Morphological, biochemical, and transcriptomic characterization of iPSC-derived human RPE cells from normal and Smith-Lemli-Opitz syndrome patients

Basal media and antibodies

Basal culture media and serum were Dulbecco’s modified Eagle medium (DMEM) with HEPES (D6171, Millipore Sigma, Burlington, MA), DMEM without calcium chloride (21,068,028, Thermo Fisher Scientific, Grand Island, NY). KBM-Gold™ (192,151, keratinocyte basal medium, calcium ion-free; Lonza, Basel, Switzerland) and bovine calf serum (SH30073, HyClone Laboratories Inc., Logan, UT). Antibody and concentrations for Western blotting were anti-BEST1 (bestrophin) rabbit polyclonal at 1:1000 (LAgen laboratories, Rochester, MN); anti-CRALBP (cellular retinaldehyde binding protein) rabbit polyclonal (1:1000, 15,356, Proteintech, Rosemont, IL); anti-RPE65 rabbit polyclonal (1:1000, GTX103472, Genetex, Irvine, CA); anti-β-actin mouse monoclonal (1:2000, 8H10D10, Cell Signaling Technology (CST), Danvers, MA); anti-ezrin (1:1000, 3145, CST); anti-vimentin (1:1000, ab92547 clone EPR3776, Abcam, Cambridge, MA), anti-E-cadherin monoclonal G10, (1:1000, sc-8426, Santa Cruz, Santa Cruz, CA), anti-α-tubulin (TUBA1) monoclonal, clone DM1A, (1:2000, T6199, Millipore Sigma); anti-α-smooth muscle actin (SMA) rabbit polyclonal (1:1000, ab5694, Abcam). Secondary alkaline phosphatase conjugated antibodies were obtained from Millipore Sigma: goat anti-rabbit IgG (A9919, 1:10,000) and goat anti-mouse IgG (AP124A, 1:10,000). For immunohistochemistry, we used anti-occludin monoclonal clone OC3-F10 (1:100, 33–1500, Thermo Fisher Scientific), anti-CRALBP clone B2, (1:100, NB100–74392, Novus Biologicals, Centennial, CO), anti-zonula occludens-1 [a.k.a. tight junction protein-1, TJP1] rabbit mAb D6L1E (1:100, 13,663, CST), fibronectin rabbit mAb E5H6X, (1:200, 26,836, CST), and a phalloidin-Alexa Fluor 568 conjugate to detect F-actin (1:500, OD044-T, Biotium, Freemont, CA).

RPE cell lines and culture conditions

Human iPSC-derived RPE cells originating from skin fibroblasts obtained initially from a “mildly” (CWI)-affected 6-month-old female, as well as from a normal neonatal male patient (Dys0100, DYS; ATCC, cat #ACS1019), were generated as described in detail previously [19]. Herein, we have further extended the analysis by including a “severe” SLOS phenotype. Briefly, A2 SLOS iPSC Clone 3F#1 was used at passage 17, and CWI SLOS iPSC Clone 4F#5 was used at passage 23 [20] for generating CWI and A2 SLOS iPSC-derived RPE cells (Neural Stem Cell Institute, New York, USA). The iPSC-derived RPE cells were then further passaged until P5–P6 (as described below) and used for further experimentation. We reconfirmed the SLOS genotypes in the CWI and A2 SLOS iPSC-RPE cells by PCR from isolated genomic DNA and direct sequence analysis (Figure 1E). Genomic DNA isolated from ARPE-19 cells (ATCC, cat#CRL2302), a commonly used transformed RPE cell line, was included as a control.

Passages 5–6 non-immortalized iPSC-RPE cells were seeded in 12-well tissue culture dishes (TCP011012, JET biofil, Elgin, IL USA) coated with a poly-L-ornithine solution (at 5 ug/cm²; P4957, Millipore Sigma) at full confluence (~400,000 cells/well) and maintained at 37 °C under 6% CO₂-air atmosphere at 80% humidity. Conditions for the culture of iPSC-derived RPE cells were as described previously [19,21], with “high-calcium” and “low-calcium” media preparation and supplement formulation and addition, as described and enumerated previously in detail [22,23]. Briefly, we seeded and passaged the cells in “low-calcium” medium precluding the formation of calcium dependent adhesions to form a loosely bound, budding monolayer (DMEM plus calcium-free, keratinocyte basal medium, supplemented with growth factors and hormones, plus 1% (v/v) bovine calf serum (BCS) and 0.5% bovine retina extract (BRE)), prepared as described previously [21]. Cells were then switched after culturing for 3–7 days to defined “high-calcium” media (with supplements and 0.9% (v/v) BCS, and 0.25% (v/v) BRE), enabling synchronized cell–cell adhesion formation and promoting optimal differentiation with subsequent maturation of iPSC-RPE cells in culture [21,24]. All three iPSC-RPE cell lines were maintained in the presence of Plasmocin™ (ANTMPT, InvivoGen, San Diego, CA, at 37.5 μg/ml initial concentration, with 12.5 μg/ml maintenance dosage) to ensure the cultures were cleared and free of mycoplasma contamination, and were routinely verified using a commercial mycoplasma detection and removal kit (B39032, Bimake.com; Houston, TX). The iPSC-RPE monolayers were maintained in a “high-calcium” culture medium for at least two months before use. For routine inspection and imaging, cell cultures were inspected using an Axiovert 25 CFL inverted photomicroscope (Carl Zeiss Microscopy LLC, Thornwood, NY), equipped with an Evolution™ MP digital CCD 5.0 MP camera (MediaCybernetics, Rockville, MD). Digital images were captured and stored as .tiff files on a standard PC computer.

For experiments where the steady-state sterol composition (see below) of cells was assessed, the iPSC-RPE cell cultures were maintained as follows. Mature cellular monolayers cultured on 24-well plates were initially seeded and grown in 0.9% (v/v) BCS-containing “high-calcium” medium for about 3 months (see above), and were then transferred to “high-calcium” medium containing 0.9% (v/v) delipidated (sterol-depleted) calf serum (DLCS). The latter was prepared as previously described [25], and Chol depletion was verified to be >98% [19]. The final medium preparation excluded BRE supplementation. In parallel, cells maintained in a 0.9% BCS-containing medium were analyzed for sterol content. Prior to harvesting the cells for sterol analysis, the monolayers were washed three times in sterile PBS.

Immunohistochemical (IHC) analysis

For imaging by confocal fluorescence microscopy, iPSC-RPE cells were seeded in “low-calcium” medium on alcohol/UV-sterilized Rinzl plastic coverslips (72,261-22, Electron Microscopy Services (EMS), Hatfield, PA), coated with polyornithine solution (5 µg/cm²), and switched after 2–7 days of culture to “high-calcium” medium to promote the formation of maturing polygonal iPSC-RPE cell monolayers. The cells were washed with PBS and then fixed in PBS-buffered 4% paraformaldehyde. Free aldehydes were quenched by treatment with a solution of 100 mM glycine (pH 8.0) for 15 min at room temperature. Blocking and permeabilization were performed by incubating the cells in Tris-buffered saline with 0.1% (v/v) Tween (TBST), supplemented with 0.5% (w/v) BSA, 1% fish skin gelatin (w/v), and serum from appropriate host species corresponding to secondary antibodies (5%, v/v), for 45 min at room temperature. Primary antibodies were diluted as indicated above (Basal media and antibodies) in TBST containing 0.5% BSA, incubating overnight at 4 °C in a humidified chamber. Unbound antibodies were removed by three washes in TBST. We detected primary antibody binding by incubating the samples with appropriate dye-conjugated secondary antibodies (corresponding to host species of primary antibodies) at 1:500 dilution in TBST. Cell nuclei were counterstained with DAPI (4′,6-diamidino-2-phenylindole dihydrochloride). Rinzl coverslips were mounted onto glass microscope slides using Fluoro-Gel with PIPES buffer (1,798,540, EMS) face-side-up. Images were acquired with a scanning laser confocal fluorescence microscope (Leica TCS SPEII DMI4000, Leica Microsystems, Buffalo Grove, IL), applying the same laser settings for imaging DYS, A2, and CWI cells in each experiment, and using either a 40X or a 63X oil immersion (RI:1.518) objective.

Oil Red O staining of iPSC-RPE cells

All three RPE cell lines were cultured (n = 3/cell line) on Rinzl coverslips in high calcium-containing media for ~2.5 months. Detection of lipid droplets was performed by Oil Red O (MAK194, Millipore Sigma) staining, as per the manufacturer’s protocol. Oil Red O working solution was prepared by adding 3 parts of Oil Red O solution to 2 parts of water, followed by filtration (after 20 min of incubation), using a 0.4 µm filter. RPE cells were fixed in 4% formaldehyde (buffered in PBS) for 10 min on ice. The fixative was washed with PBS three times, followed by treatment of coverslips with 60% isopropanol for 5 min. Cells were treated with freshly prepared Oil Red O working solution for 20 min at room temperature. Cells were washed with deionized water at least 3 times to remove any precipitate. Cells were incubated in Mayer’s hematoxylin solution (MHS32, Millipore Sigma) for 1 min, followed by three washes in water. Cells were mounted and coverslipped, and light images were captured using an Olympus BH2 light/epifluorescence microscope.

IVS8G-C mutation analysis in SLOS iPSC-RPE cells

DHCR7 mutations in CWI and A2 iPSC-RPE cells were confirmed by PCR amplification of the appropriate genomic region (T93M mutation and intron-exon junction IVS8G-C), as described elsewhere [19]. Normal control iPSC-RPE (DYS) cells and ARPE-19 cells did not contain DHCR7 mutations. Primers used to amplify the mutant region were as follows: Fwd 5ʹ – CAGAGGCAGAGCTGGGG – 3ʹ and Rev 5ʹ- CAGCAGGCGGTAAGGCA – 3ʹ (yielding a 557 bp PCR product upon gel electrophoresis). Gel-excised PCR products (QIAquick® Gel Extraction Kit, Qiagen, cat. #28704) .were sequence verified (pay-for-service) at Roswell Park Cancer Institute (RPCI) Genomics Shared Resource (Buffalo, NY). Sequence readouts corresponding to the IVS8G-C region from ARPE-19 cells and DYS, CWI, and A2 iPSC-RPE cells were aligned and compared using the Clustal Omega Multiple Sequence Alignment program [26].

Steady-state sterol analysis

Sterol analysis of cultured iPSC-RPE cells was performed essentially as described in detail elsewhere [19,27]. In brief, cells were saponified, the non-saponifiable lipids were extracted with petroleum ether, the extract was dried under Argon gas and re-dissolved in MeOH, and analysis was performed using reverse-phase HPLC. The response factors and retention times for authentic standards of Chol and 7DHC were determined empirically. The following retention times were obtained: Chol, 16.65 ± 0.1 min; 7DHC, 15.1 ± 0.1 min. The integrated peak areas for Chol and 7DHC peaks were quantified using Flo-ONE^™ software (Packard Instrument Co., Inc., Meriden, CT), and the 7DHC/Chol mole ratios were calculated; the data were subjected to statistical analysis using an unpaired Student’s t test, with the cutoff for statistical significance set at p <0.05 (n = 3/group).

RNA isolation, library preparation, and sequencing

Total RNA was extracted from individual iPSC-RPE samples using a RNeasy Plus Mini Kit (Qiagen; Germantown, MD), according to the manufacturer’s instructions. RNA quantity and quality were determined using an RNA Nano 6000 kit (Agilent Scientific Instruments, Santa Clara, CA) on an Agilent BioAnalyzer 2100. RNA libraries were prepared from 100 ng of total RNA of each sample with a RIN >9.0 using a SureSelect Strand-Specific RNA Library Prep kit (Agilent), according to the manufacturer’s protocol, and each sample was uniquely indexed for multiplexing. Library sample quantity and quality were determined using the DNA High Sensitivity kit (Agilent) on an Agilent Bioanalyzer 2100. Six samples were multiplexed and clustered at 8 pM in two lanes of a flow cell. We performed a 101 bp paired-end sequencing on an Illumina HiSeq 2500.

SDS-PAGE/Western blot analysis

Protein lysates from iPSC-RPE were prepared from triplicate monolayers cultured concurrent with those harvested for RNA isolation, using RIPA buffer (9806, CST, Danvers, MA), supplemented with protease inhibitors (P8340, Millipore Sigma). Samples were sonicated (4X for 3 s pulses, on ice), and clarified by centrifugation at 15,000 × g for 15 min at 4 °C. The supernatants were assayed for total protein content using a Micro BCA™ Protein Assay Kit (23,235, Thermo Fisher Scientific), and equal amounts of total protein were subjected to SDS-PAGE (10% Bis-Tris NP0316, Thermo Fisher Scientific). Proteins were electrophoretically transferred to low-fluorescence PVDF membrane (1,620,264, Bio-Rad, Hercules, CA) for Western blot analysis as previously described [28] and probed with primary antibodies and alkaline phosphatase conjugated secondary antibodies (see the section Basal media and antibodies). Antibody binding was detected using a chemifluorescent enzyme substrate (45–000–947, Cytiva, Marlborough, MA) and imaged using a Chemidoc MP system (Bio-Rad). Densitometric analyses were performed using ImageLab® 5.0 software (Bio-Rad), and probed target band intensities were normalized to α-Tubulin levels. Data were subjected to statistical analysis using an unpaired Student’s t test, with the cutoff for statistical significance set at p <0.05 (n = 3/group).

RNA-seq data analysis

Reads were aligned to the human genome version hg38 analysis set, which does not include alternate contigs, using STAR (v2.5.2b). A non-redundant transcriptome database was derived from Gencode v24 and lincRNA annotation databases available on the UCSC Genome Browser [29] Raw counts for each exon were generated using RSubread (v1.24.1) and were summed over each isoform [30,31]. Only reads uniquely mapped to one genomic location were included in the counts. Uniquely mapped reads falling within overlapping isoforms were counted as belonging to both isoforms, unless a splice junction unique to one particular isoform of a gene had zero reads, in which case that isoform was considered not expressed and given a zero count. Splice junctions were quantified using the output provided by the STAR alignment. Differential expression analysis was performed using DESeq, assessing all pairwise comparisons, and transcripts having at least twofold expression change and an adjusted p <0.01 were considered significantly differentially expressed. Differentially expressed gene sets were imported into Panther to determine pathways with overrepresented transcripts. Downstream analyses were performed using custom scripts in Perl and R.

Comparison of control and SLOS iPSC-RPE cell morphology and genotype

We evaluated morphological differences between control (DYS) and SLOS-derived (CWI, A2) iPSC-RPE cells by phase-contrast microscopy and IHC analysis of mature, confluent monolayer cell cultures. Figure 1 shows IHC images obtained using anti-cellular retinaldehyde-binding protein (CRALBP, Figure 1A), fluor-conjugated phalloidin (to label F-actin; Figure 1B), and anti-zona occludens protein (tight junction protein 1 (TJP1); Figure 1C). Mature DYS and CWI RPE cells exhibited cytoplasmic CRALBP immunolabeling, albeit with noted cell-to-cell variability in staining intensity (yellow asterisks, Figure 1A). By contrast, CRALBP expression was low to absent in mature A2 RPE cells (white asterisks,Figure 1A). The F-actin cytoskeleton distribution in DYS RPE cells appeared predominantly circumferential (yellow arrow,Figure 1B), with occasional sparse stress fibers. By comparison, CWI and A2 RPE cells exhibited increased actin stress fiber formation (white arrow, Figure 1B) while maintaining areas of normal circumferential F-actin distribution (yellow arrowheads, Figure 1B), consistent with observations by others regarding cultured iPSC-derived RPE cells [32]. A2 RPE cells were highly pleomorphic, exhibiting marked variability in cell shape and size, regular polygonal packing, areas with highly irregular cell arrangement and borders, and F-actin stress-fiber formation (white arrows, Figure 1B). At the tight junctions, DYS RPE cells exhibited intense cell border-specific TJP-1 immunolabeling (yellow arrows,Figure 1C). However, the TJP1-positive cell borders in some CWI RPE cells appeared irregular (white arrows,Figure 1C). A2 RPE cells formed continuous ZO-1 borders with both straight cell–cell contacts and areas exhibiting highly irregular and undulating TJP1 immunolabeling patterns (white arrows,Figure 1C). Bright-field images of parallel cultured monolayers used for transcriptomics and protein analyses are shown in Figure 1D. Both DYS and CWI RPE monolayers exhibited a polygonal cobblestone-like appearance typical of normal mature RPE cells, with pigmentation (melanin) and phase-bright cell borders. By contrast, A2 RPE cells exhibited more heterogeneity, exhibiting areas with a regular polygonal arrangement, and areas exhibiting highly irregular cellular packing, density, and shape (Figure 1D). Genotype and DNA sequence analyses (Figure 1E) verified that DYS RPE cells were negative for the IVS8G-C DHCR7 mutation (large G peak, green arrow, Figure 1E), whereas CWI RPE cells were confirmed heterozygous for the IVS8G-C DHCR7 mutation (small G peak, merged multi-C peak, yellow arrow, Figure 1E), and A2 RPE cells were homozygous for that mutation (G peak absent, merged multi-C peak, red arrow, Figure 1E). ARPE-19 cells served as biologic controls and were negative for the IVS8G-C DHCR7 mutation.

Occludin is a member of the tetraspanin 4 superfamily (containing four transmembrane α-helices) of proteins, which binds to cytoplasmic ZO1/TJP1 domains at tight junctions and is a key regulator of paracellular permeability in epithelial cells, including RPE cells [33,34]. DYS RPE cells exhibited co-labeling with anti-occludin antibody and fluor-conjugated phalloidin (the latter binds specifically to F-actin), with sharply defined cell borders and predominantly continuous anti-occludin labeling at the tight junctions (Appendix 1). CWI RPE cells exhibited some irregular cell borders, with areas of undulating occludin labeling, whereas A2 RPE cells exhibited areas of discontinuous or highly undulating occludin labeling (yellow asterisks, Appendix 1). As described above, both CWI and A2 (SLOS-derived) RPE cells exhibited prominent stress fiber formation (white arrows, Appendix 1).

We further characterized the regional epithelial-mesenchymal transition (EMT) of A2 RPE cell monolayers by performing co-immunolabeling against fibronectin, a key extracellular matrix (ECM) component and an EMT marker in vitro and in vivo [35,36], and against occludin. Although DYS RPE cell monolayers were mostly uniformly labeled along their cell borders for occludin (red arrows, Appendix 2) and circumferential F-actin (yellow arrows, Appendix 2), they generally lacked fibronectin expression. A2 RPE cell monolayers exhibited some fibronectin-negative patches that possessed occludin-positive cell borders (red arrow, Appendix 2) and occludin-negative patches with marked fibronectin upregulation (yellow asterisk, Appendix 2), as well as increased actin stress fiber formation (yellow arrows, Appendix 2), indicating regional EMT. We noted that A2 RPE cell monolayers exhibited modest improvement in polygonality and phase-bright appearance of cell borders when cultured on poly-L-ornithine-coated Rinzl coverslips compared to growth on poly-L-ornithine-coated tissue culture plastic. However, A2 RPE cells cultured on poly-L-ornithine-coated Rinzl coverslips still underwent EMT, which varied in severity, with areas of marked stress fiber formation and areas exhibiting both normal and irregular cell packing (see above, and Figure 1, Appendix 1 and Appendix 2).

Characterization of overall gene expression

We generated RNA-Seq libraries (n = 6 technical replicates each) from each of the concurrently cultured iPSC-RPE cell lines. The libraries were sequenced on a HiSeq 2500, generating between 70 and 120 million raw reads. Two of the DYS samples failed to cluster before sequencing and were removed from further analysis. Gene expression was first characterized for the whole transcriptome to determine the quality of the sample preparations. Using STAR to align reads to the hg38 human genome build, over 80% of the reads from each sample were aligned. We used the GENCODE v24 annotation database to generate counts for both genes and isoforms [37]. We determined the specific isoforms expressed in the samples using the methods described previously [31]. Of the 65,439 genes and 113,246 isoforms in the annotation database, we detected 14,255, 14,891, and 13,521 genes and 31,976, 33,291, and 28,688 isoforms (RPKM >1) in the A2, CWI, and DYS iPSC-RPE samples, respectively (Figure 2A). Intra-sample concordance was high among the samples (range: 0.96–0.99). Inter-sample concordance, however, was relatively high between A2 and DYS (0.90) RPE cells but was lower when we compared CWI to DYS (0.78) and A2 to CWI (0.65) RPE cells. In concordance with the negative binomial distribution of the RNA-Seq data, most transcripts had an RPKM <1, which we considered to indicate no expression (Figure 2A). This is in agreement with a principal component analysis showing that the replicate samples were clustered together, although each condition was separate (Appendix 3).

We then compared the overlap of isoform expression. Among the three sample types, 25,589 isoforms were expressed in all, which corresponds to 77%–89% of the total isoform expression. However, the two SLOS patient-derived samples showed the highest degree of overlap (29,165). In contrast to the inter-sample concordance, CWI and DYS RPE cells showed a higher degree of overlap in isoform expression (27,317 isoforms) than A2 and DYS RPE cells (26,027 isoforms; Figure 2B).

We also considered the RPKM distribution of isoform expression as a function of relative transcript levels. We counted the number of transcripts across 6 RPKM bins: 0–1 (not expressed), 1–10, 10–100, 100–1000, and greater than 1000. Aside from the 100–1000 RPKM bin, in which CWI RPE cells showed more transcripts than average, the data across the remaining bins were consistent among the sample types (Figure 2A).

To examine the global expression differences between the conditions, we performed differential expression analysis using DESeq. We chose DESeq over DESeq2 to minimize the number of observed false positives [38]. We considered a transcript to be differentially expressed if it showed a fold change greater than 2 and an adjusted p <0.01, and we found significant differences in expression between all conditions. We saw the largest number of changes between A2 and DYS RPE cells, where we found 20,091 transcripts either significantly upregulated or downregulated (Figure 2C). Of these, the largest number (8,654 transcripts) fell into the 2- to 5-fold change group. Interestingly, the second largest number (4,251 transcripts) was either turned on or off in the A2 iPSC-RPE cell line. Comparing CWI and DYS RPE cells, we found 14,192 differentially expressed transcripts (Figure 2C). Again, most of these transcripts fell into the 2- to 5-fold change (7,348 transcripts) and the on/off group (3,403 transcripts). We also found many transcripts significantly different between the two SLOS phenotypes (“mild” versus “severe”), with 18,430 differentially expressed transcripts between the A2 and CWI RPE cells, 8,439 with a 2- to 5-fold change, and 2,933 in the on/off category (Figure 2C).

Protein and transcriptome characterization of RPE and EMT biomarker expression

We analyzed the protein and transcriptome expression profiles of key RPE and EMT markers. Protein lysates were prepared from mature (2.5 months in culture) control and SLOS-derived RPE cell monolayer cultures (n = 3 per iPSC RPE cell line). Parallel monolayers were used for Western blot and RNA-Seq analyses (see Methods, Figure 3B). Transcriptomic analysis revealed that several known EMT marker transcripts were expressed at significantly higher levels in CWI SLOS and A2 SLOS RPE compared to DYS, and that a subset was upregulated in A2 with respect to CWI (Appendix 4). Protein-level expression of RPE markers (CRALBP, BEST1, and RPE65) was observed in all three iPSC-RPE cell lines by Western blotting, albeit at different levels (Figure 3A,B). The two SLOS-derived iPSC-RPE lines exhibited significantly higher β-actin (ACTB) levels compared to DYS iPSC-RPE cells, despite uniform protein content loading (Figure 3A). Therefore, α-tubulin (TUBA1A) content was used as the “housekeeping” protein for normalization, due to its lower variance among the three cell lines than ACTB (Figure 3A, B). The visual cycle protein CRALBP was expressed at similar levels in CWI and DYS iPSC-RPE cells, but A2 iPSC-RPE cells exhibited significantly lower levels (~0.08-fold, normalized to TUBA1A, relative to DYS, p <0.01, n = 3; Figure 3B). The cytoskeletal linker Ezrin was expressed robustly in both DYS and CWI cells, albeit at slightly lower levels in CWI after normalization, whereas A2 SLOS expressed ~5-fold lower levels of Ezrin compared to DYS cells, indicating less epithelial differentiation and polarization (p <0.01, n = 3; Figure 3B). Expression of RPE65 and E-cadherin were significantly higher in CWI iPSC-RPE cells compared to DYS RPE cells (~4-fold and ~5-fold, respectively, p <0.01, n = 3, Figure 3B), whereas A2 iPSC-RPE cells exhibited the lowest expression levels of these differentiation markers (~0.5- fold and ~0.1-fold, respectively, relative to DYS, p <0.01, n = 3; Figure 3B). Bestrophin-1 (BEST1), a calcium-signaling regulator, was expressed in all three iPSC-RPE cell lines at basal levels, with lower levels in both SLOS iPSC-RPE lines versus DYS iPSC-RPE cells (p <0.05; Figure 3B). Vimentin, a known EMT marker in vivo [39], was robustly expressed in all three mature iPSC RPE lines (Figure 3A,B). However, it is well established that vimentin is expressed in immortalized bovine and other RPE culture models in vitro [40,41]. Interestingly, we observed a typical degradation pattern for vimentin, producing a species smaller by ~8–10 kDa than full-length ~58 kDa vimentin, consistent with its processing by either calpain or caspase family proteases [42,43]. A2 SLOS RPE expressed modestly higher levels of vimentin protein compared to CWI and DYS RPE, in agreement with the transcript levels of vimentin across the three cell lines (Appendix 4). We then compared the levels of α-smooth muscle actin (SMA), an established EMT marker. Compared to DYS iPSC-RPE cells, we observed significantly elevated levels of SMA in both A2 (~20-fold, p <0.01, n = 3) and CWI (~3.7-fold, p <0.01, n = 3) iPSC-RPE cells (Figure 3B). Transcriptomic analysis showed significant increases in the expression levels of several EMT markers (SNAI1, SNAI2 (SLUG), TWIST1, ZEB1, CDH2 (N-Cad), CDH1 (E-Cad), VIM, HMGA2) in A2 and CWI SLOS RPE cells versus DYS iPSC-RPE cells (Appendix 4).

We employed the RPE signature gene set developed by other [44], which consists of 86 genes and 292 isoforms [31]. We found that both CWI and DYS iPSC-RPE cells expressed 85 genes, consistent with our previous findings, whereas A2 iPSC-RPE cells expressed 82 of the 86 RPE signature genes. This corresponds to 225, 245, and 224 isoforms of RPE signature genes for DYS, CWI, and A2 RPE cell lines, respectively. A comparison across the three sample types of the relative expression levels of the RPE signature genes showed that A2 RPE cells had more low-expressing genes, whereas CWI and DYS RPE cells had more medium-to-high-expressing genes (Figure 3C). We compared the distribution of RPE signature gene expression levels across all three RPE cell lines and found that A2 cells had more transcripts with RPKM <10, but CWI and DYS cells had more transcripts with RPKM greater than 100 (Appendix 5A). This pattern was confirmed by the observed concordances. A Pearson correlation of CWI and DYS cells showed that the samples were tightly correlated (0.92), but very low concordance was observed between either CWI (0.05) or DYS (0.13) and A2 iPSC-RPE cell lines (Appendix 5B). In general, although the A2 cells expressed genes consistent with being RPE cells, their expression levels were dissimilar to those of both CWI and DYS cells.

Although gene expression data suggested that the A2 and CWI iPSC-RPE cells were distinguishable from DYS iPSC-RPE cells, pathway analysis revealed several common upregulated biologic process pathways in common between the two. DYS iPSC-RPE cells were also highly pigmented (melanized), consistent with RAB38 (ras-related protein Rab38) and TYRP1 (tyrosine-related protein 1) expression, as seen in transcriptome analysis (Figure 3C,D (inset)). By comparison, both SLOS-derived RPE cell lines, especially CWI cells (with lower TYRP1 expression, Figure 3C), exhibited reduced melanin content compared to DYS cells (Figure 3D, (inset)). Comparisons of A2 and CWI RPE cell data to the DYS line showed that eight of the top 20 most significant pathways are shared between the two disease lines (Figure 3D). These include features known to be important to RPE cells, including cell junction assembly, cell-substrate adhesion, extracellular matrix organization, and response to oxidative stress. Interestingly, the individual comparison of A2 versus DYS iPSC-RPE cells revealed that four actin pathways were significantly overrepresented in the former, which corresponds to the actin findings in the phenotype experiments (Figure 3D).

Sterol quantification of RPE cell lines

We tested the effects of DHCR7 mutations on sterol biosynthesis in the three RPE cell lines. Mature RPE cells (2.5-month cultures) were maintained in a sterol-replete medium containing 0.9% (v/v) bovine calf serum (BCS). Under this culture condition, RPE sterol homeostasis is driven by a combination of cellular de novo sterol synthesis and endocytic uptake of Chol-laden LDL particles [45]. Reverse-phase HPLC analysis of non-saponifiable lipid extracts from DYS, CWI, and A2 iPSC-RPE cells revealed that A2 cells cultured in sterol-replete medium exhibited significantly elevated 7DHC/Chol mole ratios (0.63 ± 0.08, avg. ± SEM, p <0.01, n = 3), as compared to DYS and CWI iPSC-RPE cells, which contain negligible amounts of 7DHC (Figure 4, top panel). Surprisingly, CWI iPSC-RPE cells exhibited a normal sterol profile (comparable to that of DYS cells) when cultured in BCS-containing medium (Figure 4, top panel), despite harboring verified compound heterozygous DHCR7 mutations (Figure 1E). However, when mature A2 and CWI iPSC-RPE cells (initially cultured in medium containing 0.9% BCS for ∼2 months) were switched to medium containing 0.9% delipidated calf serum (DLCS without serum lipoproteins/Chol) for 9 days, they both exhibited increased levels of 7DHC formation and accumulation, as demonstrated by elevated 7DHC/Chol mole ratios (0.75 ± 0.02 and 0.1 ± 0.005, respectively), compared to DYS RPE cells maintained in delipidated medium (Figure 4, top panel). However, A2 iPSC-RPE cells consistently exhibited significantly higher 7DHC content than CWI iPSC-RPE cells under both experimental conditions (Figure 4, top panel). This observation is consistent with our prior report of 7DHC accumulation in CWI iPSC-RPE cells cultured in a medium containing 0.9% DLCS, suggesting that these cells possess an active Chol homeostasis process modulated by the presence or absence of exogenous (serum-borne) lipoproteins in the medium [19,46].

Previous studies have shown significant lipid and sterol accumulation in SLOS patient fibroblasts [47]. We independently observed lipid droplet accumulation in the RPE of the AY9944 rat model of SLOS [18]. We used Oil Red O staining to test for lipid droplet accumulation in both SLOS-derived and control cultured iPSC-RPE cells (mature cultures maintained in 0.9% BCS-containing medium for 2.5 months on Rinzl coverslips; n = 3/cell line). We consistently observed substantially greater lipid droplet accumulation in A2 RPE cells than in CWI RPE cells; very few of the latter exhibited lipid droplets (black arrows, Figure 4, bottom panel). By contrast, DYS RPE cells appeared healthy and did not show lipid storage abnormalities (Figure 4, bottom panel).

Limitations of the present study

To characterize individual iPSC-RPE lines and assess differentiation, several limitations need to be considered. A limitation of the use of reprogrammed iPSC-derived RPE cell models is the low passage numbers compared to immortalized, transformed cell lines such as ARPE-19 [54–56]. Differentiated iPSC-derived RPE cells in culture may lose their epithelial-like characteristics and exhibit loss of specialized cell–cell contacts [57], reversion to a fibroblastic-like phenotype, and acquisition of mesenchymal expression profiles and migratory properties; that is, they may undergo EMT [58,59]. The reorganization of F-actin and stress fiber formation are also additional indicators of EMT [60]. We identified the challenges in delineating iPSC-associated versus SLOS mutation-associated dedifferentiation/EMT that might have occurred in our cultured cell lines. However, comprehensive iPSC-based modeling of rare disorders, such as SLOS, is challenging. The generation of multiple iPSC-RPE lines per donor, along with appropriate isogenic controls, is time/labor/resource-intensive. The appearance of “normal” phase-bright regions among dedifferentiated areas of the same monolayers (that is, as seen in CWI and A2 SLOS-derived RPE cells) also suggests heterogeneity in the cultures, and that there may be more than one set of drivers for EMT. The observed elevation in the expression level of SMA, an EMT marker, in A2 and CWI RPE cells could be driven partly by the decreased functionality of DHCR7 in those cells. Indeed, the level of RPE marker expression observed in differentiated iPSC-RPE cells from multiple biologic and clonal sources is likely to vary from culture to culture, and we acknowledge that varied levels of EMT may arise under different culture conditions, for example, when using different basement membrane adhesion substrates or other growth conditions. These limitations of iPSC-derived RPE cells have been discussed elsewhere [61]. However, when iPSC-RPE cells were seeded and cultured under identical conditions, with identical (replicate) cultures used for both biochemical and transcriptome analyses, we were able to discern distinct control versus SLOS-dependent cellular phenotypes within the limits of the in vitro system used. The degree to which the observed EMT of A2 iPSC-RPE cells is disease-driven versus iPSC-driven is a complex issue. However, we have repeatedly observed that A2 iPSC-RPE cells are prone to undergoing EMT; hence, this appears to be a bona fide feature of the A2 phenotype.