Generating minicorneal organoids from human induced pluripotent stem cells

Cornea is the transparent, avascular tissue on the ocular surface through which light enters the eye. Any damage to its epithelial, stromal or endothelial cell layers can lead to visual impairment. The annular limbus surrounding the cornea harbors adult stem cells that regenerate different parts of the cornea (Schermer et al., 1986; Cotsarelis et al., 1989). Cell replacement therapy using autologous or allogeneic adult limbal grafts has been the standard treatment for patients with severe limbal stem cell deficiency (LSCD) (Rama et al., 2010; Sangwan et al., 2011; Basu et al., 2016). However, in the case of bilateral epithelial defects and for the treatment of conditions affecting the stromal and endothelial cell layers, alternative stem cell sources such as embryonic stem cells (ESCs) and induced pluripotent stem cells (iPSCs) have been explored with a view to generating the various corneal cell types (Ahmad et al., 2007; Shalom-Feuerstein et al., 2012; Hayashi et al., 2012; Sareen et al., 2014; Mikhailova et al., 2014; Chan et al., 2013; Zhang et al., 2014; Chen et al., 2015; McCabe et al., 2015).

A recent report has shown coordinated development of corneal epithelium, neural crest cells, lens epithelium and retinal cells from iPSCs in a two-dimensional (2D) culture system and employed FACS to establish pure cultures of corneal epithelial cells (Hayashi et al., 2016). However, the requirement of rigorous cell enrichment protocols imposes a major hurdle in tissue-specific cell expansion, but can be overcome by establishing three-dimensional (3D) culture systems. This method exploits the inherent self-organizing capacity of differentiating progenitor cell populations, together with the surrounding niche cells, to generate complex tissue structures in vitro. This has been demonstrated successfully with the generation of neuroretinal tissues using PSCs (Eiraku et al., 2011; Gonzalez-Cordero et al., 2013; Assawachananont et al., 2014; Reichman et al., 2014; Zhong et al., 2014; Hiler et al., 2015; Kaewkhaw et al., 2015; Völkner et al., 2016). A recent report has described a method of generating immature corneal organoids from human iPSCs and shown them to express a few corneal markers (Foster et al., 2017). We report here a much simpler and efficient culture method that can generate complex 3D corneal organoids from both human ESCs and iPSCs. We report the establishment of long-term cultures and the characterization of these organoids at different stages of maturation. The mature organoids developed into complex, multilayered, minicornea-like 3D tissues and recapitulated the early developmental events in vitro. We also show that they offer an alternative tissue source for various ocular cell types and report the generation of transplantable sheets of corneal epithelium suitable for regenerative applications.

As pluripotent stem cells are valuable cell sources for generating various ocular cell types and for the study of organ development in vitro, we derived and characterized several human iPSC lines from human dermal fibroblasts as described earlier (Takahashi et al., 2007). As shown in Fig. S1A, the hiPSC-F2-3F1 line formed typical ESC-like colonies both on mouse embryonic fibroblast (MEF) feeders and on Matrigel-coated surfaces. This line expanded well under standard human iPSC culture conditions and the cells were passaged more than 25 times. They remained pluripotent and expressed the stem cell markers OCT4 (POU5F1), SOX2, SSEA4 and alkaline phosphatase (Fig. S1B). When passage 25 cells were transplanted into the subcutaneous space of nude mice, they proliferated and developed into teratomas comprising all three germ layers within 6 weeks (n=8, 6/8 animals developed teratomas) (Fig. S1C). The gross karyotype of this female line was found to be normal at passage 20 (Fig. S1D). Genotype analysis confirmed the presence of integrated copies of the three transgenes OCT4, SOX2 and KLF4, but not cMYC. The endogenous copies of all four genes were active in passage 25 cells and were expressed at levels comparable to that of the human ESC line BJNhem20 (Inamdar et al., 2009; Mariappan et al., 2015) (Fig. S1E).

To induce ocular differentiation, the iPSCs and ESCs were grown to 70-80% confluence under feeder-free conditions and differentiation was initiated in situ as described in Fig. 1A. At 4 weeks of differentiation, distinct clusters of raised, circular to oval-shaped eye field primordial (EFP) clusters (or ʻEFPs') had developed (Fig. S2A). Starting from 1×10⁶ PSCs, an average of 27.33±13.63 EFP clusters could be generated from a well of a 6-well plate (n=6). To confirm that these are EFPs, we manually collected the clusters for total RNA isolation and marker expression analysis by reverse transcription PCR (RT-PCR). As shown in Fig. S2B, the expression of PAX6, OTX2, SIX6 and RX (RAX) confirmed that these 3D clusters consisted of eye field-committed progenitor cells. When the EFPs were allowed to differentiate further in situ, they gave rise to lens epithelial clusters (Fig. S2C) and a SEAM (self-formed ectodermal autonomous multizone) of ocular surface epithelium by 6-8 weeks, as described by Hayashi et al. (2016). The central island of neuroretinal (NR) cells was CHX10 (VSX2)⁺ and RCVRN⁺, and the SEAM of ocular surface ectodermal (OSE) sheets was P63 (TP63)⁺ (Fig. 1Bi,ii). A wave of SOX10⁺ pigmented neural crest cells (NCCs) marked the boundary between NR and OSE cell zones (Fig. 1Biv). Retinal pigmented epithelial (RPE) cells emerged as a compact, non-pigmented epithelium surrounding the NR clusters and later matured to acquire pigmentation. Lentoid clusters expressing gamma-crystallin developed at a precise location adjacent to NR clusters (Fig. 1Bv, Fig. S2C). The NR clusters developed into optic cups and CHX10⁺ precursors self-organized to form the NR layer (Fig. S2D). Pigmented melanocytes were also observed interspersed within the zone of migrating epithelial cells (Fig. 1Bvi, Fig. S2E).

Apart from the emergence of SEAMs, rare EFP clusters developed into 3D, miniature eyeballs, with transparent anterior-segment-like structures on the surface and complex NR structures beneath. A wave of pigmented NCCs set the boundary for the cornea-like structures (Fig. 2Aiii,iv). When the EFPs at 4 weeks were manually collected and cultured under suspension in non-adherent dishes, 40.05±3.89% gave rise to distinct corneal primordial (CP) structures distinct from the generally observed retinal primordia (RP) (n=6; Fig. 2Avi-ix). In the earlier method described by Hayashi et al. (2016), the corneal epithelial cell enrichment was achieved by approximate zoning of cell outgrowths within the SEAM region and by FACS of SSEA4⁺ and ITGB4⁺ cells. However, the method described here enables the self-organization of different CP cells (OSE cells and NCCs) into 3D minicorneal organoids that can directly serve as valuable tissue sources to study corneal development in vitro and also to establish pure cultures of different corneal cell types.

The minicorneas (MCs) ranged from ∼1-7 mm in diameter (Table S3). A magnified view of MCs revealed the presence of a uniform epithelial cell lining (Fig. 2Aix, Movie 1). Transmission electron microscopy images of an 8-week-old MC revealed the presence of a layer of epithelium with tight junctions and numerous apical microvilli, which is a feature of corneal epithelium but not that of lens epithelial cells. The mucin-secreting goblet cells had numerous microvesicles on the apical surface. The middle stromal layer consisted of well-organized collagen fibrils interspersed with stromal cells. A monolayer of flat endothelium-like cells was observed on the inner surface, with a Descemet's-like basement membrane (Fig. 2B). About 10-15 MCs at 8 weeks of differentiation were pooled for the isolation of total RNA, and RT-PCR analysis revealed the expression of the cornea-specific markers PAX6 and P63, but not K12 (KRT12). Variant-specific PCR indicated P63α as the major variant expressed in the developing corneas (Fig. 2C).

Hematoxylin and Eosin (H&E) staining of immunohistochemical (IHC) sections of MCs at 6, 8, 10 and 15 weeks of maturation revealed a stepwise process of tissue layer development and self-organization of cells. The transparent MCs grew from 1 to 4 mm in diameter until 10 weeks of differentiation and developed into opaque structures (Fig. 2Di-iv). The 6-week-old transparent MCs under suspension culture (MC-1) consisted of a double-layered epithelium and a fluid-filled lumen, without any stroma. These structures collapsed immediately after fixation, with leakage of internal fluids (Fig. 2Dv). At ∼8-10 weeks of differentiation, the MCs became strengthened by the subepithelial infiltration of spindle-shaped cells (MC-2), resulting in the development of a thick stromal cell layer, which occupied the entire fluid-filled lumen (Fig. 2Dvi). Surprisingly, IHC examination of intact MCs developing in situ in adherent cultures revealed complex tissue patterns, with orderly layers of different cell types that constitute a normal cornea. The adherent MCs at 10 weeks of differentiation (MC-3) revealed the formation of anterior-segment-like structure, consisting of a thin lid-like structure above a cornea-like tissue (Fig. 2Dvii). At 15 weeks of differentiation the adherent MCs exhibited mature corneal features (MC-4), with well-formed corneal and conjunctiva-like surface epithelia, separated by a limbal crypt-like margin zone (Fig. 2Dviii). Pigmented NCCs marked the boundary between the clear corneal surface and the surrounding epithelium (Fig. 2Dix).

To confirm that the MCs are authentic ocular structures and to understand the spatiotemporal distribution of cells within complex tissues, we carried out IHC examinations on MCs at different stages of development (n=8). At 6 weeks, the fragile MC1-1 comprises a double-layered epithelium, with a fluid-filled lumen (Fig. 3Ai). The epithelial cells were VIM⁺ PAX6⁻ P63⁻, suggesting an undifferentiated primitive state (data not shown). Interestingly, a pair of circular niche-like organizers was observed at the connecting base, and Ki67⁺ proliferating cells emerged from there (Fig. 3Aii). At 8 weeks, the MC1-2 showed significant stromal cell expansion and stratification of surface epithelium. The basal epithelial cells expressed P63α and PAX6 (Fig. 3Bi,ii) and the entire stroma was populated by VIM⁺ cells (Fig. 3Biii). At 10 weeks, the MC-2 developed a thick stratified epithelium, with highly ordered collagen-filled stroma. The epithelial cells expressed P63α, PAX6 and the cornea-specific cytokeratins K3/12 and the stromal cells expressed VIM (Fig. 3C).

At 10 weeks, the MC-3 that developed in situ was strikingly similar to a developing anterior segment, with lid-like structures connected by a periderm-like epithelial lining above the corneal surface, as described elsewhere (Findlater et al., 1993; Huang et al., 2009) (Fig. 4A). The stratified corneal surface epithelium expressed PAX6, P63α and low levels of K12; the stromal cells were VIM⁺ and the endothelium-like cell layer was VIM⁺, CD200⁺ and GPC4⁺ (Fig. 4B, Fig. S3A). Infiltrating αSMA⁺ cells were observed within the anterior stroma (Fig. 4Biv) and the surrounding adnexal cell layers, possibly indicating the development of smooth muscle structures of limbal vasculatures and Schlemm's canal (Fig. S3B). The stromal, endothelial and lid surface epithelial cells were VIM⁺, whereas the stratified corneal epithelial cells were VIM⁻ (Fig. 4Ci,ii). Interestingly, a pars plicata-like ciliary process with a pigmented epithelium was observed at the periphery of MC-3, as reported previously (Kuwahara et al., 2015; Kinoshita et al., 2016). VIM⁺ cell clusters flanked the ciliary processes, suggesting the development of trabecular meshwork and choroid-like structures (Fig. 4Ciii,iv).

The 15-week-old MC-4 was morphologically identical to an adult ocular surface (Fig. 4D), with a distinct limbus-like transition zone separating the PAX6⁺ P63α⁺ K3/12⁺ K10⁻ corneal epithelium on one side and the periodic acid-Schiff (PAS)⁺ and Alcian Blue⁺ goblet cell-enriched epithelium on the other (Fig. 4E, Fig. S3C). The cornea-like structure measured ∼2 mm in diameter (∼1/6th the size of an adult cornea) and expressed most of the cornea-specific markers observed in adult corneal tissues (Fig. S4). The goblet cells were PAX6⁻ P63⁻, which suggested their development from OSE independent of PAX6 and P63 regulation. Surprisingly, except for a few newly emerging cells, the majority of the goblet cells did not express the adult conjunctival goblet cell-specific mucin MUC5AC (Fig. 4Evi). Therefore, we further checked for expression of the other secretory mucin, MUC2. IHC examinations confirmed that the goblet cells were MUC2⁺ (Fig. 4Ev). Interspersed between the goblet cells were a few brightly stained PAX6⁺ and K19⁺ epithelial cells (Fig. 4Ei, Fig. S5Bvii), which suggests the late emergence of conjunctival epithelium and its dependence on PAX6 for development and maturation. Interestingly, a distinct vasculature-like structure with a central lumen and αSMA⁺ cell lining was observed within the stroma of the transition zone. CD34 staining indicated the presence of vascular endothelium-like cells on the inner lining of the lumen, thus confirming the initiation of vascular network development along the conjunctival margin (Fig. S5Avii,viii). Another cluster of spindle-shaped CD34⁺ cells in the peripheral stroma indicated the emergence of a mesenchymal cell wave (Fig. S5Bix).

The frequent detachment of an intact epithelial monolayer from the MC surface suggested that it constitutes a separate cell layer that is possibly embryonic periderm in origin. Mouse skin periderm cells are known to emerge from P63⁺ surface epithelium during the early stratification events and are P63⁻, K17⁺ and K6⁺ (Richardson et al., 2014). During development, the periderm layer plays an important role in preventing pathological cell adhesions between the epithelial linings of adjacent organs, thus ensuring normal tissue formation. Our IHC examinations confirmed that the limbal stem cell marker K15 was exclusively expressed by the basal epithelial cells and that K13 marked the surface and suprabasal epithelium, as reported previously (Ramirez-Miranda et al., 2011; Yoshida et al., 2006). Also, the entire epithelium of MC-2, including the loose surface layers, expressed K13 and the periderm markers K17 and K19 (Fig. 5A). In MC-3, the corneal surface, lid surface and the connecting periderm were lined by K17⁺ and K19⁺ cells (Fig. 5B). When we examined the adult tissues, we observed that the flat wing cells at the corneal epithelial surface retained K13 and K19 expression, while the entire adult ocular surface epithelium was K17⁻ (Fig. S6).

The lid, forniceal and bulbar conjunctiva, limbal and corneal surfaces are lined by a contiguous sheet of epithelium and are distinguished based on minor differences in marker expression and the presence of additional cell types, such as the conjunctival goblet cells. It is well known that the basal cells of the entire epithelial lining express P63 (Fig. S7i). However, the eye-specific PAX6 is expressed only by the corneal and conjunctival epithelial cells (Fig. S7ii). To check if such higher-order cellular organization becomes established in mature corneal organoids, we examined 15-week-old MCs (MC-5) in long-term suspension cultures. As the organoids matured, the lid structures expanded simultaneously and occupied the major volume. The basally positioned corneal structure showed a remarkable cellular organization, with surface epithelium and orderly arranged, compact stromal cells, resembling that of a mature corneal tissue. Similar to MC1-1, a niche-like organizer formed the origin of P63⁺ cells, which appeared to be migrating away from the center in an outward spiraling fashion, as a double-layered epithelium (Fig. 6i). The lid epithelium was stratified and the basal cells expressed P63. However, PAX6 expression was limited to the corneal surface epithelium, with very weak or no expression in epithelial cells along the lid margins (Fig. 6iii). Similarly, K10 expression was restricted to the lid surface epithelium, with negligible expression in the corneal epithelium (Fig. 6iv). The corneal epithelial basal cells were PAX6⁺ P63⁺, whereas the mature suprabasal cells were PAX6⁺ P63⁻ (Fig. 6v). Other adnexal structures, such as the lentoid bodies derived from OSE, could be identified as PAX6^high αA-crystallin⁺ P63⁻ cell clusters (Fig. 6ii,iii). In addition to the lens, the surface epithelial cells also expressed αA-crystallin, as observed in developing mouse eyes (Fig. S7iii).

In 15-week-old adherent MC-4, P63⁺ basal cells were restricted to the corneal side and MUC2⁺ goblet cells were restricted to the conjunctival side, at the transition zone (Fig. 4Eii). In agreement with an earlier report (Richardson et al., 2014), upon epithelial stratification K17 expression became downregulated in the surface periderm cells, and the basal epithelial stem cells were P63⁺ and K17⁺ (Fig. 7Ai) and established a clear transition zone resembling that of a limbal margin. The abundance of Ki67⁺ proliferating cells within the goblet cell-enriched epithelium indicates that the progenitor cell proliferation, differentiation and tissue expansion proceed from the transition zone (Fig. 7Aii). Dual staining for MUC2 and P63 or K3/12 expression further confirmed the presence of a transition zone (Fig. 7Aiii,iv). The surface epithelium formed a collagen IV-enriched basement membrane, while the VIM⁺ stromal cells laid out a well-organized collagen I-enriched extracellular matrix (Fig. 7Av,vi). When the expression patterns of other epithelial keratins were examined, we observed that K13⁺, K15⁺ and K17⁺ cells were restricted to the limbal margin. Within the stratified corneal epithelium, K13 marked the surface and suprabasal cells, while K15 and K17 marked the basal and suprabasal cells (Fig. 7B). However, all of the surface epithelial cells expressed K19, with intense staining in the basal cells, surface periderm-like cells and in a few developing conjunctival epithelial cells, which suggested the late emergence of conjunctival epithelium.

When the cell outgrowths from EFPs were analyzed in adherent 2D cultures, we found that SSEA4⁺ primordial cells tended to organize into ruffled structures resembling limbal crypts and gave rise to P63⁺ and PAX6⁺ ocular surface epithelial cells (Fig. 7Ci-iii, Movies 2 and 3) and waves of NES⁺, SOX10⁺ and PAX6^low NCCs. NES⁺ cells are a component of the native limbal niche (Fig. S4x) and we show that iPSC-derived PAX6⁺ epithelial outgrowths are lined by NES⁺ NCCs that co-express PAX6, as reported previously (Mariappan et al., 2014) (Fig. 7Civ). Patches of PAX6^low neural crest-like cells differentiated into a corneal endothelium-like phenotype by downregulating PAX6 (Fig. 7Di,ii); they appeared as hexagonal, compactly arranged, non-pigmented flat cells with ZO-1 (TJP1)⁺ tight junctions (Fig. 7Diii,iv). The migratory NCCs were found to be SOX10⁺ P63⁻ cells (Fig. 7Dv).

Explant cultures of 8- to 10-week-old MCs on glass coverslips resulted in a spiraling wave of P63⁻ K17⁺ OSE cells at the leading edge, followed by a compactly arranged monolayer of P63⁺ K17⁺ corneal epithelial cells (Fig. S8). In an attempt to generate transplantable sheets of corneal epithelium, we established explant cultures on denuded human amniotic membrane (hAM) substrates, using mature cornea-like organoids at 8-10 weeks of maturation. IHC examination of 10-day-old cultures confirmed that the uniform sheets of PAX6⁺ K12⁺ epithelium generated using corneal organoids were comparable to those generated using adult limbal explants. The basal epithelial cells were PAX6⁺ P63⁺ P75 (NGFR)⁺ and also expressed K17, K19 and VIM (Fig. 8). Interestingly, the resting corneal epithelium was found to be VIM⁻ K17⁻ (Figs 4 and 6). However, the actively proliferating cells in both the limbal and organoid explant cultures were VIM⁺ K17⁺, suggesting a possible activation of these markers during acute regeneration and wound healing responses.

An autologous iPSC-derived corneal cell source will offer a promising alternative for the treatment of patients with bilateral LSCD. A few earlier reports have demonstrated the possibility of deriving PAX6⁺ P63⁺ K3/12⁺ corneal epithelial cells from PSCs in 2D cultures. Here, we report for the first time an efficient method of generating complex 3D corneal organoids using iPSCs, which circumvents the need for complicated cell enrichment procedures as are involved in establishing limbal cultures.

The adoption of a simple differentiation protocol by the direct shifting of growing cultures to retinal differentiation conditions in the absence of noggin has resulted in successful induction of EFP clusters. We also emphasize that it is crucial to excise EFPs and initiate suspension cultures at 4 weeks of differentiation, before the commencement of surface ectodermal cell and NCC migration, in order to ensure successful induction of corneal organoids in ∼40% of the EFPs by 6 weeks. We believe that the inhibition of migration of proliferating progenitor cells away from EFPs in suspension culture enables the autonomous self-assembly of various cell types, resulting in the generation of complex 3D corneal organoids.

The bilayered epithelium of the newly emerged MCs was derived from the primitive periderm-like P63⁻ PAX6⁻ VIM⁺ OSE cells. The presence of internal fluid appears to help in establishing a circular and convex shape for the developing corneas. The subsequent wave of VIM⁺ stromal cells and the deposition of collagen matrix helped to strengthen the outer scaffold of OSE cells. The developing lid-like structure in MC-3 was lined and connected by a continuous periderm-like epithelium above the corneal surface. This observation is in agreement with the fact that the developing eyelids fuse and form a continuous covering over the developing cornea. The connecting periderm disintegrates and enables lid separation and eye opening during advanced stages of embryonic development in humans and at postnatal stages in rodents (Findlater et al., 1993; Huang et al., 2009). Unlike the developing skin periderm that is shed after birth (Richardson et al., 2014), the presence of K13⁺ K17⁺ K19⁺ periderm-like surface epithelium in developing MCs and in adult corneas suggests their probable role in normal ocular surface development and in adult tissue homeostasis. We hypothesize that this unique surface lining may help in preventing abnormal cell fusions between the corneal and lid surface epithelium during embryonic eye development and in wound repair processes during adult tissue regeneration.

The presence of ciliary margin zone (CMZ)-like pigmented and ruffled epithelium, flanked by VIM⁺ structures, at the corneal periphery prompted us to speculate that CMZ development might precede or coincide with ocular surface periderm formation (∼5-6 weeks). The secretions of the CMZ cells might contribute to setting the initial corneal shape, which becomes further strengthened by the infiltration of VIM⁺ NCCs. The NCCs also contributed to the formation of a monolayer of VIM⁺ CD200⁺ GPC4⁺ endothelium-like cells beneath the thick stroma, thus resulting in the generation of a complete anterior-segment-like structure.

As the MCs matured, the lid and the limbal margins became established by the spatiotemporal pattern of expression of P63, PAX6 and keratins. The P63⁺ cells were restricted to the corneal and limbal basal epithelial cells, thus establishing a sharp boundary between the cornea and the future conjunctiva. Whereas the entire surface epithelium expressed PAX6 at low levels, well-differentiated central corneal cells and a subset of cells within the conjunctival region were brightly PAX6⁺, confirming its key role in corneal maturation and the emergence of conjunctival epithelium. The majority of the goblet cells on the conjunctival side were MUC2⁺ P63⁻ PAX6⁻, which suggests that the goblet cells emerge from the primitive OSE cells independently of P63 and PAX6 expression. However, the adult conjunctival goblet cells expressed very low levels of MUC2 (McKenzie et al., 2000) and were predominantly MUC5AC⁺ (Fig. S4Av). An earlier report has confirmed that goblet cell development is normal in Muc5ac^−/−; Muc5b^−/− mice (Marko et al., 2014). Taken together, we believe that MUC2 and MUC5AC are the developing and mature conjunctival goblet cell markers, respectively. The presence of niche-like organizing structures consisting of Ki67⁺ P63⁺ cells suggests that the tissue growth and expansion proceeds from such transition zones. Further anatomical maturation of corneal tissue was mediated by the infiltration of CD34⁺ mesenchymal stem cells (Sidney et al., 2014) and other neural crest-derived cell types, such as the smooth muscle cells, which contributes to the formation of limbal and episcleral vasculatures.

Earlier evidence has confirmed the roles of PAX6 in regulating NCC migration and their differentiation into ocular cell types (Baulmann et al., 2002; Kanakubo et al., 2006) and the involvement of NCC-dependent signaling in feedback regulation on PAX6 (Grocott et al., 2011). Our observations indicate that the PAX6^low NCCs differentiated into flat, non-pigmented, endothelium-like hexagonal cells by downregulating PAX6 expression. Explant cultures of 8- to 10-week-old MCs on hAM has enabled the generation of transplantable sheets of PAX6⁺ P63⁺ K12⁺ corneal epithelial sheets, similar to adult limbal tissue-derived grafts intended for regenerative applications. We further plan to use these tissue grafts in xenotransplantation studies in rabbit LSCD models, to test their clinical suitability in corneal surface reconstruction procedures.

In summary, we show for the first time that complex 3D corneal organoids can be generated from iPSCs and that the MCs undergo maturation in vitro and recapitulate the steps of normal corneal development, as depicted in Fig. 9. The availability of such MCs at 10 weeks of maturation circumvents the need for complicated cell enrichment protocols and offers a simpler method of establishing enriched cultures of corneal epithelial cell sheets for basic research needs and for regenerative applications.

This study was approved by our Institutional Review Board (IRB) of the LV Prasad Eye Institute, Hyderabad, India. All research involving human samples followed the tenets of the Declaration of Helsinki. Experiments involving animals were conducted in adherence to the ARVO statement for use of animals and with the approval of the Institutional Animal Ethics Committee (AEC) of the National Institute of Nutrition, Hyderabad, India.

Full-thickness punch biopsies of skin were taken from volunteers with their informed consent. The biopsies were used to establish human dermal fibroblast (HDFs) cultures. A retroviral cocktail containing individual vectors expressing the OCT4, SOX2, KLF4 and cMYC (OSKM) transgenes were used to transduce passage 3 HDFs at an MOI of ∼2. The cells were then split and cultured under standard human ESC culture conditions. The reprogrammed clones that emerged after 3 weeks were manually picked based on colony morphology and five clones were passaged for further expansion. The clones were also adapted to feeder-free culture conditions on Matrigel (Corning) coated plates using the mTeSR™1 kit, as per manufacturer's instructions (STEMCELL Technologies). The reprogramming efficiency was 0.005% and the clone hiPSC-F2-3F1 was expanded beyond 25 passages and characterized for stemness and pluripotency.

Growing cultures of the human ESC line BJNhem20 and the normal human iPSC line hiPSC-F2-3F1 were differentiated towards eye field commitment as described below. When the cultures reached 70-80% confluence, the growth medium was replaced with differentiation medium [DM: DMEM/F12, 4% knockout serum replacement (KOSR), 4% fetal bovine serum (FBS), 1× non-essential amino acids (NEAA), 1× Glutamax, 1× Pen-Strep; Thermo Fisher Scientific] to induce spontaneous differentiation for 2 days. Subsequently, the cultures were shifted to retinal differentiation medium (RDM: DM plus 2% B27) and maintained for 1 month to induce eye field specification. Noggin was omitted from the RDM cocktail in order that uninhibited TGFβ and BMP intrinsic signals could direct OSE development. The distinct EFP clusters that emerged at 4 weeks were either continued as adherent cultures in situ or excised manually for suspension cultures as described below.

The EFPs were further continued in situ as adherent cultures in RDM for another 4 weeks to allow whole eyeball-like structure development, with transparent CP on the surface and NR cup on the basal side. These cultures were maintained in corneal differentiation medium (CDM: DM plus 1% N2, 5 μg/ml insulin, 5 ng/ml FGF, 10 ng/ml EGF; Thermo Fisher Scientific) for a further 6-8 weeks to enable maturation of the ocular surface structures. However, a majority of the EFPs gave rise to concentric cell outgrowths, as described previously (Hayashi et al., 2016). Alternatively, the EFPs were manually scooped out intact and cultured in RDM for 4 weeks in non-adherent plates. It is crucial to excise the EFPs at 4 weeks before the commencement of initial waves of surface ectodermal cell and NCC migration. Within 2 weeks of suspension culture, distinct RP and transparent, bubble-like CP structures emerged from the floating EFPs. At 6-8 weeks of differentiation, the delicate CP structures were dissected out of floating EFPs and cultured separately for a further 8-10 weeks in CDM for tissue maturation, as depicted in Fig. 1A. Alternatively, the MCs were processed directly for explant cultures or RNA isolation or fixed in 10% formalin for IHC examination.

MCs at different stages of maturation (6-10 weeks) were taken and the basal stalk that carries the niche-like organizer along with the adjoining epithelium was chopped out under a microscope and cut into fine pieces in a few drops of CDM. The tissue explants were picked using a needle and explanted on to the surface of de-epithelialized human amniotic membrane (hAM). Alternatively, the explants were placed on Matrigel-coated glass coverslips. The cultures were maintained in CDM and incubated at 37°C with 5% CO₂. The epithelial cells migrated out of the explants and formed growth zones that merged with each other to form uniform epithelial sheets within 10 days. The cell sheets were fixed and processed for IHC examination as described below.

Genomic DNA and total RNA were isolated from cell samples using standard procedures. cDNAs were prepared by reverse transcription using the SuperScript II reverse transcriptase kit (Invitrogen, Life Technologies). PCRs were performed using either genomic DNA or cDNA as the reaction template (n=3). Template concentrations were normalized based on eEF1α (EEF1A1) expression. Table S1 summarizes the primers used. The amplicons were resolved on 1% (w/v) agarose gels, stained with ethidium bromide and imaged using the Gel Doc XR+ system (Bio-Rad).

The MCs were fixed in 10% formalin and paraffin embedded for further sectioning. Thin (4 μm) sections were processed for evaluation by H&E, PAS and Alcian Blue staining by standard procedures. For IHC examination, antigen retrieval was achieved by heating at 100°C with sodium citrate buffer (pH 6.2) and the slides were processed for blocking and antibody incubations. DAB staining of samples was performed as per the manufacturer's instructions (Super Sensitive One-Step Polymer-HRP IHC detection system, Biogenex), with counterstaining with Hematoxylin, propidium iodide (PI) or DAPI (1 μg/ml each). Table S2 summarizes the antibodies used, including dilutions. Alkaline phosphatase staining was undertaken as per manufacturer's instruction (Chemicon, Millipore). The samples were finally mounted with DPX (SD Fine Chemicals) or glycerol and imaged using an epifluorescence (IX71, Olympus) or confocal (LSM 510, Carl Zeiss) microscope. The images were analyzed using ImagePro Express (Media Cybernetics) and LSM 510 Meta version 3.2 (Carl Zeiss) software, respectively, and the composites were prepared using Adobe Photoshop CS.

iPSCs at passage 25 were suspended in 20% Matrigel in DMEM/F12 and kept on ice. About 1×10⁶ cells in 200 μl were aspirated into tuberculin syringes fitted with a 26 G needle and injected into the subcutaneous space above the rear right haunch of 6-week-old nude mice (n=8). Teratomas that developed at 6-8 weeks post-injection were surgically dissected after euthanizing the animals. The tissues were fixed overnight in 4% paraformaldehyde and processed for paraffin embedding. The tissue blocks were sectioned and processed for IHC examination as described above.

The cells at passage 8 and 20 were grown under standard iPSC culture conditions. About 70-80% confluent cultures were treated with colcemid (0.1 μg/ml; Sigma-Aldrich) for 2-3 h to induce metaphase arrest and trypsinized to prepare single-cell suspensions. The cells were further treated with a hypotonic solution, fixed and then dropped onto clean glass slides (Fisher Scientific) and air dried. After a brief trypsin treatment, the chromosomes were G-banded by Giemsa staining. Well-spread metaphases were imaged and analyzed using CytoVision automated (Applied Imaging).

Tissues were fixed in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.2) for 24 h at 4°C and then washed with 1× PBS thoroughly and post-fixed in 1% aqueous osmium tetroxide for 2 h. The samples were then washed, dehydrated through a graded alcohol series, embedded in Spurr's resin and incubated at 80°C for 72 h for complete polymerization. Ultra thin (60 nm) sections were prepared using an ultramicrotome (Leica Ultra Cut UCT-GA-D/E-1/00), mounted on copper grids and stained with saturated aqueous uranyl acetate and counterstained with Reynolds lead citrate. The sections were viewed using a Hitachi H-7500.

The mean values of experimental repeats are given as ±s.d.

We thank Dr M. Lakshman of the RUSKA Lab, College of Veterinary Sciences, Prof. Jayshankar Telangana State Agricultural University (PJTSAU), Rajendranagar, Hyderabad, India, for providing TEM support; Dr Lakshmi Rao Kandukuri, Clinical Research Facility-Medical Biotechnology (CRF-MB), Centre for Cellular and Molecular Biology (CCMB), Hyderabad, India, for karyotyping; Dr Suresh Pothani, National Centre for Laboratory Animal Sciences, National Institute of Nutrition, Hyderabad, India, for providing animal facility support; and Prof. Dorairajan Balasubramanian, Dr S. Shivaji and Dr Gullapalli Nageshwara Rao for their critical review and useful suggestions in finalizing the manuscript.