|Home | About | Journals | Submit | Contact Us | Français|
Differentiation methods for human induced pluripotent stem cells (hiPSCs) typically yield progeny from multiple tissue lineages, limiting their utility for drug testing and autologous cell transplantation. In particular, early retina and forebrain derivatives often intermingle in pluripotent stem cell cultures, owing to their shared ancestry and tightly coupled development. Here, we demonstrate that three-dimensional populations of retinal progenitor cells (RPCs) can be isolated from early forebrain populations in both human embryonic stem cell (hESC) and hiPSC cultures, providing a valuable tool for developmental, functional, and translational studies. Using our established protocol, we identified a transient population of optic vesicle-like (OV) structures that arose during a time period appropriate for normal human retinogenesis. These structures were independently cultured and analyzed to confirm their multipotent RPC status and capacity to produce physiologically responsive retinal cell types, including photoreceptors and retinal pigment epithelium (RPE). We then applied this method to hiPSCs derived from a patient with gyrate atrophy, a retinal degenerative disease affecting the RPE. RPE generated from these hiPSCs exhibited a disease-specific functional defect that could be corrected either by pharmacological means or following targeted gene repair. The production of OV-like populations from human pluripotent stem cells should facilitate the study of human retinal development and disease and advance the use of hiPSCs in personalized medicine.
Human pluripotent stem cells, which include both human embryonic stem cells (hESCs) and human induced pluripotent stem cells (hiPSCs), hold the potential to differentiate into any cell type. In doing so, they can serve as comprehensive model systems of human cell genesis, particularly at early developmental stages that would otherwise be inaccessible to investigation1,2. In addition, patient-derived hiPSC lines have a unique capacity to model human disease3–6, although the scope of disorders amenable to this form of study is limited7–9. Major considerations when creating hiPSC disease models include the capacity to efficiently generate, identify and isolate relevant cell populations, as well as recapitulate and assay critical aspects of the disease mechanism.
Retinal cell types are particularly well-suited for the investigation of cell development and dysfunction using pluripotent stem cell technology. The vertebrate retina harbors a modest repertoire of major cell classes sequentially produced via a conserved series of events10–16. Furthermore, the effects of inherited and acquired retinal degenerative diseases (RDD) are often limited initially to a specific cell class, which simplifies the study of cellular mechanisms that incite RDD and the evaluation of potential therapies.
Previous studies have demonstrated the ability of human pluripotent stem cells to differentiate along the retinal lineage with varying efficiencies17–20, with one protocol achieving a near uniform retinal cell fate using the WA01 hESC line18. However, pluripotent stem cell-derived retinal cells, particularly those from hiPSCs, are most often found in mixed populations that include some non-retinal or unidentified cell types17,20–22. Further complicating matters is the fact that several markers used for retinal cell identification (e.g., calretinin, PKCα, Tuj1) also label cells found in other regions of the CNS. As such, a means to isolate developmentally synchronized populations of multipotent retinal progenitor cells (RPCs) across multiple hESC and hiPSC lines would be desirable. The RPCs and their definitive retinal progeny could then be used to study mechanisms of human retinal development and disease, examine retinal cell function, and devise and test RDD treatments.
We recently described a method to differentiate human pluripotent stem cells to RPCs, retinal pigment epithelium (RPE), and photoreceptor-like cells in a manner that mimicked normal human retinogenesis17. However, a means to separate and track the fate of the RPCs in live culture was not available. In the current study, we used transient morphological features to isolate structures with characteristics reminiscent of the optic vesicle (OV). Using these OV-like structures, it was possible to study principles of early human retinal development, monitor the sequence and timing of neuroretinal cell genesis, and optimize RPC and RPE production efficiencies in recalcitrant hiPSC lines. We then utilized our system to evaluate pharmacologic and genetic strategies to correct a functional defect in hiPSC-RPE cultures derived from a patient with an inherited RDD.
hESCs (WA09 and WA01) and hiPSCs (IMR90-423, 6-9-12T24, iPS-1225, iPS-12.425, and three patient-specific lentivirus-derived hiPSC lines) were maintained in ESC medium (see Supporting Information Methods for media descriptions) using an established method17,23. Differentiation toward anterior neuroectodermal and retinal fates was accomplished using a modification of a previously published protocol17. Briefly, human pluripotent stem cell colonies were cultured as suspended aggregates for 4 days in EB medium, whereupon they were switched to Neural Induction Medium. For some experiments, cultures were treated from day 2–4 with 100 ng/ml Noggin (or Dorsomorphin) and DKK1 (or XAV939). After a total of 6–7 days, aggregates were plated onto laminin-coated substrates to permit formation of neural clusters. On day 16, the loosely adherent central portions of the neural clusters were mechanically lifted and grown as cellular aggregates in Retinal Differentiation Medium (RDM). Between day 20–25, a subset of these aggregates adopted a vesicle-like appearance, which was manually separated and differentiated in RDM. Nonvesicular aggregates, or spheres, were likewise separated, pooled, and differentiated in RDM.
Generation of RPE from vesicle-like cultures was enhanced by adding Activin A (100 ng/ml) to RDM from day 20–40 of differentiation. RPE from manually isolated pigmented OV-like structures was expanded by plating them onto laminin-coated substrates in RDM + 20 ng/ml FGF2/EGF and 5 μg/ml heparin for up to one week as described for prenatal RPE cultures26. Large patches of RPE-like cells could then be manually isolated, dissociated, and re-plated in the presence of mitogens for at least one additional passage.
RNA was isolated using the RNAeasy (Qiagen) or Picopure RNA isolation kit (Arcturus), reverse transcribed, and amplified by PCR (30 cycles) or quantitative PCR (qPCR) (40 cycles) as described17. Primer sets are listed in Supporting Information Table S2. For a description of the microarray analysis method, see Supporting Information Methods.
Gyrate atrophy fibroblasts were obtained from Coriell Institute for Medical Research (GM06330) and propagated in DMEM with 10% FBS. In addition, fibroblasts from two individuals without known genetic diseases were cultured from skin samples, as approved by the University of Wisconsin-Madison IRB. Reprogramming was performed as previously described23 using lentiviral vectors expressing OCT4, SOX2, NANOG, LIN28, c-MYC, and KLF4.
OV-like structures were differentiated in suspension culture for a total of 100 days and plated onto poly-D-ornithine- and laminin-coated coverslips for an additional 2 days to allow for adherence and peripheral cell spread/migration. Ionic currents were measured using conventional whole cell recordings performed on a fixed stage of a Nikon FN-1 microscope. Cells were superfused continuously at ~2 ml/min in standard bath solution (see Supporting Information Methods). Recordings were acquired using patch pipettes (3–5 MΩ) filled with pipette solution (see Supporting Information Methods), an Axopatch 200B amplifier, Digidata 1440 interface and pCLAMP 10 software (Molecular Devices). Stored data was analyzed using Clampfit-10 and plotted using Excel and Origin-8. Current responses were generated using either a voltage ramp (−160 to +40 mV) or step (−160 to +40 mV in 10 mV steps) protocol from a holding potential of −70 mV. A junction potential of 11 mV was calculated using pCLAMP software and all records were compensated post hoc.
For Ca2+ imaging, cells were loaded with 5 μM Fura-2 AM (Invitrogen) for 60 min at 37°C. After washing for 30 min, coverslips were placed at the bottom of the recording chamber. 1 mM probenecid was added to prevent dye extrusion, and cells were sequentially exposed to 340 and 380 nm excitation wavelengths while registering fluorescence emission at 510 nm. Images were collected every 10 s with 2×2 pixel binning using a cooled CCD camera (CoolSNAP HQ, Photometrics) and a 60X water immersion objective. NIS Elements was used for image acquisition and offline analysis to determine changes in [Ca2+]i using the equation listed in Supporting Information Methods. Results were expressed as mean ± SEM.
The OV is derived from the primitive anterior neuroepithelium (PAN) and shares numerous features with the early forebrain28–30. We previously demonstrated the ability to differentiate human pluripotent stem cells into PAN, which in turn gave rise to free-floating cell aggregates containing either RPCs or early forebrain cells17. In the present study, we looked for features that could be used to separate the RPC and early forebrain populations in live culture without the need for genetic manipulation.
Using our protocol17, densely packed, neural rosette-containing cell colonies were isolated intact from WA09 hESCs after 16 days of differentiation and cultured in suspension. Starting at 20 days, two populations of cell aggregates could be distinguished via light microscopy (Fig. 1A). A minority population (20.4 ± 4.3%) was phase-bright and appeared vesicle-like due to the presence of a compact outer mantle of cells. The remaining population was spherical and had a more uniform cell distribution, although internal rosettes were commonly observed. Some cell aggregates possessed elements of both populations; however, these hybrid structures were excluded from this study.
Cell aggregates with an unequivocal vesicle-like or nonvesicular configuration were separated between day 20–25 and cultured independently (Fig. 1B,C). Because this period of differentiation corresponded developmentally to the formation of OVs in humans, we asked whether CHX10, a marker of RPCs in the distal OV, was differentially expressed between the two populations. Prior to separation, CHX10 was found in a subset of cell aggregates (Fig. 1D). Afterward, CHX10 immunoreactivity was abundant (90.88 ± 4.23% of total cells) in the vesicle-like structures (Fig. 1E) and nearly absent in the nonvesicular spheres (Fig. 1F). By contrast, ISLET-1, a homeodomain protein involved in projection neuron differentiation and early forebrain development, was found only in the nonvesicular population at this stage (Fig. 1G–I).
Given the disparate expression of CHX10 and ISLET-1 between the vesicle-like structures and nonvesicular spheres, PCR was performed to investigate gene expression patterns of other transcription factors involved in early retinal and/or forebrain development (Fig. 1J). After 20 days, the vesicle-like structures selectively expressed multiple retinal transcription factor genes appropriate for the OV stage of retinogenesis, whereas the nonvesicular spheres expressed transcription factors indicative of the embryonic forebrain. Factors involved in the early development of both the retina and forebrain were likewise present in both culture populations.
The progenitor state of the CHX10+ vesicle-like structures was further queried using the cell proliferation marker Ki67. ICC analysis of sections revealed coexpression of CHX10 and Ki67 and confirmed the compact, radial arrangement of nuclei within these structures (Fig. 1K). With further maturation, they often lost their distinctive morphology and developed internal rosettes, rendering them indistinguishable from the nonvesicular spheres (Fig. 1L). However, most cells remained CHX10+ and/or Ki67+ until approximately 50 days of differentiation (Fig. 1M), after which time the expression of these markers progressively declined in lieu of more mature retinal markers (see Fig. 2).
We next performed comparative gene microarray analysis to further highlight differences in transcription factor gene expression between day 20 vesicle-like structures and nonvesicular spheres (Supporting Information Fig. S1). Many transcription factor genes associated with retinal development were present at higher levels in the vesicle-like structures, including SIX6, RAX, MITF, TBX2, TBX5, and VAX2. Nonvesicular spheres, on the other hand, expressed higher levels of transcription factors implicated in early forebrain or general neural development, including DLX1, DLX2, SOX1, and ISLET-1. Taken together, results from the comparative ICC, PCR, and microarray analyses indicated that the vesicle-like structures and nonvesicular spheres harbored RPC and early forebrain populations, respectively. Furthermore, the gene and protein expression profiles of the RPCs reflected a differentiation state akin to the OV stage of retinal development. Therefore, the vesicle-like structures and nonvesicular spheres were re-designated OV-like structures and early forebrain (EFB) spheres, respectively.
Many markers used to distinguish non-photoreceptor cell types within the neuroretina are found elsewhere in the CNS, compromising efforts to study retinogenesis in mixed lineage pluripotent stem cell cultures. To circumvent this issue, we performed longitudinal differentiation analyses on isolated populations of OV-like structures and EFB spheres. Because OV-like structures contained a nearly exclusive population of RPCs, differentiated progeny from these cultures could be assigned to the retinal lineage with a high degree of certainty.
Cultures of WA09 hESC OV-like structures were sampled every ten days from day 20–120 of differentiation and analyzed by qPCR (Fig. 2A). BRN3 and CRX, neuroretinal markers of retinal ganglion cells (RGCs) or early post-mitotic photoreceptors, respectively, were among the first genes to be expressed. The CALRETININ gene (amacrine cells, horizontal cells, and some RGCs) was expressed at a nearly constant level after day 40. Later-expressed genes included NRL (post-mitotic rod precursors) and PKC-α (bipolar cells). The presence of multiple neuroretinal cell types was confirmed at day 80 by ICC using 1° antibodies against BRN3 and βIII-TUBULIN (Fig. 2B), CALRETININ (Fig. 2C), PKC-α and CHX10 (Fig. 2D), CRX and RECOVERIN (Fig. 2E–G), and NRL (Fig. 2H). Cells expressing photoreceptor markers often displayed a distinct morphology, bearing one thin process opposite a thicker, tubular or conical process (Fig. 2F,G). ICC quantification revealed that most progeny of OV-like structures produced by day 80 expressed photoreceptor or RGC markers (CRX: 55.9 ± 6.6%; BRN3: 14.6 ± 3.2% of all cells). Immunostained sections of OV-like structures showed segregation of retinal cell populations, with most BRN3+ cells concentrated near the periphery, whereas RECOVERIN+ cells tended to congregate internally in clusters or linear configurations (Supporting Information Fig. S2).
EFB spheres maintained an anterior neural identity upon further differentiation, generating a variety of phenotypes found within the forebrain. At 20 days, EFB spheres expressed PAX6 (Fig. 3A), SOX1 (Fig. 3B), OTX2 (Fig. 3C), and βIII TUBULIN (Fig. 3A–C). By day 70, EFB spheres yielded more mature cell phenotypes, including GABAergic neurons (Fig. 3D), tyrosine hydroxylase (TH)+ dopaminergic neurons (Fig. 3E), and GFAP+ astrocytes (Fig. 3F). No retina-specific markers were present in EFB spheres at any time point tested (up to day 70). The anterior neural character of day 20 EFB spheres was further confirmed by PCR, which revealed expression of forebrain, but not midbrain (EN-1) or hindbrain (HOXB4) genes (Fig. 3G). The expression of some of these genes was maintained at day 70 (e.g., FOXG1), while others were down- (e.g., PAX6, OTX2) or up- (e.g., MAP2, TH) regulated.
Examination of electrophysiological properties offers a powerful means of evaluating both cell identity and functional capacity. WA09 hESC OV-like structures were differentiated for 100 days and plated on coverslips, whereupon cells with photoreceptor-like morphology were recorded using standard patch clamp techniques. To verify identity, recorded cells were loaded with sulphorhodamine (Fig. 4A) and later immunostained for RECOVERIN (Fig. 4B). PCR analysis of these cultures confirmed expression of genes associated with phototransduction, including the cyclic nucleotide-gated channel subunits CNGA1, A3, B1, and B3, the guanylyl cyclase gene RETGC, the cGMP phosphodiesterase gene PDE6B, and ARRESTIN (Fig. 4C). RECOVERIN+ cells produced an outward positive current that was activated at depolarizing voltages between −50 and +40 mV from a holding potential of −70 mV (Fig. 4D). Upon achieving whole-cell configuration, these cells registered a resting membrane potential of −44 ± 4 mV and a current at +40 mV of 27 ± 8 pA/pF (n = 15), compared to −29 ± 2 mV and 9 ± 1 pA/pF for control, non-photoreceptor cells (n=3) (Fig. 4D). The current-voltage (I-V) plot revealed a large outward current with a linear I-V relationship between −10 to +40 mV, but no inward current. The voltage-dependent outward current was suppressed with 15 mM tetraethylammonium (TEA), and measurements at both +20 and +40 mV showed that the TEA-sensitive component possessed fast activation kinetics without deactivation during the 500 ms voltage pulse (Fig. 4E). I-V curves confirmed the selective reduction of outward current by external TEA from an average of 468 ± 139 pA to 89 ± 25 pA (measured at +40 mV; n=5 cells) (Fig. 4F). Given the low [Ca2+] pipette solution used for these experiments and the TEA sensitivity of the current, we concluded that delayed rectifier potassium channels were responsible for the observed voltage-dependent outward current, consistent with photoreceptor electrophysiology31–35.
In the dark, photoreceptors are maintained in a depolarized state via influx of Ca2+ and Na+ through nonselective cGMP-gated plasma membrane cation channels36–38. Light-stimulated hydrolysis of cGMP results in closure of these channels and membrane hyperpolarization39. To further test the functional identity of the photoreceptor-like cells, we exposed them to membrane-permeable 8-br-cGMP, which resulted in an instantaneous increase in inward current amplitude of >4-fold (measured at −70 mV holding potential) (Fig. 4G), consistent with the known kinetics of cyclic nucleotide-gated ion channels40. A similar sustained increase in current amplitude was observed at depolarized membrane potentials (Fig. 4H,I). Overall, exposure to 8-br-cGMP converted the outward rectifying I-V plot to a linear I-V plot that crossed close to 0 mV (Fig. 4I), indicating a switch to nonselective ion conductance40. Furthermore, 8-br-cGMP treatment resulted in membrane depolarization from −44 ± 4 mV to −7.7 ± 1.8 mV, similar to the difference between the light and dark resting membrane potentials of photoreceptors39. These features are found solely in cells that undergo phototransduction31,39, and provide further evidence that RECOVERIN+ cells generated from these cultures possess an electrophysiological signature highly reminiscent of photoreceptors.
RPE was rarely observed when OV spheres were cultured in isolation, regardless of the duration of differentiation. Previous reports demonstrated that the TGFβsuperfamily member Activin A can promote an RPE fate at the expense of neuroretina41,42. To test whether a similar effect could occur in our cultures, 100 mg/ml Activin A was added to cultures of OV-like structures from day 20–40. Activin A-treated cultures produced subsets of structures with deep pigmentation beginning between day 40–60 (Fig. 5A,B). Pigmented OV-like structures were manually separated from non-pigmented OV-like structures (Fig. 5C) and adhered to substrate in the presence of mitogens to promote cell proliferation. Under these conditions, cells proliferated and formed monolayers (Fig. 5D). Removal of mitogens resulted in reestablishment of cellular pigmentation, polygonal morphology (Fig. 5E), and gene and protein expression patterns characteristic of RPE (Fig. 5F; Supporting Information Fig. S3). qPCR analysis (Fig. 5G,H) revealed a reciprocal influence of Activin A on the expression of MITF (8.3 ± 2.5-fold increase) and CHX10 (5.3 ± 1.4-fold decrease), transcription factors associated with the development of RPE and neuroretina, respectively. Activin A-treated cultures also expressed RPE genes such as RPE65 and BEST1 at higher levels (4.9 ± 1.8- and 19.1 ± 5.4-fold, respectively) than untreated OV-like structures (Fig. 5G), and lower levels of genes associated with neuroretina, including CRX and BRN3B (2.4 ± 0.6- and 1.9 ± 0.4-fold, respectively) (Fig. 5H).
To assay the functional potential of hESC-RPE cells, we chose to examine their response to ATP, a molecule postulated to govern light-induced activation of purinergic signaling pathways, leading to intracellular Ca2+ mobilization and directional fluid transport across the RPE43. hESC-RPE cells exhibited an increase in [Ca2+]i (58 ± 4.5 nM) immediately after exposure to 100 μM ATP, followed by a decline to steady state levels (Fig. 5I–K). These responses were comparable to those of prenatal human RPE cultures (Supporting Information Fig. S3). Therefore, similar to other human pluripotent stem cell differentiation protocols42,44–46, RPE derived with our method displayed important physiological properties that can be exploited to study cell function and test therapeutics.
Lineage-specific differentiation efficiencies vary across hiPSC lines, which restricts their utility7,17,45,47. A better understanding of the factors governing hiPSC differentiation potential may help optimize methods for the targeted production of neural and retinal cell types. We screened four hiPSC and two hESC lines via qPCR for expression of selected genes involved in early neuro- and retinogenesis. Significant variability was found in the expression levels of anterior neural/eye field genes across these lines at day 10 (Fig. 6A,B; Supporting Information Fig. S4). Lines displaying the lowest levels of these genes mainly produced flat, epithelioid colonies (Fig. 6C) that lacked expression of anterior neural/eye field markers (Fig. 6D).
We next investigated very early gene expression levels of DKK1 and NOGGIN, two pro-anterior neural/eye field factors that antagonize the canonical Wnt and BMP signaling pathways, respectively20,22. Prolonged addition of one or both of these factors is common to human pluripotent stem cell neural and retinal differentiation protocols20,22; however, it is unclear whether endogenous DKK1 and NOGGIN expression levels predict competency to produce anterior neural progeny. We found that hiPSC lines expressing the highest levels of DKK1 and, to a lesser extent, NOGGIN at day 2 (Fig. 6E,F) also expressed the highest relative levels of anterior neural/eye field genes at day 10 (Fig. 6A,B; Supporting Information Fig. S4) and CHX10+ OV-like structures at day 20 (Fig. 6G,H). OV-like structures isolated from hiPSCs were indistinguishable from those derived from hESCs and, upon further differentiation, generated neuroretinal cell types (Supporting Information Fig. S5). RPE produced from hiPSC cultures was also similar to human prenatal and hESC-derived RPE based on appearance, gene and protein expression, and physiological response to ATP (Supporting Information Fig. S3).
Given the apparent correlation between early endogenous DKK1 and NOGGIN expression and the ensuing acquisition of an anterior neural/eye field fate, we hypothesized that DKK1 and NOGGIN (or their small molecule equivalents) need only be added during a short developmental time window to augment neural and retinal specification in recalcitrant hiPSC lines. To test this theory, DKK1 (or XAV-939) and NOGGIN (or Dorsomorphin) were added to our lowest anterior neural/eye field gene-expressing hiPSC line (Lenti iPSC #2) from day 2–4 and analyzed by qPCR six days later. Treated cultures expressed 84.8 ± 1.9- and 156.3 ± 59.5-fold higher levels of PAX6 and RAX, respectively, compared to untreated cultures (Fig. 6I). Morphologically, treated hiPSCs produced mounded colonies of neuroepithelial cells at day 10 (Fig. 6J) that expressed anterior neural/eye field transcription factors (Fig. 6K). The neuralization of treated hiPSCs was further confirmed through FACS analysis, which revealed a 2.7-fold increase in the percentage of PAX6+ cells at day 10 (72.6 vs. 27.3%; Fig. 6L) and a 6.7-fold increase in the percentage of CHX10+ cells at day 20 (34.1 vs. 5.1%; Fig. 6M) compared to untreated cultures.
A reliable means of isolating RPCs from patient-specific iPSCs would facilitate the study and treatment of retinal diseases4. Gyrate atrophy (GA) is a rare autosomal recessive RDD that primarily affects RPE, causing secondary photoreceptor loss and blindness27,48–50. GA is caused by a wide range of defects in the gene encoding ornithine-δ-aminotransferase (OAT), a vitamin B6-dependent enzyme that catalyzes the interconversion of L-ornithine and Δ1-pyrroline-5-carboxylate48. hiPSCs were derived from fibroblasts of a GA patient bearing the A226V OAT mutation (Fig. 7A,B) and screened for expression of pluripotency genes, ability to form teratomas, retention of the OAT gene mutation, and maintenance of a normal karyotype (Supporting Information Fig. S6).
Using our protocol, (A226V)OAT hiPSCs formed neuroepithelial structures (Fig. 7C) and expressed anterior neural/eye field markers at day 10 (Fig. 7D). OV-like structures isolated from these cultures generated neuroretinal progeny (Fig. 7E), but rarely became pigmented in the absence of Activin A. OAT activity in RPE derived from (A226V)OAT hiPSCs (Fig. 7F) was then measured using a colorimetric assay27. In the presence of 50 μM vitamin B6, (A226V)OAT hiPSC-RPE had very low OAT activity (5.9 ± 3.0 μmol/hr/106 cells; n=3) (Fig. 7G) compared to control RPE cultures derived from IMR90-4 hiPSCs, WA09 hESCs, and prenatal human eyes (68.9 ± 9.9, 142.7 ± 19.7, and 91.2 ± 10.0 μmol/hr/106 cells, respectively).
Because the particular mutation in this GA patient is postulated to affect vitamin B6 binding to the enzyme51, we next investigated whether elevated levels of vitamin B6 could restore OAT activity. Increasing vitamin B6 to 200 μM was ineffective; however, 600 μM vitamin B6 greatly enhanced enzymatic activity (24.5 ± 9.8-fold; n=3) in (A226V)OAT hiPSC-RPE (Fig. 7H). In contrast, 600 μM vitamin B6 only increased OAT activity in control RPE cultures by 1.7 ± 0.1-fold (n=3) and in (A226V)OAT fibroblasts by 8.1 ± 3.2-fold (n=5). Higher levels of vitamin B6 (1200 μM) failed to increase OAT activity further.
Lastly, we applied our differentiation protocol to a gyrate atrophy hiPSC line (iPS-12.4) that previously underwent repair of the OAT gene mutation via bacterial artificial chromosome (BAC)-mediated homologous recombination25. Contrary to RPE derived from the (A226V)OAT parent line (iPS-12), the gene-corrected hiPSC-RPE displayed normal levels of OAT activity (130.0 ± 12.4 μmol/hr/106 cells; n=3) (Fig. 7I). Therefore, OV-like structures provide a practical and versatile intermediary in the production of retinal cell types from hiPSCs.
The ability to isolate tissue-specific progenitor populations from live, differentiating human pluripotent stem cells enhances their scientific and clinical utility. Beyond supplying a potentially expandable and relatively uniform cell source for transplantation, such populations facilitate in vitro studies of human cellular development and disease.
The development of the vertebrate retina can be traced morphologically, beginning with the evagination of OVs from the lateral aspects of the primitive forebrain. At this stage, chick OVs can be dissected from surrounding tissues and cultured short-term as explants, during which time they maintain their characteristic vesicular structure41. However, a similar in vitro hallmark of early retinal populations had not been described in human pluripotent stem cell cultures prior to this report. OV-like structures isolated from hESCs and hiPSCs were anterior neuroectodermal in origin, expressed early retinal markers, and gave rise to mature retinal cell types, confirming their status as RPCs. In addition, the order of appearance of retinal cell types from OV-like structures approximated the conserved cell birth order found during vertebrate retinogenesis10,12.
While OV-like structures resemble true OVs in some respects, the similarities between them are limited. Unlike true OVs, hESC- and hiPSC-derived OV-like structures did not invaginate to form bilayered cups using the minimal culture conditions we employed. Instead, they “filled in” and became relatively unorganized, presumably as a result of continued RPC proliferation in the absence of appropriate signaling cues. In addition, during normal vertebrate OV development, production of Chx10+ neuroretinal progenitors is restricted to the distal OV, whereas the proximal OV remains Mitf+ and yields RPE29,30. This segregation of the neuroretinal and RPE domains is governed in part by reciprocal effects of FGFs and TGFβ superfamily members produced by surface ectoderm and extraocular mesenchyme, respectively41,52. Similarly, canonical Wnt signaling has been shown to promote RPE generation while antagonizing formation of neuroretina53. The lack of RPE production in isolated OV-like structures, which are cultured in defined medium without these factors, suggests an intrinsic bias toward a neuroretinal fate. In support of this hypothesis, we found that hESC-derived sphere cultures endogenously expressed FGFs along with inhibitors of TGFβ superfamily and Wnt signaling17. In the present study, we further showed that this neuroretinal bias could be partially overcome via the addition of the TGFβ superfamily member Activin A. An analogous situation has been described in the chick model, where OV explants grown in isolation failed to generate RPE, but did so when exposed to Activin A41.
Beyond adopting distinct morphologies and expression profiles, we also demonstrated that retinal cells derived from human pluripotent stem cells possessed important functional properties. Patch clamp recordings of RECOVERIN+ cells revealed a current-voltage relationship consistent with mammalian photoreceptors31–35, as well as robust depolarization in response to administration of cGMP analogs. The latter finding is particularly intriguing, since cGMP is a critical second messenger of the phototransduction cascade38, responsible for modulating the circulating ‘dark current’39. While RGCs can also depolarize after cGMP stimulation54, all cGMP-responsive cells tested in our cultures were RECOVERIN+ and lacked long neurite projections characteristic of RGCs. Therefore, the identity of the recorded cells was most consistent with photoreceptors. To document functionality of hESC- and hiPSC-RPE cells in a novel manner, we examined their response to ATP, an extracellular signaling molecule linked to the generation of IP3 and release of intracellular calcium55. Through this mechanism, ATP is proposed to regulate ion and fluid flux across the RPE43, among other critical functions. Exposure of hESC- and hiPSC-RPE to ATP resulted in rapid and reversible increases in intracellular calcium, analogous to responses elicited from prenatal human RPE cultures.
While the isolation of OV-like structures from differentiating hESCs provides a dynamic tool for the study of human retinal development, it may prove more useful for hiPSC research. Multiple reports have detailed variability in fate potential between hiPSC lines7,17,45,47, some of which displayed little capacity to yield retinal cell types. In the present study, we found that OV-like structures were generated by every differentiating hiPSC line tested, although sometimes in very low abundance. Even so, their distinctive appearance allowed them to be isolated and cultured independently. It was also noted that retinal differentiation efficiency across hiPSC lines correlated with early endogenous expression of factors (DKK1 and NOGGIN) known to influence anterior neural fate specification. We have since used this information to quickly and inexpensively screen new hiPSC lines for competency to produce OV-like structures.
As methods of hiPSC production and differentiation continue to improve, so will the potential to apply this technology to model and treat human diseases. Inherited RDDs appear particularly amenable to hiPSC modeling, since many exhibit gene-specific, cell autonomous features that can be tested in vitro. In addition, broad sequelae common to multiple inherited RDDs, such as loss of rods, can be monitored in culture and used to assess drug efficacy5. For the present study, we chose to derive hiPSCs from a patient with GA, an RPE-based RDD for which a simple assay exists to measure activity of the affected gene product, OAT27. This enzyme is expressed in multiple tissues, but has particularly high activity in RPE27, which may explain why clinical manifestations of GA are largely restricted to the retina. We confirmed that (A226V)OAT hiPSC-RPE had very low OAT activity, and further showed that these cells were exceptionally responsive to treatment with vitamin B6, a pharmacological agent known to reduce serum ornithine levels and improve fibroblast OAT activity in a small cohort of GA patients. However, direct tests of vitamin B6 efficacy in RPE derived from a living GA patient have heretofore been unfeasible. We found that 600 μM vitamin B6, but not 200 μM, completely restored OAT activity in (A226V)OAT hiPSC-RPE, while higher concentrations showed no additional benefit. These findings are potentially important since excessive vitamin B6 supplementation can cause painful and ultimately irreversible neurological side effects56.
Perhaps the most imminent clinical role of hiPSC technology is in pharmacologic testing of rare, genetically heterogeneous diseases that 1) display variations in drug efficacy between individuals and 2) affect tissues that cannot be routinely or safely biopsied. Even when surrogate tests of drug response are available, patients may be better served by direct testing of cell types targeted by the disease. For example, information from the repository where we obtained the GA fibroblasts indicated that the donor was vitamin B6 unresponsive. However, other patients with the A226V OAT mutation51, as well as the donor’s own hiPSC-RPE, were shown to respond to such treatment. It is therefore possible that the patient would have had a therapeutic response to vitamin B6 at the RPE level, but because efficacy was determined by less sensitive means, the treatment was erroneously deemed unbeneficial.
A more distant application of hiPSCs may be as a source of autologous donor cells for replacement therapies. To serve in this capacity, repair of disease-causing gene defects will likely be necessary. Recently, Howden et al. corrected the OAT mutation in our patient using BAC-mediated homologous recombination25 and showed that this process did not introduce new mutations into the hiPSC genome. Here, we further demonstrated that OAT activity was fully restored in the gene-corrected hiPSC-RPE. To our knowledge, this is the first report of functional correction of disease-specific hiPSCs using both pharmacologic and genetic approaches.
We demonstrated that vesicle-like structures corresponding to the OV stage of retinal development could be manually isolated from hESC and hiPSC lines. The production of highly enriched populations of physiologically active retinal cell types from hiPSCs provides a platform to investigate patient-specific drug effects, gene repair strategies, and disease mechanisms in vitro. However, caution must be exercised when extrapolating data obtained from cell culture systems to whole organisms. As such, hiPSC technology is envisioned to complement, not supplant, existing laboratory models of disease.
While our manuscript was under review, Eiraku et al.57 published a report demonstrating that mouse ESCs could form OV structures, which subsequently self-organized into bilayered optic cups that produced both RPE and laminated neuroretinal tissue. Interestingly, when the mouse hESC-derived OVs were separated from attached non-retinal neuroectoderm and cultured in isolation, they failed to form optic cups and became biased to a Chx10+ neuroretinal fate, similar to our findings with isolated human ESC- and iPSC-derived OV-like structures. In light of our results and those of Eiraku et al.57, we suspect that OV-like structures from human ESCs and iPSCs are also capable of higher order structure and organization under certain culture conditions.
This work was supported by the Foundation Fighting Blindness (Wynn-Gund Translational Research Acceleration Program Award and Walsh Retinal Stem Cell Consortium - DMG), the National Institutes of Health (R01EY21218 to DMG, and P30HD03352 to the UW-Madison Waisman Center), Lincy Foundation (DMG), Retina Research Foundation (RRF) Gamewell Professorship (DMG), E. Matilda Ziegler Foundation for the Blind (DMG), Rebecca Meyer Brown Professorship (BP) and UW-ICTR NIH grant 1UL1RR025011 (BP). DMG is a recipient of a Research to Prevent Blindness McCormick Scholar Award and the UW Eye Research Institute/RRF Murfee Chair. SEH is supported by a NHMRC Overseas Biomedical Fellowship. The content is solely the responsibility of the authors and does not necessarily represent the official views of the NEI or NIH.
COMPETING INTERESTS STATEMENT
The authors declare competing financial interests: J.A.T. is a founder, stockowner, consultant and board member of Cellular Dynamics International (CDI). He also serves as scientific advisor to and has financial interests in Tactics II Stem Cell Ventures. CDI currently has no products related to the retina. The other authors have no financial interests to disclose.
Author ContributionsJason S. Meyer: Conception and design, Collection and/or assembly of data, Data analysis and interpretation, Manuscript writing, Final approval of manuscrip
Sara E. Howden: Conception and design, Collection and/or assembly of data, Data analysis and interpretation, Manuscript writing, Final approval of manuscrip
Kyle A. Wallace: Collection and/or assembly of data, Data analysis and interpretation
Amelia D. Verhoeven: Collection and/or assembly of dat
Lynda S. Wright: Data analysis and interpretation, Manuscript writin
Elizabeth E. Capowski: Conception and design, Data analysis and interpretatio
Isabel Pinilla: Collection and/or assembly of dat
Jessica M. Martin: Collection and/or assembly of dat
Shulan Tian: Collection and/or assembly of data, Data analysis and interpretation, Manuscript writin
Ron Stewart: Collection and/or assembly of data, Data analysis and interpretation, Manuscript writing, Final approval of manuscrip
Bikash Pattnaik: Conception and design, Financial support, Collection and/or assembly of data, Data analysis and interpretation, Manuscript writing, Final approval of manuscrip
James Thomson: Conception and design, Financial support, Data analysis and interpretation, Final approval of manuscrip
David M. Gamm: Conception and design, Financial support, Data analysis and interpretation, Manuscript writing, Final approval of manuscript