seems likely that the exogenous expression of OTX2 may help to reestablish RPE gene expression networks and restores RPE identity. In addition, roles of OTX2 in cell cycle regulation also were reported. Despite OTX2 being highly expressed in quiescent cells, such as photoreceptors and RPE, its role in regulating the cell cycle in proliferative cells is largely context-dependent. During development, OTX2 maintains proliferation of mesencephalic dopaminergic progenitors in a dose-dependent manner.²³ In contrast, genetic ablation of OTX2 in thalamic progenitors resulted in an increase in the proliferation activity.²⁴ This inconsistency also was observed in medulloblastoma cell lines. In OTX2 amplified medulloblastoma lines, OTX2 directly regulates expression of cell cycle genes and promotes cell proliferation.²⁵ However, overexpressing OTX2 in non-OTX2 amplified medulloblastoma cell lines, MED8A and DAOY, resulted in cell senescence.²⁶ This discrepancy may be due to chromosome accessibility or presence of transcription coregulators of OTX2 in different cell types. 

The basic helix-loop-helix (bHLH) transcription factor MYCN, is a member of the MYC protein family (MYC, MYCN, MYCL). Members of the MYC family share a homologous transactivation domain and redundantly orchestrate multiple cellular processes.^27–31 MYC family members are famous for their roles in oncogenesis, cell cycle regulation, and control of cell potency. In addition, recent studies have shown that MYC family proteins universally regulate genes associated with RNA and protein biosynthesis.^17,32 As evident by the upregulation of ribosomal genes in the MOA-RPE, it is possible some of the beneficial effects of overexpressing MYCN directly relate to general improvement of protein biosynthesis and, therefore, improve overall cell growth. Furthermore, as well as acting as primary transcription factors, MYC family proteins can bind to thousands of additional genomic loci to potentially facilitate transcription of multitudes of genes; thus, acting as a universal transcriptional amplifier.^33–36 It is possible MYCN is acting similarly here to facilitate the action of OTX2 as well as other RPE transcription factors whose levels have decreased with passage, hence boosting overall RPE gene expression. Finally, MYC family proteins regulate cell potency. MYC is one of the original Yamanaka factors to generate induced pluripotent stem cells (iPSCs),³⁷ and codeletion of MYC and MYCN abolished the pluripotency.³⁸ Later, it was shown that histone deacetylase inhibitors such as valproic acid (VPA) and trichostatin A, can replace MYC during the reprogramming process.^39,40 In line with roles of MYC family proteins in histone modification and euchromatin regulation,^34,35,41 it is possible that MYCN may reestablish an epigenetic status that favors OTX2 restoration of RPE gene transcription networks, and consequently facilitate the reprogramming process. 

The transcription factor MITF is a master controller of RPE terminal differentiation that physically interacts with OTX2 through its bHLH domain to regulate expression of RPE genes.⁴² Inhibition of DNA binding proteins (ID1-4) are bHLH proteins without a DNA binding domain that dimerize with bHLH transcription factors and abrogate their DNA binding and transcription activity.⁴³ It has been reported that IDs interfere with MITF function in osteoclast differentiation,^44,45 thus it is possible that a similar mechanism may be applied to regulate RPE differentiation. In minimally passaged RPE, ID1-4 are expressed in subconfluent cultures, but are downregulated as the cells differentiate (previously published data⁵). In contrast, highly passaged RPE fail to downregulate ID3 and ID4 expression and this misregulation cannot be restored by inhibition of the TGFβ signaling pathway or by MOA treatment. These observations suggest that IDs may act as upstream negative regulators of RPE differentiation. It is possible that MYCN is acting by directly binding to IDs, thus preventing their inhibition of MITF and allowing for its normal interaction with OTX2. 

Cellular Senescence and MOA-Mediated Restoration

An organism's biological functions progressively decline over its lifespan. Despite numerous biological factors (genetic variants, accumulation of DNA damage, telomere shorting, stem cell regulation, and dysregulation of the immune system)^46–49 that have been associated with this deteriorative process, it is generally a consequence of cells being unable to regenerate and maintain their normal functions. In vitro senescence can be introduced through repetitively passaging primary cells,⁵⁰ in which cells progressively lose their proliferative potential and eventually reach their replication limit. Senescent cells often have different morphology, metabolic activity, gene expression, and secretome when compared to their original cells.^50,51 It has been appreciated that RPE cells undergo a similar process after repetitive passaging. Highly passaged RPE cells have lower proliferation potential, develop a mesenchymal phenotype, have higher SA-β-gal activity, and possess a distinct gene expression profile. In this study, we have shown that MOA treatment “restores” highly passaged RPE cultures to a status similar to its original state and allows cells to be repetitively passaged without showing signs of reoccurred cellular senescence. Whether the reprogramming process prevents the onset of senescence or reverses senescence remains to be determined. 

Indications for PVR and AMD

Currently, surgical removal of retinal membranes and anti-VEGF therapeutics are the best available clinical management strategies for PVR and nAMD, respectively.^52,53 However, while the anatomic success rate is relatively high for PVR, visual outcomes can be poor and recurrence is relatively common. In the case of nAMD, VEGF blockers are quite effective at limiting immediate loss of vision. Nevertheless, nearly 50% of those treated with anti-VEGF blockers have retinal scars and loss of visual acuity within 2 years.⁵⁴ To date, there are no approved pharmacologic treatments directed at preventing retinal fibrosis. Although the proliferation-dependent loss of RPE programming reported here may contribute to the progressive and persistent nature of RPE fibrosis in PVR and nAMD, it seems doubtful that reversing this process would be desirable once RPE leave the basement membrane and scar formation ensues. Doing so could lead to pigmentation of the ectopic RPE and further visual decline. Promoting a productive wound response to prevent the undesired migration and establishment of scar formation may be a more effective approach for these conditions. 

From a therapeutics perspective, our findings relating to the loss of RPE programming may be most relevant to regenerative medicine approaches directed at treating advanced dry AMD or geographic atrophy (GA). Progressive RPE atrophy is a hallmark feature of GA and while the major focus is on the cell death component of this process, it is important to recognize the contribution of the failure of RPE to repair or regenerate with disease advancement. The lessons learned in this study give further insight into the inherent limitations of RPE to regenerate. Any future efforts directed at treating AMD by eliciting and enabling an intrinsic RPE repair response will likely have to overcome this limitation. In theory, it may be possible to overcome this barrier using a reprogramming-based approach; however, finding pharmacologic agents that block the loss of RPE identity may hold more promise. In addition, agents directed at extending the effective lifespan of RPE would lead to improved efficiencies in the production of stem cell-derived RPE for treatment of AMD and improvements in the use of primary RPE in cell culture-based studies. 

Acknowledgments

The authors thank Dean Bok and Jane Hu for the original cultures of human fetal RPE used in this study. 

Supported by the BrightFocus Foundation (M2011064; MJR), California Institute for Regenerative Medicine (LA1-02086; PC), Beckmann Initiative for Macular Research (BIMR-1312; PC and MJR), and the Garland Initiative for Vision Research Grant which provided funding for this study or provided funding for or resources used by this study. 

Disclosure: Y.-H. Shih, None; M.J. Radeke, None; C.M. Radeke, None; P.J. Coffey, None