Molecular characterization and sub-retinal transplantation of hypoimmunogenic human retinal sheets in a minipig model of severe photoreceptor degeneration Free

In most retinal degenerative diseases (RDs), loss of visual function results from the death of photoreceptors (PRs), the specialized light-sensitive cells involved in phototransduction (Pacione et al., 2003). Photoreceptors exist in two types, rods and cones. Rods are the most abundant photoreceptors in the human retina and respond to dim light. They are involved in night vision and important for peripheral vision. Cones respond to intense light and are required for color, daylight and high-resolution central vision (Michaelides et al., 2006; Aboshiha et al., 2016). In contrast to other mammals, the eye of modern primates (apes, monkeys and hominins) contains a unique circular structure of 4-5 mm diameter called the macula located near the center of the retina (Curcio et al., 1990; Franco et al., 2000). The macula is highly enriched in cone photoreceptors and has a cone-only smaller region of ∼1.5 mm diameter called the fovea (Curcio et al., 1990; Franco et al., 2000). Retinal degenerative conditions (late-stage retinitis pigmentosa, macular degeneration and inherited retinal dystrophies) all share the common features of loss of cone photoreceptors and central (macular) vision (Michaelides et al., 2006). Retinitis pigmentosa is one of the most common and genetically heterogeneous retinal degenerative diseases (Daiger et al., 2013). While the disease primarily affects rods, it usually progresses toward legal blindness when cones located in the center of the retina are lost. Loss of central vision is also predominant in cone-specific diseases such as Stargardt disease, cone dystrophies, cone/rod dystrophy, Leber congenital amaurosis (LCA) and macular degeneration (Gill et al., 2019).

Despite its major impact on quality of life, therapeutic interventions against vision loss remain of limited effectiveness and are confined to a small number of cases. Gene therapy approaches aiming at correcting or silencing the disease-causing mutation or replacing the defective gene are promising avenues. They rely on the injection of viruses and have, to date, demonstrated limited success, primarily restricted to specific cases, such as patients with LCA resulting from a mutation in RPE65 (Takkar et al., 2018). Gene therapy is also limited because it focuses on a single gene or mutation at a time, when RDs are associated with mutations in hundreds of genes (Nash et al., 2015; Kiel et al., 2017). Hence, while gene therapy can prevent photoreceptor degeneration, it cannot replace lost photoreceptor cells. Prosthetic sub-retinal implants (e.g. Argus II) have also been developed. However, these devices are invasive and can only provide limited visual stimuli in patients (Parmeggiani et al., 2017). Because human vision depends mainly on the macula for most activities, there is a need for cone cell replacement therapy to restore central vision.

Although promising, photoreceptor transplantation therapy in animal models has been debated for its efficiency and suitability. It has been suggested that grafted rod precursors can undergo material transfer (RNA, protein and exosomes) with the host photoreceptors, potentially leading to non-cell-autonomous effects (Pearson et al., 2016; Ortin-Martinez et al., 2017). On the other hand, work from other groups suggests that dissociated human photoreceptors grafted in the sub-retinal space can integrate and differentiate efficiently into the photoreceptor cell layer of immunodeficient mice, and extensive analyses confirmed that material transfer or cell fusion does not occur in this context (Zhu et al., 2017). Notably, when grafted into the retina of late-stage rd1 mice lacking photoreceptors, mouse induced pluripotent stem cell (iPSC)-derived retinal tissues or dissociated human iPSC-derived cone precursors can form presumptive synapses with endogenous bipolar neurons and improve visual function, thus also generally excluding the thesis of material transfer to endogenous photoreceptors (Assawachananont et al., 2014; Ribeiro et al., 2021; Gasparini et al., 2022). Results from these important experiments also suggest that, even in congenitally blind mice, the central connections from the retina are still relatively intact and functional (Assawachananont et al., 2014; Yao et al., 2018; Ribeiro et al., 2021). However, grafting dissociated cone and/or rod photoreceptors results in a disorganized group of cells lacking apicobasal polarity and a retinal tissue-like structure. Although pioneer work from Yoshiki Sasai's laboratory provided a revolutionizing method to generate human retinal organoids (Eiraku et al., 2011; Nakano et al., 2012; Kuwahara et al., 2015; Fathi et al., 2021), their heterogeneous nature may represent a limitation to retinal transplantation therapy, such as the formation of neural rosette-like structure within the host sub-retinal space (Assawachananont et al., 2014; Shirai et al., 2016). Human iPSC-derived retinal organoids have recently been employed to produce sheet-like grafts for transplantation in rats, yielding promising outcomes, but lacking apicobasal polarity and developing neural rosette-like structures within the host sub-retinal space (Watari et al., 2023). Transplanted rat and human fetal retinas have shown connectivity and visual recovery in a rat model of retinal degeneration (Seiler et al., 2017; Lin et al., 2018), but the availability of human embryonic retinas is limited, and their usage faces strong ethical concerns. Thus, the current protocols to purify dissociated cone precursors or to generate sheet-like grafts from retinal organoids involve laborious and specialized manual procedures, resulting in limited quantities and scalability for human clinical application.

We previously reported on an independent method to produce abundant cone precursors from human embryonic stem cells (ESCs) and iPSCs, forming uniform and polarized retinal sheets (RSs) with a 3D architecture in a 2D culture environment. These features allow for easy customization by cutting the material into the desired size and shape and provide ample material for multiple transplants (Zhou et al., 2015; Barabino et al., 2020). The method relies on the use of human recombinant COCO (also called DAND5), a Cerberus-like family member protein that can antagonize soluble BMP, TGFβ and WNT ligands. The method allows the spontaneous formation of a ∼100 µm thick, multilayered and polarized RS containing up to ∼60% of cone progenitors and precursors, thus possibly suitable for macula replacement therapy (Zhou et al., 2015).

We herein describe the molecular and cellular characterization of RSs produced from isogenic control and hypoimmunogenic universal donor (UDC) iPSC lines. Using a cobalt chloride-induced model of severe photoreceptor degeneration in the Yucatan minipig, we show that RS punches transplanted in the sub-retinal space can form a new but immature cone photoreceptor cell layer within the host retina, with evidence of visual response to bright light at the lesion site.

We induced the differentiation of the isogenic WTC/Actin-GFP (WTC^Act), WTC/Tubulin-RFP (WTC^Tub) and WTC-UDC (WTC^UDC) iPSC lines into RSs (Fig. S1A; Materials and Methods). CRX is a master regulator of photoreceptor development, and flow cytometry analysis at 60 days in vitro (DIV60) revealed that, when compared with unstained control cells, ∼65% of the cells were positive for CRX (12% CRX-high and 53% CRX-low) (Fig. S1B). Using RNA-sequencing (RNA-seq), global gene expression of DIV60 RSs was compared with that of the human embryonic retinal development atlas (Hoshino et al., 2017). This revealed that RSs express the pan photoreceptor genes CRX, OTX2 and ROM1, and the cone-specific genes OPN1SW (encoding S-opsin) and PDE6B (encoding for a light-responsive phosphodiesterase) at levels that matched day 67-94 of the human embryonic retina (Fig. S1C). To test the functionality of RSs in vitro, we measured the proportion of endogenous cyclic (c) GMP degraded by light-responsive phosphodiesterases. This revealed light-induced cGMP degradation in RSs when compared with those treated with a phosphodiesterase inhibitor, thus showing a biochemical response to bright light (Fig. S1D) (Kawamura and Murakami, 1986; Wensel, 1993; Calvert et al., 1998; Chen et al., 2015; Turunen and Koskelainen, 2019).

To deepen the characterization of the cell populations, we performed single-cell RNA-seq (scRNA-seq). DIV60 RSs were found to contain nearly all cell types present in the human embryonic eye (control human retina samples ranging from day 94 to day 129 post-conception) (Cao et al., 2020), except for gene sets specific to bipolar neurons, astrocytes and Müller glia (Fig. 1A). UMAP analysis revealed that retinal progenitor cells (RPCs) were the most abundant cell types present in RSs (Fig. 1A). Developmental kinetics analysis of RSs at DIV5, DIV30 and DIV60 revealed a linear relationship in pseudo-time representation (Fig. 1B), with the identification of markers for early (MKI67 and PAX2), intermediate (ASCL1 and VSX2) and late (RCVRN and RPE65) groups (Fig. 1C). RAX, PAX6, LHX2, SIX3, SIX6 and VSX2 are expressed at the earliest stages of retinal development (Furukawa et al., 1997a; Loosli et al., 1999; Nishina et al., 1999; Livne-Bar et al., 2006; Hägglund et al., 2011; Shaham et al., 2012; Diacou et al., 2018, 2022; Raeisossadati et al., 2021) and delineated a prominent RPC population present in RSs at DIV5, DIV30 and DIV60 (Fig. 1D; Fig. S2A).

OTX2 and CRX are closely related transcription factors that are crucial for photoreceptor development. OTX2 is expressed early in retinal development, including in retinal progenitor cells, photoreceptor precursors and mature photoreceptors. In contrast, CRX expression begins later, becoming prominent only as retinal progenitor cells differentiate into photoreceptor precursors and mature photoreceptors (Furukawa et al., 1997b; Bibb et al., 2001; Nishida et al., 2003; Ruzycki et al., 2018; Yamamoto et al., 2019; Tegla et al., 2020). Both OTX2 and CRX are also expressed in a subset of bipolar neurons. Accordingly, OTX2 was expressed earlier and more broadly than CRX and all other photoreceptor markers in RSs at DIV5, DIV30 and DIV60 (Fig. 1E; Fig. S2B). Although OTX2 is expressed in all photoreceptor lineage cells, it is also found in the retinal pigment epithelium (RPE), in contrast with CRX. Our annotation of cell types in RSs at DIV5, DIV30 and DIV60 revealed that a significant fraction of OTX2-expressing cells at DIV60 were indeed RPE cells (Fig. 1E,F). This also revealed the presence of two main cell populations at DIV5, i.e., RPCs and ocular surface ectoderm. The ocular surface ectoderm can give rise to both the lens and cornea. This cell population expressed keratin genes as well as PAX6 and SIX3, both required for ocular surface ectoderm specification in mice (Fig. 1F; Fig. S2D,E) (Kamachi et al., 2001; Aota et al., 2003; Liu et al., 2006; Wolf et al., 2009; Machon et al., 2010; Xie et al., 2013).

Rare putative pluripotent stem cells expressing both POU5F1 and NANOG at DIV5 were found very close to the ocular surface ectoderm cell population but distant from RPCs (Fig. 1F; Figs S2D,E and S3A,B). Between DIV5 and DIV30, the early RPC population (expressing high levels of MKI67 and PCNA) gradually diminished and disappeared by DIV60, making room for RPCs, late RPCs, photoreceptor precursors and a distinct retinal precursor cell population classified here as retinal neuronal fate (Fig. 1F; Fig. S2D,E). Notably, putative pluripotent stem cells were not observed at DIV30 and DIV60, consistent with the absence of teratoma formation in NOD/SCID mice grafted with dissociated DIV60 RSs (Fig. 1F; Figs S2D,E and S3A-C). Heat map representation of annotated genes in DIV60 RSs further revealed key markers distinguishing the identified cell populations, including lens placoid (CRYGA and CRYBA1) and corneal cells (KRT5, KRT15 and CLDN7) (Fig. 1G) (Wistow, 2012; Sun et al., 2015; Limi et al., 2019; Ligocki et al., 2021). Pax2 expression in early mouse eye development occurs in the optic stalk and marks the boundary between the presumptive neural retina (Pax6⁺) and the optic nerve head (Pax2⁺) (Schwarz et al., 2000). The PAX2⁺, VAX1⁺ and NKX6-2⁺ group thus represents a defined ocular neuroectoderm cell population corresponding to the presumptive optic stalk and later to its derivatives, such as the optic nerve and optic nerve head (Fig. 1G; Fig. S2B) (Mui et al., 2005; Moreno-Bravo et al., 2010).

In each cluster, we observed notable cellular diversity. For example, late RPCs, while retaining their characteristic identity, also contain a distinct subgroup (6.4%) that begins expressing RCVRN, signifying a transition towards precursor-like states (Fig. 2B,D; Fig. S2C). Similarly, in the precursor cell population, many cells still retain expression of PAX6 and VSX2 (27.8%), signifying their persistence in a more immature state (Fig. 1D; Fig. S2A). Intriguingly, a subset of precursor cells displays a shift in identity by expressing PDE6H, GNGT1, GNGT2 and ARR3 suggesting a progression toward a more mature cellular state (Fig. S2C) (Lukowski et al., 2019). These observations emphasize the dynamic nature of cell identities within clusters and highlight the possibility of transitional states within specific cell populations.

While RSs are enriched for RPCs and photoreceptor precursors, they also contain other cell types that could interfere with their application in therapy. Large neural rosettes were isolated using a biopsy punch and analyzed by scRNA-seq (Fig. 2A). When compared to DIV60 RS cultures (i.e. the bulk population), RS punches were highly enriched for the late RPCs cluster, with about the same proportion of photoreceptor precursors (Fig. 2A′-D; Fig. S4A-E). To identify rod-committed cells within the CRX⁺ subset (encompassing the whole photoreceptor cell lineage), we assessed co-expression with NRL, a key regulator of rod differentiation. About 12.5% of the cells were CRX⁺/NRL⁺ in bulk RSs, and this was reduced to 5.5% in RS punches (Fig. 2B,C). Notably, RS punches exhibited a significant depletion of RPE, amacrine, horizontal, lens placoid and corneal epithelium cells (Fig. 2E). Consequently, this resulted in a highly enriched population of late RPCs and photoreceptor precursor cells suitable for transplantation therapy.

We next conducted immunofluorescence (IF) analyses on RS punches at various time points of the differentiation procedure (Fig. 3A). DIV60 RS punches were composed of a layer of 4-8 nuclei expressing CRX, OTX2, VSX2 and RCVRN (Fig. 3B; Fig. S5A). The tissue was polarized, with RCVRN and PNA labeling being present at the apical side over the nuclear layer (Fig. 3B; Fig. S5A). The photoreceptor precursor nuclei on the apical side were strongly positive for CRX and OTX2, with the nuclei localized in the center and the basal side of the tissue positive for VSX2 (Fig. 3B; Fig. S5A). In RS punches from both WTC^UDC and WTC^Act cell lines, 20-25% of the cells show strong expression of CRX and OTX2 (Fig. 3C). These findings align with the scRNA-seq results and are comparable with fluorescence-activated cell sorting (FACS) analysis of the bulk population when focusing on CRX-high cells (Fig. 2B; Fig. S1B).

Transmission electronic microscopy analysis at DIV60 further revealed the presence of tight junctions at the border of the mitochondria-rich nascent inner segments (Fig. S5B). Dense organelles corresponding to the basal body of primary cilia were also observed (Fig. S5B). At DIV90, we observed an initial but consistent S-opsin signal spanning the entire tissue, with some signal concentrated at the most apical side. Additionally, we noticed a distinct and well-defined Peanut agglutinin (PNA) staining in the emerging bud-like outer segments (OS) of the immature cones (Fig. 3D, arrows). Interestingly, DIV60 RS punches exhibited a natural tendency to fold inwards and form organoid-like structures when lifted from the plate and maintained in suspension cultures (Fig. 3E). This inherent folding behavior improves their viability during extended ex vivo maintenance. Hence, retinal organoid-like structures were employed for long-term culture to evaluate photoreceptor maturation. At DIV150-DIV160, retinal organoid-like structures were strongly positive for CRX, OTX2 and expressed apically PNA, RCVRN, S-opsin and M/L-opsin (Fig. 3F,G), with relatively mature OSs resembling those found in the adult human retina (Fig. 3G). These results show that RPCs and PRs precursors present within DIV60 RS punches can ultimately develop into mature cones ex vivo when grown as retinal organoid-like structures.

Pig eye anatomy and dimensions closely resemble those of humans and include a visual streak enriched in cones, bearing some resemblance to the macula found in primates (Choi et al., 2021). We induced severe photoreceptor degeneration in this ‘pseudo-macula’ of adult Yucatan minipigs through sub-retinal cobalt chloride injections (Fig. 4A,B; Movie 1) (Hara et al., 2006; Shirai et al., 2016). Using a single apparatus, we conducted Optical Coherence Tomography (OCT) and multifocal Electroretinogram (mfERG) analyses under photopic conditions. These tests confirmed retinal lesions measuring 4-9 mm in diameter 1 month after treatment, with loss of visual function at the lesion site in correspondence with the outer nuclear layer (ONL)-free area (Fig. 4C). Variations in lesion size and severity were primarily linked to solution reflux post-injection, affecting the final cobalt chloride volume in the retinal bleb (Table S1). One-month post-treatment, IF analyses confirmed complete loss of the ONL in the damaged area (Fig. 4D-F). While the RPE and inner nuclear layer (INL) generally remained intact, we observed more severe damage in some eyes, especially in the central portion of the lesion, which could lead to some RPE and INL loss (Fig. 4D-F). Staining for VSX2, which labels bipolar neurons in the mature retina, indicated a generally intact INL at the lesion site (Fig. 4G).

To overcome a robust innate immune response previously observed in pilot experiments, minipigs were immunosuppressed using a multiple-drug regimen starting 1 week before surgery and maintained throughout post-transplantation (Fig. 5A; Materials and Methods). For transplantation, we induced a sub-retinal bleb (∼5 mm diameter) using a saline solution. RS punches were isolated from fresh DIV60 RS cultures and loaded into a custom injector. Next, one or multiple RS punches, each ranging from 1 to 3 mm in diameter, were injected into the sub-retinal bleb using shock waves (Fig. 5B; Table S1; Movie 2).

At the time of surgery, the circular retinal lesion was visualized by fundus observation, and we could confirm the persistence of the graft at the lesion site by OCT analysis at days 30, 45 and 60 post-transplantations (Fig. 5C; Fig. S6; Table S1). OCT and histological analyses revealed one or multiple issues with most grafts, including folding onto themselves (D60-03L, D60-09R, D60-09L), inverted apicobasal orientation (D60-04L), placement within severely damaged areas with affected host INL (D60-05R, D60-04L) or degeneration of the graft for its proximity to the incision site resulting in infiltration and partial destruction by macrophages (D60-07L). Another common issue was the positioning of the graft in the periphery or outside the fully damaged area (D60-01L, D60-04R), preventing proper interaction with the host retinal tissue and meaningful functional studies (Fig. S6; Table S2). Histological analysis of the D60-03R eye, however, revealed the presence of a flat graft forming a relatively uniform retinal tissue within the host sub-retinal space and expressing the human-specific markers human nuclear antigen (HuNu) and STEM-121 (Fig. 5D; Fig. S7C). We observed that two portions of this graft were unfolded and properly oriented, with CRX⁺ and PNA⁺ cells of the graft located apically and in close association with the host RPE (Fig. 5D, inset; Fig. S7A). While photoreceptor markers, such as CRX and RCVRN, were expressed apically in the entire and new ∼3-4 nuclei thick photoreceptor-like layer (Fig. 5D), the formation of proper OSs was not observed, and only low levels of S-opsin and M/L-opsin were visible at the apical region of the graft (Fig. S7D). Within the graft, we noticed a complementary distribution of CRX and VSX2, with VSX2⁺ retinal progenitors located inward of the graft and facing the endogenous INL (Fig. 5D). Some of these retinal progenitors exhibited low levels of Ki67 expression while testing negative for PCNA (Fig. S7E). This state is commonly observed during development in quiescent adult stem cells and progenitors, which remain in a non-proliferative resting state while remaining metabolically active and prepared to re-enter the cell cycle (Manoir et al., 1991; Sobecki et al., 2017; Sun and Kaufman, 2018). Between these two populations, we observed the presence of presumptive photoreceptor precursors co-expressing CRX and VSX2 (Fig. 5D, inset; Fig. S7A,B). Notably, expression levels of CRX and OTX2 in this human iPSC-derived graft were similar to that of the adult pig's endogenous photoreceptors (Fig. 5E). The graft was also positive for HuNu (Fig. 5E, white labeling). The tight junction protein ZO-1 demarcates the outer limiting membrane located between the photoreceptor's cell body and the inner segments (IS) (West et al., 2008; Omri et al., 2010; Pearson et al., 2010). Histological analyses revealed that the graft contained genuine but immature photoreceptors having a ZO-1⁺, RCVRN⁺ and PNA⁺ apical edge in direct contact with the host RPE (Fig. 5F).

To assess graft function, we conducted post-surgery follow-up analyses using mfERG recordings under photopic conditions to specifically measure the cone response, excluding the rod contribution. These assessments were performed in all transplanted animals at days 30, 45 and 60 post-transplantation, except in cases where technical challenges, such as cataract formation or eye movement during analysis, prevented data collection (Table S2).

To visually represent the presence or absence of N1 and P1 waves within damaged and transplanted areas, we performed a double-blinded qualitative analysis. N1 and P1 waves represent retinal responses to visual stimuli, with N1 indicating the initial negative deflection mainly associated with cone photoreceptor hyperpolarization. In contrast, P1 indicates the positive deflection associated with the depolarization of bipolar cells and inner retinal neurons (Wördehoff et al., 2004; Lai et al., 2007). In the healthy region of the host retina, both P1 and N1 waves were observed (green hexagons), whereas the damaged area often lacked both waves (red hexagons) or one of the two waves (yellow hexagons) (Fig. 6A). Within the grafted region (highlighted by the dotted circle), traces consistently indicated the presence of N1 and P1 waves (Fig. 6A). Next, we performed quantitative analysis of latency and amplitude of N1 and P1 waves in healthy and damaged areas with grafts. We observed a partial but significant rescue of the P1 amplitude in the grafted area starting from day 45 and maintained till the endpoint at day 60 (Fig. 6B). This result is notable as the host photoreceptor ONL was completely ablated at this position before transplantation (Figs 4C-F and 5C). Limited N1 response can be attributed to its intrinsic characteristics since it is much smaller than the P1 wave. As a negative control, we conducted similar analyses on the D60-03L eye, which showed similar expression levels of photoreceptor markers like CRX, PNA and RCVRN as the D60-03R eye but was completely folded and improperly polarized, and it showed no functional improvement (Fig. S8A-C). These results suggest that the proper unfolding and orientation of the graft appears to be crucial for visual function recovery and suggest that the signal observed in the D60-03R animal is not due to interaction between photoreceptors and other cells within the graft.

Using IF, we observed a robust signal of the pre-synaptic marker synaptophysin within the basal portion of the graft in close proximity with the VSX2⁺ bipolar neurons of the host (Figs S7B and S9A). Consistent with the positive mfERG results in the D60-03R eye, we detected colocalization of the pre-synaptic and post-synaptic button markers vGLUT1 and PSD95, respectively, in the outer plexiform layer between the basal side of the graft and the host bipolar neurons (Fig. 6C). Similar observations were made when using the pre-synaptic markers vGLUT1, synapsin, synaptophysin, and CTBP2 (Ribeye) in combination with the post-synaptic markers PKCα, MAP2 and PSD95 (Fig. 6C; Fig. S9A-D). Collectively, these results indicate the potential formation of synaptic connections between the grafted photoreceptors and the host bipolar neurons.

Recruitment of microglial cells (IBA1⁺) was frequently observed near the injection site and around the grafted area, with little evidence of cell proliferation (Fig. S10A-C) (Norte-Muñoz et al., 2022). The innate immune response observed in eyes grafted with grafts from the WTC^Act cell line was more robust compared to those from the WTC^UDC cell line, with significant infiltration of macrophages and IBA1⁺ activated microglia within the grafts (Fig. S10A-D). When compared at the same timepoint (D60), IBA1⁺ microglial cell invasion was observed in 100% of isogenic control grafts (5/5) and in only ∼11% of UDC grafts (1/9). Despite immunosuppression, we also observed complete destruction of 3/8 grafts generated with the WTC^Act cell line (Fig. S10D; Table S2). IBA1⁺/CD4⁺ macrophages (anti-inflammatory M2 healing macrophages) were rare in WTC^Act grafts and absent in WTC^UDC grafts (Fig. S10A). Acquired immune response (CD4⁺ and CD8⁺) was nearly absent at 60 days post-transplantation in WTC^UDC grafts and very low in WTC^Act grafts (Fig. S10A,D) (Anosova et al., 2001; Detrick and Hooks, 2010).

In this study, we have confirmed the capacity of three isogenic human iPSC lines to differentiate into RSs containing nearly all cell types present in the human eye, including RPCs and cone precursor cells. RSs were responsive to bright light stimulation in vitro, as shown using the cGMP degradation assay. Developmental kinetic analysis using scRNA-seq revealed that, although present at DIV5, pluripotent stem cells were absent from DIV30 and DIV60 RSs, as further confirmed by the absence of teratoma formation in transplanted NOD/SCID mice. RS punches were found to be highly enriched for RPCs and photoreceptor precursors minus horizontal, amacrine, RPE, lens and corneal cells, making them more suitable for photoreceptor transplantation therapy. RS punches in suspension cultures could differentiate into mature S- and M/L-cones ex vivo, while sub-retinal transplantation in minipigs revealed their capacity for integration and survival within the damaged host retina.

When compared to the developing human retina, DIV60 RS punches most resemble D-67-D80 human fovea, based on the RNA-seq and histological data (Hoshino et al., 2017). In the human fovea, OTX2 labels the one-nucleus-thick ONL and the much larger INL that also expresses VSX2. A four- to five-nuclei-thick ganglion cell layer is also present and that is negative for both markers (Hoshino et al., 2017). While a near identical OTX2⁺ ONL is also present in the RS punches, the presumptive INL only partially co-expresses OTX2 and VSX2, and the ganglion cell layer is indistinguishable with only ∼2% of RGCs being present (Figs 2A′ and 3B). DIV60 RS punches also share a comparable developmental state with freshly dissected human fetal RSs used in sub-retinal transplantation studies (Radtke et al., 2008; Seiler and Aramant, 2012; Seiler et al., 2017). These studies have consistently demonstrated promising outcomes in both animal models and human clinical trials, characterized by the preservation and restoration of visual responses (Radtke et al., 2008; Seiler and Aramant, 2012; Seiler et al., 2017). Nevertheless, they have not been adopted as standard clinical practice due to ethical constraints associated with the use of human fetal tissue. The pivotal role of synaptic connections between the transplant and host in achieving this visual improvement has been well-documented. Notably, phase II trials in patients with retinitis pigmentosa and age-related macular degeneration have demonstrated substantial enhancements in visual acuity following transplants of fetal retinas along with their RPE (Radtke et al., 2002, 2004, 2008). Intriguingly, it has been observed that immunosuppression may not be requisite for fetal retina transplantations if the intact retinal barrier is preserved during surgery (Radtke et al., 2002, 2004, 2008). Our iPSC-derived RSs, while possessing a similar cellular composition to the fetal retina, present a distinct advantage for macular transplantation because they lack bipolar, amacrine and horizontal cells (Fig. 2E). Additionally, the ease of handling RSs offers the potential for integration with RPE monolayers, providing a promising solution for cases involving concurrent damage to both photoreceptor and RPE cells.

As suggested by previous studies, fully mature cells often display limited integration capacity within host tissue, while excessively immature progenitors may fail to properly mature into a polarized and mature layer, potentially resulting in in vivo cell proliferation (MacLaren et al., 2006; Reh, 2016). Herein, we chose to employ DIV60 RS punches for retinal transplantation. These cells have predominantly lost their proliferative traits and consist of progenitors and precursors largely poised for photoreceptor cell fate. Interestingly, unlike observations from other studies (Gasparini et al., 2022), we did not observe enhanced graft maturation due to donor–host interaction in vivo; instead, grafts followed a developmental pace relatively similar to the ex vivo cultures. The current experimental approach involves a xenogeneic and heterochronic paradigm, as human embryonic tissue is transplanted into the retina of adult minipigs. Achieving full cone differentiation and the development of completely mature OSs in the transplanted animals may necessitate an additional 6 months (Zhong et al., 2014; Capowski et al., 2019; Kruczek and Swaroop, 2020). By extrapolation, prospective patients treated with this regenerative medicine approach may require up to 1 year before observing optimal therapeutic effects.

We built a ‘performance scoring sheet’ of all grafting experiments, which revealed a strong correlation between the presence of a polarized photoreceptor sheet in situ and a functional response at the grafted site (Table S2). The correlation between the structure of the graft and its function helped us define key elements for its successful engraftment. Hence, proper differentiation, polarization and functionality of the newly introduced photoreceptor layer depend on the uniform distribution of the graft as an unfolded retinal tissue. Graft folding resulted in the formation of rosette-like structures, similar to reports from other groups using dissociated photoreceptors and retinal organoids (Assawachananont et al., 2014; Ribeiro et al., 2021; Gasparini et al., 2022; Watari et al., 2023). Growing RSs on a semi-solid biodegradable matrix or a flexible solid support may prevent graft folding during transplantation. The use of matrix or similar supports has proven effective in transplanting pluripotent stem cell-derived RPE monolayers (Diniz et al., 2013; Kashani et al., 2018; Sharma et al., 2019). Coupled with the use of more advanced delivery devices, this approach has the potential to significantly increase transplant success rates (Brant Fernandes et al., 2017).

Although clinical-grade autologous iPSC lines can be created for each patient, the associated expenses and time requirements are considerable. Additionally, patients with inherited retinal degenerative diseases will require genetic correction of the identified mutation before the development of autologous iPSC lines. Consequently, the availability of hypoimmunogenic iPSC lines created worldwide offers promising possibilities for treating patients with retinal degeneration, circumventing the need for autologous iPSC and immune suppression (Torikai et al., 2013; Trounson et al., 2019; Xu et al., 2019; Petrus-Reurer et al., 2020). This work paves the way for future long-term functional investigations using RS grafts and holds great promise for the treatment of macular degeneration and retinal dystrophies.

The surgical procedure to ensure correct graft localization, apicobasal orientation and unfolding needs to be improved. The small number of surgeries answering all of these requirements represents the main limitations of this study.

The isogenic WTC^Act and WTC^Tub (Allen Institute) and WTC^UDC iPSC lines (Healios KK, Japan), and the iPSC#1 and iPSC#2 iPSC lines (Flamier et al., 2018), and the H9 (WiCell) human embryonic stem cell line were dissociated using ReLeSR™ (StemCell Technologies, 05872) and plated on growth factor reduced iMatrix in StemFlex cell media (Gibco, A3349401), supplemented with ROCK inhibitor (Y-27632; 10 nM Cayman Chemical, 10005583). WTC^UDC is a hypoimmunogenic line with inactive α-chain HLA Class 1a and HLA Class II genes expressing the exogenous α-chain HLA Class 1b, PD-L1 and PD-L2 genes (patent number WO2021241658A1). The iPSC#1 and iPSC#2 lines were generated by reprogramming normal human fibroblasts using non-integrative plasmids expressing the Yamanaka factors and a small hairpin-RNA against p53. All cell lines were regularly tested for contamination. To generate retinal sheets, upon confluency cells were differentiated with CI media: DMEM-F12 medium (Invitrogen) containing 1% N2, 2% B27, 10 ng/ml IGF1 (PeproTech, 100-11) and 5 ng/ml bFGF (PeproTech, AF-100-18B), Heparin (Sigma-Aldrich, H3149), and 30 ng/ml COCO (R&D Systems, 3047-CC-050) (Zhou et al., 2015). At DIV60, cultures were observed by phase contrast microscopy using a binocular to identify neural rosettes. To generate retinal organoid-like structures, isolated DIV60 RS punches were cultured in CI medium as suspension cultures in ultra-low adherence plates (Corning, CLS3473). Human pluripotent stem cells were approved by the ‘Comité de Surveillance de la Recherche sur les Cellules Souches’ of the Canadian Institute Health Research (CIHR) and Maisonneuve-Rosemont Hospital Ethics Committee and used by following the CIHR guidelines. All methods were carried out in accordance with relevant guidelines and regulations. Fibroblasts were obtained from Coriell Biorepository, with informed consent from all participants.

Total RNA from two independent biological replicates was extracted using the manufacturer's instructions for the use of Qiagen spin columns and assayed for RNA integrity using QIAxcel RNA kits. cDNA was prepared according to the manufacturer's instructions (New England Biolabs library) and sequenced using the Illumina platform. Base-calling and feature count were carried out using Illumina software. For differential expression analysis, Dseq2 (Love et al., 2014) was used on the R program (https://www.r-project.org/). The samples were analyzed alongside a public dataset of bulk RNA at various stages of the human fetal retina (GSE104827). A heatmap was plotted with the photoreceptor genes to match the maturity of our retinal sheets with the developmental stage of the fetal retina.

Cells were dissociated and prepared using the 10x Genomics single cell gene expression kit aiming for the collection of 10,000 cells. The resulting library was sequenced using the Illumina platform. Base-calling and feature count were carried out using Illumina software. Read count was performed with Cell Ranger software 3.0.0 (Zheng et al., 2017) from 10x Genomics to analyze and map the reads on the HG38 genome. The counts of this experiment were aggregated with the analyzed data from the publicly available Descartes dataset of the eye using the Seurat package (Stuart et al., 2019). The Descartes dataset is a published dataset of the human fetal retina (Cao et al., 2020). Using anchor points between these two datasets, we plotted them on the same 2D space using UMAP (McInnes et al., 2018 preprint) distribution to understand the identity of our cells. Monocle3 (Cao et al., 2019) was used to calculate the pseudo-time (Trapnell et al., 2014) between the DIV5, DIV30 and DIV60 RSs. In summary, the three datasets were combined using Monocle3, then a UMAP (McInnes et al., 2018 preprint) was generated from the combined dataset. On those data, Monocle3 was able to learn a differentiation graph that would allow us to plot the pseudo-time. Loupe 4.1.0 (Hao et al., 2021) was instrumental in visualizing the data and performing an unbiased ontology analysis to identify the cell lineage present in our cultures. In summary, cells were clustered without bias using the k-means strategy. The differentially expressed genes in each group identified by the cell ranger were compared to the literature and their ontology origins to identify the corresponding group.

All experimental procedures were approved by the Integrated Health and Social Services University Network (CIUSSS) animal ethics committee under the project number 2021-2300. Experimental procedures are described in the supplementary Materials and Methods.

In contrast to non-human primates or rodents, pigs display a very strong innate immunity response which must be suppressed by a specific drug regimen. The immunosuppressants used for all groups were started 9 days before transplant (−D9) and continued until euthanasia: Doxycycline 10 mg/kg once a day by mouth (SID PO); Methylprednisolone 5 mg/kg intramuscularly (IM) on −D9 only, followed by administration of Prednisone 5 mg/kg SID PO starting on −D8; Tacrolimus 0.5 mg SID PO; Triamcinolone (Kenalog) 40 mg/ml 2.8 mg/eye intravitreal (IVT) on D0.

Animals received Mydfrin 2.5% Tonopen Trab (TP) and Cyclogil 1% TP approximately 60 min, 45 min and 15 min before the surgery to induce mydriasis. Drops of Mydriacyl 1% were also given before the imaging procedures to optimize mydriasis. Optixcare eye lube was applied to keep the eyes lubricated during the imaging procedures. The animals were anesthetized by administration of butorphanol (0.20 mg/kg), ketamine (10 mg/kg) and dexmedetomidine (0.04 mg/kg) given IM. Upon induction of anesthesia, animals were intubated and supported with mechanical ventilation. A corneal instillation of lidocaine hydrochloride 2% was given in each eye. A Jet-Electrode (Roland Consult) was placed on the eye with an appropriate amount of Optixcare eye lube. OCT was performed using the RETImap animal system (Roland Consult). The region of interest (ROI) was placed in the center and an averaged single scan across ROI was performed. mfERG was recorded following the OCT, using the RETImap system (Roland Consult). Briefly, a Jet-Electrode was placed on the eye, and a reference and ground electrode (Genuine Grass Platinum Subdermal Needle Electrode, Natus) were placed near the ipsilateral outer canthus and on the chin, respectively. A minimum of ten cycles were used for each recording. A low bandpass of 10 Hz and a high bandpass of 300 Hz were used. The mfERG waveform includes an initial depolarization (N1), followed by a repolarization (P1). The N1 wave reflects the initial negative deflection corresponding to cone photoreceptor cell activity. This wave component directly measures the cone function. The P1 wave is the amplified response generated by bipolar and amacrine cells in response to the cone activity (N1), thus reflecting phototransduction by the whole cone system. Interestingly, while N1 can exist in the absence of P1, as it occurs when bipolar cells are destroyed or nonfunctional and photoreceptors are preserved, P1 cannot exist in the absence of N1. Due to its low amplitude, N1 is more challenging to detect and is lost faster during photoreceptor degeneration (Wördehoff et al., 2004; Behbehani et al., 2005; Lai et al., 2007). To visually represent the presence of N1 and P1 waves within damaged and transplanted areas, we conducted a qualitative assessment of mfERG traces. Each hexagon, superimposed onto a fundus microscopic image of the area of interest, was color-coded based on the scores assigned by five assessors in a double-anonymized approach. Scoring criteria included a score of 0 for traces without N1 or P1 waves, 1 for traces with either an N1 or P1 wave, and 2 for traces with both N1 and P1 waves. Assessors were provided with standard traces for reference. The results were merged, averaged and mapped to the corresponding hexagon on the reading, assigning a color code ranging from red (0) to yellow (1) to green (2) to indicate the presence of N1 and P1 waves. For the quantitative evaluation of mfERG, we conducted a detailed analysis of the latency and amplitude of N1 and P1 waves in a carefully selected subset of hexagons within specific ROIs, including healthy, damaged and grafted areas. Approximately five to seven hexagons were quantified for each area/condition to ensure comprehensive coverage and statistical robustness. This selection strategy accommodated the graft coverage, which typically involved around five hexagons. In the healthy and damaged regions, our focus centered on a central hexagon with a precisely defined boundary in the respective region (healthy or damaged). This central hexagon was quantified alongside its six surrounding hexagons, excluding those situated on the border between the healthy and damaged areas. For the graft, quantification involved hexagons visibly overlapping with the graft, allowing for a targeted assessment of the functional response in this particular area.

Thirty days after surgery, Yucatan minipigs were killed by inducing deep anesthesia followed by a lethal injection of pentobarbital (Euthanyl Forte 540 mg/ml 10 ml/50 kg) as a rapid bolus. Both eyes from each treated animal were enucleated, incisions were made through the pars plana, and the globes were immersed in 4% paraformaldehyde in PBS (Sigma-Aldrich) on ice. The anterior chambers were then removed and the vitreous removed, and the eyecups were fixed for another 10-30 min in the same fixative depending on how much vitreous was left. After three washes of 10 min each in 1× PBS, the minipig eyecups were cryoprotected in sucrose solutions prepared in PBS, pH 7.8 (15% sucrose for 1-2 h until sinking, then 30% sucrose for 1 h, and finally 50% sucrose overnight). The minipig eyecups were then snap-frozen in optimal cutting temperature embedding material (Neg50 Frozen Section Media, Thermo Fisher Scientific) in a beaker filled with dry ice. They were then stored at −80°C until sectioned. The cryostat (Leica) was used to produce 10 μm serial sections of minipig eyes.

Retinal cryosections were permeabilized with Triton X-100 for 10 min. Unspecific antigen blocking was performed using 2% bovine serum albumin in PBST for 30 min. Slides were incubated with the primary antibody overnight at 4°C in a humidified chamber (see dilutions in Table S3). Secondary antibodies were incubated for 1 h at room temperature. After incubation with the secondary antibody, slides were washed, counterstained with DAPI and mounted using Vectashield^® Antifade Mounting Medium (H-1000-10) and No. 1.5H coverslips. Pictures were taken using a confocal microscopy system (Olympus). Confocal microscopy analyses were performed using 60× objectives with an IX81 confocal microscope (Olympus), and images were obtained with Fluoview software version 3.1 (Olympus).

Phototransduction activity was assessed by measuring the light-induced hydrolysis of endogenous cGMP using an enzyme immunoassay kit (Biotrack EIA system) according to the manufacturer's instructions (Amersham Bioscience GE Healthcare). Undifferentiated WTC^Act or WTC^UDC iPSC and cells differentiated into RS were kept in the dark. The cells with the ‘light condition’ were exposed to 5 min of bright light before the analysis. The PDE inhibitor IBMX (3-isobutyl-1-methylxanthine, Sigma-Aldrich) was added (1 mM) 72 h before the determination of cGMP levels.

Thanks to Bausch Health Canada and Joe Matar for essential technical support in this project, and to Drs Flavio Rezende and Cynthia Quian for their initial support and involvement in this study.

The peer review history is available online at https://journals.biologists.com/dev/lookup/doi/10.1242/dev.203071.reviewer-comments.pdf