Highlights
hPSC-derived photoreceptors express markers in a pattern similar to human development
2D/3D differentiation protocol generates sufficient cones for transplantation
hPSC-derived cones incorporate into the adult retina following transplantation
Summary
Transplantation of rod photoreceptors, derived either from neonatal retinae or pluripotent stem cells (PSCs), can restore rod-mediated visual function in murine models of inherited blindness. However, humans depend more upon cone photoreceptors that are required for daylight, color, and high-acuity vision. Indeed, macular retinopathies involving loss of cones are leading causes of blindness. An essential step for developing stem cell-based therapies for maculopathies is the ability to generate transplantable human cones from renewable sources. Here, we report a modified 2D/3D protocol for generating hPSC-derived neural retinal vesicles with well-formed ONL-like structures containing cones and rods bearing inner segments and connecting cilia, nascent outer segments, and presynaptic structures. This differentiation system recapitulates human photoreceptor development, allowing the isolation and transplantation of a pure population of stage-matched cones. Purified human long/medium cones survive and become incorporated within the adult mouse retina, supporting the potential of photoreceptor transplantation for treating retinal degeneration.
Introduction
Retinal degenerations involving the loss of photoreceptors are the leading cause of untreatable blindness in the developed world. Human vision is critically dependent on cone-mediated vision, and diseases affecting these cells, such as age-related macular degeneration and Stargardt disease, are particularly debilitating. Replacement of the lost photoreceptors by transplantation represents one of few options for the reversal of end-stage degeneration (Barber et al., 2013, Singh et al., 2013, Barnea-Cramer et al., 2016, Neves et al., 2016). Most macular diseases involve loss of both the retinal pigment epithelium (RPE) and the photoreceptors that they support. Clinical trials involving the transplantation of human embryonic stem cell (hESC)-derived RPE to treat patients with macular degenerations have shown safety and some limited efficacy (Schwartz et al., 2015), but the ability of human cones to integrate and survive following transplantation has yet to be demonstrated.
There have been numerous pre-clinical studies investigating the transplantation of rod photoreceptor precursors derived both from neonatal and stem cell sources, with many reporting improvements in visual function (Pearson et al., 2012, Singh et al., 2013, Barnea-Cramer et al., 2016, Neves et al., 2016). Until recently, it had been understood that the observed improvements in visual function, at least in those models where some ONL remained, were due to donor cells integrating within the host retinae. Recently, however, we (Pearson et al., 2016), and others (Santos-Ferreira et al., 2016, Singh et al., 2016, Ortin-Martinez et al., 2017, Decembrini et al., 2017), have demonstrated that the majority of rescued cells seen in the host ONL arise as a result of the exchange of cytoplasmic material between donor and host photoreceptors. The cellular mechanisms mediating this material transfer remain to be determined, but it permits the acquisition by the host photoreceptors of proteins that they otherwise lack and, in the case of diseased cells, may render these cells functional. Previous studies have shown rescue of visual function following transplantation of hESC-derived retinal cells (Lamba et al., 2009, Zhu et al., 2017). However, in the light of these new studies (Pearson et al., 2016, Santos-Ferreira et al., 2016), unequivocal proof that the GFP-labeled cells within the host ONL are integrated donor cells is lacking. Here, we sought to establish whether human pluripotent stem cell (hPSC)-derived cones can survive and integrate into murine models of retinal degeneration following transplantation.
A fundamental requirement in the development of cell-based therapies is the establishment of robust protocols that permit the derivation of large numbers of donor cells from a renewable source that faithfully recapitulate the characteristics of the endogenous cell types they are designed to replace. The differentiation of hPSCs toward retinal lineages has evolved considerably in the last few years and several protocols have been reported (Meyer et al., 2009, Nakano et al., 2012, Boucherie et al., 2013, Zhong et al., 2014, Reichman et al., 2014, Mellough et al., 2015). The accuracy with which these protocols replicate normal retinal development can vary and the choice of differentiation system to generate cells for therapy will be of great importance. For example, murine ESCs (mESCs) differentiated using a 2D/adherent culture-based differentiation protocol supported incomplete differentiation of photoreceptors (West et al., 2012), while the same cells could be differentiated using a 3D/suspension culture system that gave rise to large numbers of postnatal-staged rod and cone photoreceptors (Gonzalez-Cordero et al., 2013, Kruczek et al., 2017).
Here, we report the recapitulation of human photoreceptor development from hPSCs using a combined 2D/3D differentiation system and the transplantation and incorporation of purified hPSC-derived cones within the diseased host retina. Together, these support the potential utility of hPSC-derived cells in future therapeutic applications.
Results
Efficient Generation of hPSC-Derived Photoreceptors Using a 2D/3D Differentiation Protocol
We sought to establish a method that would maintain both the differentiation niche and the layered morphology of the retina, as well as recapitulating human photoreceptor development. Our method involves a combination of 2D/3D differentiation, based on the protocols reported by Goureau and Canto-Soler (Reichman et al., 2014, Zhong et al., 2014), and is shown schematically in Figure 1A. In brief, hPSCs were grown to confluence. When 90% confluent (denoted as day 0 of differentiation), fibroblast growth factor (FGF) was removed for 2 days, followed by a neural induction period of up to 7 weeks. By ∼3–4 weeks of differentiation, islands of lightly pigmented RPE cells appeared (Figure 1B, white arrows) and from within these regions optic vesicle-like structures formed, which contained presumptive neuroretinal vesicles (NRVs) (Figure 1B, black arrowheads) bearing retinal neuroepithelium. Between weeks 4 and 7 NRVs were manually dissected, together with small amounts of RPE, and grown in suspension culture in the presence of fetal bovine serum, taurine, and retinoic acid (Figures 1B and S1). The presence of RPE cells facilitated the separation of NRVs under bright-field microscopy (Figures S1A–S1C) from other forebrain-like neuroepithelium, which can also form in these cultures. RPE islands can be purified and expanded to yield functional RPE cells (data not shown). By 10 weeks, 90% (±8%) of H9 ESC-derived NRVs examined continued to display retinal neuroepithelial morphologies, while 10% (±8%) exhibited non-neural epithelia morphologies (Figures 1C–C’’; p < 0.0001, paired t test; N = 10 differentiations) and were excluded from further culture. Variation between cultures was evident, with 52% of differentiations (N = 81) forming NRVs surrounded by RPE. Of these, 23% resulted in good differentiations (classified as >40 NRVs/differentiation; mean = 177 ± 65 vesicles) and 29% resulted in moderate differentiations (between 10 and 40 NRVs per culture; mean = 17 ± 10 vesicles), with respect to NRV formation (summarized in Table S3).
All NRVs, without exception, expressed markers of photoreceptor differentiation (n > 300 NRVs). Immunohistochemistry (IHC) revealed well-formed ONL-like regions with numerous cells immuno-positive for the pan photoreceptor marker RECOVERIN, and the rod-specific transcription factor NRL (Figures 1D–F′). By 17 weeks, 36% (±6%) (n = 20 NRVs; N = 4 differentiations) of the cells within the NRVs were RECOVERIN+, as assessed by flow cytometry (Figure 1H) and 95% (±5%) (n = 30 images; N = 3 differentiations) of RECOVERIN+ cells co-expressed the cone-rod homeobox protein, CRX (Figure 1G). This differentiation protocol also appeared to support the differentiation of other retinal cells types, as demonstrated by IHC for ganglion cells (NEUN+ and RXRγ+), horizontal cells (PROX1+ and CALBINDIN+), amacrine cells (CALRETININ+), bipolar cells (PKC+), and Müller glia cells (CRALBP+) (Figures S1D–S1L).
Time Course of hPSC-Derived Photoreceptor Development In Vitro Reflects that Seen In Vivo
While protocols for differentiating photoreceptors from hPSCs have been reported previously (Lamba et al., 2009, Meyer et al., 2009, Nakano et al., 2012, Zhong et al., 2014, Reichman et al., 2014, Mellough et al., 2015), a detailed time course of their genesis in vitro, compared with normal human development, has yet to be elucidated. We examined the time course of expression of a panel of photoreceptor-specific proteins in our 2D/3D differentiation system (Figure 2). In vivo, fetal week (Fwk) 14 human retina already contains defined nuclear layers (O’Brien et al., 2003). Similarly, by 15 weeks of differentiation, hPSC-derived CRX+ cells, which first appear at week 6, were present in a defined ONL-like layer at the apical edge of the developing neuroepithelium (Figure 2A). A comparable pattern of staining was observed for RECOVERIN, with most RECOVERIN+ cells organized within a recognizable ONL-like structure by 10 weeks (Figure 2B). While well-organized neuroepithelia were typical at this stage, in some instances NRVs became disorganized and formed rosettes. Rods were abundant by 15 weeks, as shown by widespread presence of NRL and RHODOPSIN (Figures 2C and 2E) with 63% ± 8% of cells in the neuroepithelia positive for NRL (n = 30 images; N = 3). In the mammalian retina, the transcription factor NRL interacts with CRX to regulate the expression of Rhodopsin. NRL+ rods were first detected around week 7, while the onset of RHODOPSIN was observed around week 12. Proliferation of the neuroblastic layer (NBL) ceased completely by 15 weeks, with no KI67+ cells observed in the presumptive ONL after week 12 (Figures S1J and S1K). The pattern of expression of photoreceptor markers described here for hPSC-derived NRVs closely follows the timings reported for human photoreceptor development in vivo(Figure 2D) (O’Brien et al., 2003, Hendrickson et al., 2008, Hendrickson et al., 2012; J.C.S., unpublished data). We confirmed substantial expression of RECOVERIN and NRL within the well-formed ONL by Fwk 20 (Figure S2). The comprehensive characterization of photoreceptor differentiation described above was performed on H9 hESC-derived NRVs. To further validate our system, we assessed photoreceptor differentiation using a second ESC line (H1 Wicell; data not shown) and a hiPSC line (IMR90-4 Wicell; Figure S3.) and observed similar patterns of expression.
hPSC-Derived Photoreceptor Precursors Develop Several Key Mature Structures In Vitro
Robust formation of cilia and outer segments (OS) by hPSC-derived photoreceptors is an important requirement for their utility in understanding human retinal development and for disease modeling. Previous studies have reported the formation of rudimentary OS on a small number of hPSC-derived photoreceptors (Mellough et al., 2015, Zhong et al., 2014, Parfitt et al., 2016). We sought to determine whether our culture conditions would better support the differentiation of structures typically associated with mid-late stages of photoreceptor maturation: bright-field images of NRVs consistently revealed a semi-transparent neuroepithelium that appeared laminated from 17 weeks of differentiation (Figure 3A, white bars) and, similar to a recent report (Wahlin et al., 2017), pronounced brush-like protrusions developed by 28 weeks from the apical border of the neuroepithelium (Figures 3B and B′). By 20 weeks, the rod-specific phototransduction protein α-transducin (GNAT1) was expressed throughout hPSC-derived rods, revealing basal processes extending toward a presumptive outer plexiform layer (OPL) and elongated inner segments (IS) (Figure 3C). Mitochondria, typically enriched in the IS, were highly concentrated just apical to the outer edge of the presumptive ONL (Figure 3D). ZO-1, a marker of adherens junctions, was present at the apical margin of the ONL, basal to the photoreceptor ISs (Figure 3E), indicating formation of a correctly positioned outer limiting membrane (OLM) (Figure 3L). The connecting cilia (CC) is responsible for transporting proteins between the IS and OS of photoreceptors. In our hPSC-derived NRVs, ciliary rootlet protein, ROOTLETIN, was observed correctly located, basal to the OS protein, RETGC (Figures 3F and F′), and levels increased over time (Figures 3F and F′). Another ciliary protein, RPGR, presented in a clear, punctuated pattern between the presumptive IS- and OS-like regions (Figure 3G), while PERIPHERIN-2 (PRPH2), a protein essential for OS formation, was detected at week 17 and co-localized with RHODOPSIN (Figure 3H). Photoreceptor maturation continued in vitro: RHODOPSIN was initially distributed throughout the rod cell bodies, but became restricted to the PRPH2+ OS-like regions, which were apical to the mitochondria-rich IS (Figures 3I, 3J, and S4A–S4A″). ABCA4, another OS protein, was also localized in a pattern very like that observed for PRPH2 (Figure 3K).
We have previously reported that, despite the presence of phototransduction proteins in mESC-derived embryoid bodies in regions resembling OSs, we were unable to detect OSs by electron microscopy (Gonzalez-Cordero et al., 2013). To verify the formation of IS, cilium, and OS regions, we performed ultrastructural examination of hPSC-derived NRVs (Figures 3L–3N). At 17 weeks, we could readily identify the OLM, numerous ISs packed with mitochondria (Figures 3L and 3M), and CC-like structures with characteristic 9 + 0 microtubular basal body composition (Figure 3L, insert). Importantly, protruding from the CC were OS-like structures containing defined, albeit disorganized, membranous discs (Figures 3M and S4B–S4G). While most of these structures were relatively short and stubby up to week 17 (Figure 3L; N = 2 NRVs), OS-like structures were more pronounced at later stages (Figure 3M; N = 4 NRVs) and a minority were notably longer and resembled the elongated OSs of rods (Figure 3N).
In the retina, photoreceptors project processes basally toward the OPL, where they form synaptic connections with bipolar and horizontal cell interneurons. We sought to determine whether hPSC-derived photoreceptors exhibited the machinery necessary for synapse formation. Both the presynaptic ribbon marker, RIBEYE, and SYNTAXIN3, which is expressed in both photoreceptor and bipolar synapses were observed in an OPL-like region, basal to the presumptive ONL, where cone pedicles (Figures S5A–S5C) and rod boutons (Figure S3E) terminate. Ultrastructural analysis showed that this OPL-like region comprises numerous synaptic terminals containing vesicles. These vesicles were often seen close to one or more electron-dense bars (Figures S5D–S5G), an arrangement typical of synaptic ribbons. Together with the IHC data, these findings suggest that hPSC-derived photoreceptors can develop key structures necessary to receive and transfer visual information and do so in a time course that closely reflects what is known about normal human photoreceptor development.
Characterization of hPSC-Derived Cone Photoreceptor Development In Vitro
While previous studies reported the generation of hPSC-derived cone-like cells (Zhong et al., 2014, Zhou et al., 2015), a detailed characterization of their development has not been described. Three key transcription factors, ONECUT1, OTX2, and OLIG2, are reported to be involved in early cone specification in the mouse (Hafler et al., 2012, Emerson et al., 2013). ONECUT1 suppresses early rod gene expression and defines competence for cone and horizontal cell production. In our cultures, ONECUT1+ RPCs were distributed throughout the NBL of NRVs at 7 weeks, many becoming restricted to the INL-like region and were most likely PROX1+ horizontal cells (Figures 4A and S1E). In early murine development, ONECUT1 is also expressed in photoreceptor precursors. Accordingly, at 7 weeks a proportion of ONECUT1+ cells were also CRX+ (Figures 4B and B′, arrowheads), whereas ONECUT1 was restricted to cells in the INL that were negative for CRX by 12 weeks (Figures 4C and C′).
In the murine retina, OLIG2 expression commences at embryonic day (E) 12.5, becoming restricted to cells of the INL postnatally (Shibasaki et al., 2007). Consistent with this, in hPSC-derived retina, ONECUT1 and OLIG2 were first detected throughout the NBL at early stages, becoming restricted to the INL at later time points (Figure 4D). In mouse, OTX2 and ONECUT1 co-regulate the early cone precursor-specific transcription factor, TRβ2 (Emerson et al., 2013). OTX2 was also observed at week 7 in the NBL and in photoreceptor precursors (Figure 4E). Finally, RXRγ, a transcription factor specific to cones and ganglion cells in the retina, was expressed by cone cells at the apical margin of the presumptive ONL in hPSC-derived retina as early as 8 weeks. The appearance of this marker earlier than the rod photoreceptor marker, NRL, suggests that cones differentiate before rods in this culture system (Figure 4F), in keeping with normal development. The time course of expression of these early retinal and photoreceptor markers was confirmed by RT-PCR (Figure 4G).
We next sought to further characterize hPSC-derived cone maturation, specifically components of the phototransduction cascade. In vivo, the developmental pattern of human cone opsin expression has been described; short wavelength opsin (S OPSIN) and long/medium wavelength opsin (L/M OPSIN) proteins first appear in the fovea around Fwk 11 and 15, respectively (Xiao and Hendrickson, 2000). In keeping with these timings, the first S OPSIN cones were observed from week 12 of differentiation, increasing in both level of expression and numbers of positive cells by 20 weeks (Figure 5A), while L/M OPSIN cones were not detected until week 17 (Figure 5B). Next, to determine the efficiency of cone differentiation, we quantified the number of cells expressing the pan-cone marker, ARRESTIN3, at different developmental stages (Figures 5C–C″). At week 12, 6.8% (±1.9%; n = 30 images, N = 3 differentiations) of the cells in the NRVs were ARRESTIN3+ (Figures 5C and 5D); this increased substantially, to 17.8% (±5.6%; n = 30 images, N = 3 differentiations), by week 20 (Figure 5C’’–D). Of note, ARRESTIN3+ cells were more numerous than OPSIN+ cones at all time points examined. Schematic Figure 5E summarizes the temporal expression of cone markers in vitro. A 3D view of a whole NRV at week 20 shows the efficient generation of ARRESTIN3+ cones and RHODOPSIN+ rods (Figures 5F and F′) and the different morphology of cones and rods was clear (Figures 5G–5G″): Rods had small cell bodies (∼5 μm) and elongated ISs, while cones had large cell bodies (11 ± 1 μm) (Figure 6E) and stubby bud-like ISs. Furthermore, ARRESTIN3+ hPSC-derived cones co-stained for peanut agglutinin, which labels the cone matrix sheath and cone pedicles (Figure 5H). The cone phototransduction marker, CNGB3, was located to the OS-like region (Figure 5I). qRT-PCR confirmed the expression of cone-specific phototransduction markers, which increased with development (Figure 5J). Together, these data show that our culture system supports the robust generation of large numbers of hPSC-derived photoreceptors, including a high proportion of cones. Importantly, the developmental expression profile and morphology of these cones is similar to what is currently known about developing human cones.
Transplantation of hPSC-Derived Cones into the Nrl−/− Adult Retina
Next, we sought to assess the transplantation capacity of hPSC-derived cones into degenerate adult mouse retina. There are no animal models that accurately reflect the pathology of macular degeneration. However, to mimic transplantation into the cone-rich environment of the para-foveal region of the human retina, we transplanted purified populations of cones from weeks 14 to 17 of differentiation into the adult Nrl−/− mouse, a model in which all cells that were destined to become rods adopt an S-cone phenotype (Mears et al., 2001). L/M opsin cones were virally labeled at weeks 14–16, when the neuroepithelia is already postmitotic, and ShH10.L/Mopsin.GFP-positive (L/Mopsin.GFP+) cells were isolated by fluorescence-activated cell sorting (FACS) (7% ± 3.8% GFP+ cells of NRVs) 2 weeks later. This is lower than the total number of cones in the NRVs and reflects viral transduction efficiency (Figure S6A).
Eyes were examined 3 weeks after transplantation: despite not having received immune suppression, 50% of transplanted eyes (N = 15/30) contained robust numbers of hPSC-derived cones, predominantly located within the subretinal space (Figure 6A) in close apposition to the host ONL (Figure 6A, highlighted panels). In some instances, the transplanted cells formed neuro-rosettes, similar to those formed by murine RPCs (MacLaren et al., 2006) (Figures S6B and S6C). Remarkably, a small number of L/Mopsin.GFP+ hPSC-derived photoreceptors appeared to have become incorporated into the host murine ONL (Figures 6B–6C″, 6F–6J″, and S6D–S6F′). These cells were correctly oriented within the ONL, typically close to the subretinal cell mass, and, importantly, displayed morphological features typical of mature human, as opposed to mouse cones, including comparatively large cell bodies and nuclei (significantly larger than surrounding murine rods and cones). Moreover, some displayed cone pedicle-like structures terminating within the host OPL (Figures 6B’, 6C′, 6F, 6G, and 6J, arrowheads) and nascent segment-like structures projecting toward the host RPE (Figures 6C’’, 6H′, and 6I, arrows). Punctate RIBEYE labeling was observed in the GFP+ pedicle-like structures in the host OPL (Figures 6F and 6F′, arrowheads) and incorporated L/Mopsin.GFP+/hNUCLEI+ cells were ARRESTIN3+ (Figures 6G and 6G′). As described above, despite robust viral labeling of cone cells (in which GFP expression is driven by the L/Mopsin promoter) at the time of transplantation (17–20 weeks), L/M OPSIN protein expression was restricted to a relatively small number of cones in vitro (Figure 5B). However, further maturation of the transplanted cones is possible in vivo: at 3 weeks after transplantation, robust expression of L/M OPSIN was observed in both the cell mass (Figures 6H and 6H′) and in L/Mopsin.GFP+ cones incorporated within the host ONL (Figures 6H and 6I′, arrows).
The human origin of these cells was established in three ways: co-staining the cells for human nuclei antigen (hNUCLEI+), measuring the size of the nuclei, compared with endogenous Nrl−/− murine photoreceptors, and by performing fluorescent in situhybridization (FISH) for mouse Y chromosome following the transplantation of L/Mopsin.GFP+ cones into male donors. IHC for hNUCLEI confirmed the presence of small numbers of L/Mopsin.GFP+/hNUCLEI+ cones unambiguously located within the host ONL (55 ± 38 cones per eye, n = 9 retinae; N > 4 independent experiments) (Figure 6D). The mean diameter of these L/Mopsin.GFP+/hNUCLEI+ nuclei was 13 μm (±1.8, mean ± SD, n = 42 nuclei, N = 3), compared with 6 μm (±1.0, mean ± SD, n = 52 nuclei, N = 3) for endogenous cone-like photoreceptor nuclei in the Nrl−/−mouse retina (Figure 6E). Next, we performed FISH, as described previously (Pearson et al., 2016). Here, donor cells were derived from the female H9 ESC-cell line, and were transplanted into male mouse hosts. Incorporated L/Mopsin.GFP+ cells were, without exception, negative for Y chromosome (Figures 6J, 6J″, S7A, and S7B; Movie S1) demonstrating that these cells were not endogenous mouse photoreceptors and did not arise through material transfer. Virtually all cells within the host male eyes were Y chromosome+, while transplanted human H9 female cells (both incorporated and those remaining in the cell mass) and female mouse eyes were always negative for Y chromosome, demonstrating the specificity of the probe to male mouse cells (Figures S7C and S7D). Occasionally, we observed GFP+ cells located within the ONL that were positive for the Y chromosome (Figure S7E). However, these were always negative for hNUCLEI (30 ± 34 GFP+ cells, n = 7 retinae; N > 4) (Figures S7F and S7F′) and morphologies and nuclear sizes typical of murine Nrl−/− photoreceptors (6 ± 1.0 μm; n = 33 nuclei, N = 3; Figure 6E). These are therefore likely to be the result of GFP labeling of murine host cells. This could arise from carry-over of viral particles, although this is unlikely to account for many cells (Gonzalez-Cordero et al., 2013, Pearson et al., 2016); indeed, no GFP+ cells were seen in eyes receiving control injections of the final wash from the cell preparation steps. Alternatively, GFP expression in host cells may be due to cytoplasmic material transfer between the human donor cells and the murine host photoreceptors, as recently described for murine-murine rod (Pearson et al., 2016, Santos-Ferreira et al., 2016, Singh et al., 2016, Ortin-Martinez et al., 2017) and cone photoreceptor transplantation (Decembrini et al., 2017). Interestingly, very little evidence of either donor cell incorporation or cytoplasmic material transfer was observed following the transplantation of hPSC-derived L/Mopsin.GFP+ photoreceptor precursors into wild-type animals, despite the presence of substantial cell masses (2 ± 3 GFP+ cells per eye; n = 7 retinae, N = 4) (Figure S7G).
We also assessed the transplantation capacity of the L/Mopsin.GFP-negative population isolated by FACS from week 17 cultures (Figures S7H–S7K), but only a very small number of hNUCLEI+ cells were observed within the host ONL, some of which contained rod-specific proteins (Figures S7K and S7K’, head arrow). Together, these experiments demonstrate that hPSC-derived cone photoreceptor precursors can become incorporated into an adult “cone-rich” (murine) retina, albeit in small numbers.
Replacement of hPSC-Derived Cones in Advanced Retinal Degeneration
Next, we sought to test if hPSC-derived cones have the potential to replace lost photoreceptors in a mouse model of advanced degeneration. This better reflects the clinical scenario encountered in patients with macular degeneration where there is extensive loss of central cone photoreceptors. A purified population of 100,000 17 week L/Mopsin+ cones was transplanted into the Aipl1−/− mouse model of Leber congenital amaurosis (LCA4) between 8 and 12 weeks of age, when almost all the photoreceptors have died (Dyer et al., 2004, Ramamurthy et al., 2004). In vivofundus imaging of transplanted eyes showed GFP in the superior hemisphere (Figure 7A, arrowheads). As before, hPSC-derived L/Mopsin.GFP+ cones survived in the subretinal space with or without the use of immune suppression (N = 7/13) (Figure 7B). In contrast to transplants into the Nrl−/− host, subretinal rosette formation was rare in this model. Instead, the hPSC-derived cone cells formed a distinct layer adjacent to the host INL, which was often only a single-cell layer but in some instances formed thick (∼25 μm) layers. The human origin of these cells was confirmed by human-specific mitochondria (Figures 7C and 7C′) and hNUCLEI (data not shown) IHC. The transplanted hPSC-derived cones expressed ARRESTIN3 and L/M OPSIN (Figures 7D–7E′), and in some cases nascent hMITOCHONDRIA+ and PRPH2+ segments could be seen extending toward the RPE and neurites extending toward the host INL (Figures 7F–7H, arrows). IHC showed that, in the absence of rods, the afferent terminals of host PKC+ rod bipolar cells, which normally contact rods, are in close proximity to the hPSC-derived L/Mopsin.GFP+ cones (Figure 7I). Dendrites of CALBINDIN+ horizontal cells were also found in close apposition with L/Mopsin.GFP+ neurites (Figure 7J) that expressed the presynaptic protein, RIBEYE (Figures 7K–7L′). It has been proposed that reactive Müller glia represent a physical and/or biochemical barrier to transplanted cells (Kinouchi et al., 2003). However, while GFAP+ Müller glial processes were seen throughout the Aipl1−/− retina, these did not appear to impede donor-host interactions (Figure 7M).
Together, these data suggest that hPSC-derived cones have the potential to replace lost photoreceptors in end-stage retinal degeneration. Further studies are required to determine whether hPSC-derived cones can transmit visual information to mouse interneurons and how long these cells can survive post-transplantation. A small number of transplanted cones were still seen at 6 weeks post-transplantation in the unsuppressed host (n = 3 eyes) and this might be improved by immune suppression.
Discussion
Retinal degenerations, including maculopathies due to loss of cone photoreceptors, are a major cause of blindness worldwide. In this study, we describe a modified 2D/3D retinal differentiation system for hPSCs that permits the generation of human photoreceptors in numbers sufficient to begin to assess their transplantation potential. We demonstrate that hPSC-derived rod and cone photoreceptors differentiate in vitro in a time course that closely reflects what is known about normal human retinal development (O’Brien et al., 2003, Xiao and Hendrickson, 2000, Hendrickson et al., 2008, Hendrickson et al., 2012, Hollenberg and Spira, 1973). Moreover, this system supports the development of several key structures typical of later stages of photoreceptor maturation, including the formation of ISs, CC, and nascent OS-like and presynaptic structures. Importantly, although our protocol was not specifically aimed to produce a high percentage of cones, as described by Zhou et al. (2015), cones were generated with sufficiently high efficiency to readily permit their isolation for transplantation purposes. Following transplantation, we observed incorporation of hPSC-derived cones into the adult “cone-rich” Nrl−/− mouse retina. When transplanted into the Aipl1−/− model of end-stage retinal disease, these cells formed synaptic-like structures in close apposition to host interneurons and expressed photopigments. Together, these findings suggest that hPSC-derived cone precursors have the potential to replace lost photoreceptors following transplantation into the degenerate adult retina.
Increasingly, hPSC-derived retinal cultures are being used not only to investigate cell therapy strategies, but also to model retinal development and disease (Jayakody et al., 2015). The ability to accurately model human development in vitro presents opportunities for testing developmental hypotheses that so far have been limited to animal models, due to the restricted availability of fetal tissue. In this study, we performed a detailed characterization of photoreceptor development and show the differentiation of both human cones and rods in a manner that accurately reflects what is currently known about human retinal development. We observed progenitors positive for ONECUT1, OLIG2, and OTX2, all key transcription factors in cone specification and fate decisions in the chick and rodent retina (Emerson et al., 2013, Hafler et al., 2012). In addition, we report, for the first time, the presence of ONECUT1+ photoreceptor precursors, as identified by CRX co-staining.
The formation of well-developed hPSC-derived photoreceptors, in which the majority bear structures such as cilia and nascent OSs, brings us closer to being able to use hPSC-derived retinas to study the mechanisms underlying retinal degenerations, as well as providing a testing platform for drug screening and the development of treatments such as gene therapy. However, although our protocol efficiently generated hPSC-derived photoreceptors bearing nascent segments, the disc membranes within the OSs were not tightly stacked, possibly due to the lack of close apposition to the RPE. Future studies will need to address these limitations.
A major objective of the current study was to be able to generate stage-matched hPSC-derived photoreceptors in numbers sufficiently large enough to permit the isolation of cones for transplantation. We, and others, have previously reported the transplantation of donor-derived murine cone and cone-like photoreceptor precursors (Lakowski et al., 2010, Santos-Ferreira et al., 2015, Smiley et al., 2016, Decembrini et al., 2017). These studies reported the presence of reporter-labeled cells within the host ONL, including small numbers with cone-like morphologies. However, the majority were morphologically similar to rod photoreceptors. In light of our recent findings showing the transfer of cytoplasmic material between donor and host photoreceptors (Pearson et al., 2016), it is likely that these were labeled host photoreceptors, rather than integrated donor cells (Decembrini et al., 2017). Similarly, studies reporting the apparent integration of virally labeled human PSC-derived photoreceptors in the murine host did not confirm their human origin (Lamba et al., 2009, Lamba et al., 2010). As in these studies, we also observed L/Mopsin.GFP+/hNUCLEI-labeled photoreceptors within the host ONL whose nuclei and morphology resembled that of mouse photoreceptors, albeit in small numbers. This suggests that, like murine photoreceptor precursors, human photoreceptors may also be able to transfer cytoplasmic material to host cells. Importantly, however, the results presented here also show unequivocally that hPSC-derived L/Mopsin.GFP+/hNUCLEI+ cones can be incorporated into the ONL of host retinae. Interestingly, greater numbers of hPSC-derived L/Mopsin.GFP+/hNUCLEI+ cones were found in the “cone-rich” Nrl−/− retina compared with the wild-type retina. The Nrl−/− retina displays a slow degeneration (Roger et al., 2012) as well as a disrupted OLM (Stuck et al., 2012), each factors that could have facilitated the incorporation of hPSC-derived photoreceptors observed here, as described previously for other models (West et al., 2008, Pearson et al., 2010, Barber et al., 2013). Indeed, in another study, we have found the host retinal cytoarchitecture to play a major role in the relative contributions of donor cell incorporation and cytoplasmic material transfer following transplantation of murine photoreceptors (R.A.P., unpublished observations) and both mechanisms should be considered in future studies when assessing transplantation outcomes.
Dissociated cells and retinal sheets have each been transplanted into animal models of end-stage disease (Shirai et al., 2016, Assawachananont et al., 2014). Recently, mixed populations of hPSC-derived retinal cells have been transplanted into the Rd1model, resulting in improvements in visual function (Barnea-Cramer et al., 2016). To date, no study has demonstrated the transplantation of a pure population of cones where the developmental processes and structural maturation of photoreceptor genesis have so closely mimicked normal human retinal development. We report the transplantation of hPSC-derived cone precursors into the Aipl1−/− mouse model of end-stage retinal degeneration. The transplanted cones formed nascent segment-like structures and extend basal processes toward host interneurons by 3 weeks post-transplantation. It is not yet possible to say if these cells are light responsive or if the basal terminals represent true synaptic connections. For this reason, we limit our descriptions to incorporation, rather than functional integration. Further investigation of the importance of the developmental stage of donor cells at the time of transplantation, together with assessments of extended survival and maturation of these cells is required, as are studies into the functional capabilities of these cells both within the NRV and following transplantation. It remains to be determined whether human cone photoreceptors would be able to functionally connect to the mouse retinal circuitry, which is biased toward rod photoreceptor connections. The use of larger animal models that more closely resemble the structural anatomy of the human retina may be required to determine if transplanted hPSC-derived photoreceptors can transmit visual information.
Cone transplantation may provide a potential treatment for a range of maculopathies that result in cone cell death. Our results show that it is possible to generate sufficient human L/M OPSIN cones to transplant both into a “cone-rich” retina and into a model where no endogenous photoreceptors remain. The results presented here establish that human cones can become incorporated within an adult mammalian retina and therefore further support the development of photoreceptor cell transplantation as a treatment for retinal degeneration.