Retinal Organoids from Pluripotent Stem Cells Efficiently Recapitulate Retinogenesis

The plasticity of pluripotent stem cells provides new possibilities for studying development, degeneration, and regeneration. Protocols for the differentiation of retinal organoids from embryonic stem cells have been developed, which either recapitulate complete eyecup morphogenesis or maximize photoreceptor genesis.

Here, they have developed a protocol for the efficient generation of large, 3D-stratified retinal organoids that does not require evagination of optic-vesicle-like structures, which so far limited the organoid yield.

Analysis of gene expression in individual organoids, cell birthdating, and interorganoid variation indicate efficient, reproducible, and temporally regulated retinogenesis. Comparative analysis of a transgenic reporter for PAX6, a master regulator of retinogenesis, shows expression in similar cell types in mouse in vivo, and in mouse and human retinal organoids. Early or late Notch signaling inhibition forces cell differentiation, generating organoids enriched with cone or rod photoreceptors, respectively, demonstrating the power of our improved organoid system for future research in stem cell biology and regenerative medicine.

Graphical Abstract

Pluripotent embryonic stem cells (PSCs) facilitate research on mammalian neuronal development, neurodegenerative disorders, and regenerative therapies. It has been shown in the retina that developmental processes such as optic-vesicle (OV) and optic-cup (OC) morphogenesis and signaling cascades can be reproduced using mouse and human embryonic stem cells (mESCs and hESCs). Retinal organoid and 2D culture approaches have been used for cell replacement therapy studies because efficient derivation of sufficient numbers of integration-competent cells remains a major limitation for regenerative medicine. The first reports on cell-based disease-modeling approaches, retinal neuronal morphogenesis , and function in organoids are promising. Yet, in this evolving field, benefits and limitations have not been fully explored and many questions remain.

Description of the Retinal Organoidogenesis Trisection Protocol and the Generation of Wild-Type and hPAX6GFP mESC-Derived Retinal Organoids

(A and B) Representative phase contrast images of entire mESC-derived aggregates (A) and transgenic hPAX6GFP expression in organoid cryosections (B) at different days (D) of our trisection protocol. Aggregates are shown before (D1–10) and after (D14–18) organoid trisection.

(C) Schematic overview and timeline of the trisection protocol. mESC-derived starting (mother) aggregates are formed with 100% efficiency, out of which eyefield induction occurs in 81%, so that trisection increases the yield of retinal organoids to 183% (N ≥ 4, n ≥ 20). Magenta indicates eyefield tissue on D7–10 and retina thereafter. KOSR, knockout serum replacement; N2, N-2 supplement.

For example, the question of efficient generation of large, stratified, retinal tissues has not been addressed. Sasai and colleagues pioneered a protocol that allows the self-organization of eyecup-like structures. This entails a series of complex tissue interactions, such as eyefield evagination and subsequent invagination, resulting in neural retina opposed by retinal pigment epithelium (RPE). However, this protocol relies on the evagination of the neuroepithelium and its live visualization, preferably using transgenic RAX (retina and anterior neural fold homeobox) reporter gene expression, for reliable manual isolation of the prospective retinal organoids. RAX is part of a group of transcription factors sufficient and necessary for the specification of the eyefield, which gives rise to the eye primordia and the retina. Although eyefield formation has been shown to be efficient in mouse PSC lines, the yield of retinal organoids depends on and is highly limited by a low frequency of neuroepithelial evagination. Others have adapted protocols to maximize and simplify rod photoreceptor production by omitting the evagination dissection step. This results in larger organoids, with retinal and non-retinal structures intertwined within the starting organoid, and comes at the expense of inner-retina cell types. Therefore, they speculated that unbiased neuroepithelium trisection at the eyefield stage overcomes these limitations and enables production of more numerous retinal organoids.

Low Frequency in Potential Optic-Vesicle Formation, but Not Eyefield Induction, Limits Retinal Organoidogenesis Efficiency

(A) RAX and PAX6 co-expression, indicating eyefield identity, was detected at day (D) 7.

(B) At D10, large neural retinal areas (RAX+ VSX2+) with adjacent RPE areas (MITF+ RAX− VSX2−) were observed. Dashed square indicates region-of-interest shown at higher magnification: RAX+ (b′) and MITF+VSX2+ (b″).

(C) Quantitative analysis of eyefield induction: the number of RAX+ aggregates increases over time, based on scoring immunostained aggregate sections at D2–10 (N ≥ 3, n ≥ 10; 3,000 mESCs seeding density).

(D–F) Eyefield induction efficiency was dependent on cell-seeding density and did not correlate with evaginations. (D) Phase-contrast images and immunostaining images for eyefield transcription factors RAX and LHX2 of aggregates developing from different cell-seeding densities at D1 and D7 as indicated. (E) Quantification of aggregates with evaginations and (F) RAX+ aggregates, developed from different cell-seeding densities (N = 4, n = 24).

(G) Scheme of eyefield and optic-vesicle formation efficiency (N = 7, n ≥ 10). mESC-derived starting (mother) aggregates are formed with 100% efficiency, out of which 81% develop eyefields and 28% form optic vesicles. Thus, according to the first protocol, aggregate evagination and optic-vesicle formation limit the efficiency of retinogenesis, and the majority of eyefield-containing tissues are frequently discarded.

Another question is the heterogeneity within and between organoids, which seems common to all the protocols developed so far but has not yet been studied in detail. Several processes, such as progenitor proliferation, cell differentiation, and ontogenetic cell death, could be potential sources of organoid variation.

Transgenic animals with fluorescently labeled cells have been instrumental in visualizing major processes in the developing and adult retina. However, it is unknown whether reporter expression is comparable between retinal organoids and in vivo. Thus, they investigated PAX6 transgenic reporter expression to gain an insight into retinal organoidogenesis.

Generation of Rod- or Cone-Photoreceptor-Enriched Organoids by Timed Inhibition of Notch Signaling

(A and B) Notch inhibition by DAPT at early time points (D12–14) increased the number of cone photoreceptors. (A) Images of immunostained organoid sections and (B) graphs with quantification of cone marker expressing cells (TRBETA2+, S-OPSIN+) in DAPT-treated and control (DMSO) organoids (N = 3, n = 5).

(C and D) Notch inhibition at later time points (D16–18) resulted in rod-enriched organoids. (C) Quantification and (D) images of cell-type-specific markers (BRN3+, ganglion cells; CRX+, photoreceptors; ELAVL3/4+, ganglion, amacrine, horizontal cells; GLUL+, Müller glia; VSX2+, progenitors, bipolars) in DAPT-treated and control (DMSO) organoids (N = 3, n = 5). hPAX6GFP-expressing cells were greatly reduced by Notch inhibition. Top panels of (B) and (C) show schemes of the experimental paradigms.

PAX6 is a highly conserved master regulator of neurogenesis, playing several roles in eye and retinal development, e.g., eyefield specification, stemness control, and cell-fate specification. PAX6 reporter expression might also provide an insight into the formation of retinal structure because it remains expressed in postmitotic horizontal and amacrine cells, whose synaptic processes are part of the outer and inner plexiform layer, respectively.

Efficient Retinal Histogenesis from Trisected Day-10 Aggregates

(A and B) Organoid development followed a temporal program, as evidenced by (A) mitotic marker phosphohistone H3 (PHH3) and (B) expression of the neurogenic marker ASCL1 (N = 3, n = 5).

(C) Developing large epithelial structures express markers for retinal progenitors (VSX2) and photoreceptors (CRX), indicating retinal identity. Photoreceptors were localized at the outer (apical) side of the aggregate. VSX2 is also expressed in postmitotic bipolars. Scoring analysis of (retinal) aggregates immunostained for VSX2 and CRX (N ≥ 4, n > 20).

(D) The trisection approach increased the total yield of retinal organoids at D21. Each starting aggregate was trisected at D10 into three evenly sized portions, out of which 183% ± 44% developed into retinal organoids, the minority developed as non-retinal organoids (23%), and the remainder degraded by D21.

(E) Cell birthdating showed a defined timing of retinogenesis. The scheme shows 2-hr EdU pulses (open triangle) analyzed after a 2-day chase period (closed triangle). Graphs represent quantitative analysis of birthdated cells (N = 3, n = 3/N). Representative images of birthdated ELAVL3/4+ (amacrine, horizontal, ganglion cells), BRN3+ (ganglion cells), VSX2 (progenitors, bipolar cells), and CRX+ (photoreceptors) cells.

Here, they have developed a protocol to facilitate efficient organoidogenesis of large, complex, 3D retinas derived from wild-type mESCs without requiring the formation and isolation of OV/OC-like structures. Gene-expression profile analyses of individual organoids and retinal cell birthdating experiments indicate efficient, reproducible, and temporally regulated retinogenesis. They have established retinal organoidogenesis from mESC and hESC lines carrying a human PAX6 transgenic GFP reporter, hPAX6GFP BAC, and respective transgenic mice to assess GFP-expressing cells in a comparative approach. Their results suggest that their protocol is a valuable addition to the existing organoid technologies, and will facilitate future retina research and regenerative medicine1.

hPAX6GFP in mESC- and hESC-Derived Retinal Organoids

(A and B) Analysis of transgenic hPAX6GFP expression in (A) mESC-derived and (B) hESC-derived retinal organoids. (A) At day (D) 18, weakly GFP+ cells in the neuroblastic layer co-expressed the retinal progenitor markers RAX and PAX6. At D21, GFP+ cells had accumulated in the inner nuclear layer (INL)-like region and co-localized with ELAVL3/4 and PAX6, indicating amacrine and horizontal cells. GFP+ cells are not co-labeled with photoreceptor, bipolar (OTX2, RCVRN), or ganglion cell (BRN3+) markers. GFP+ cells co-expressed the pan-amacrine marker TFAP2a and bHLHB5, expressed by GABAergic amacrines. (B) Overview and ROI images of immunostained human organoid sections at D41: GFP was detected in PAX6+, RAX+, VSX2+, and ELAVL3/4+ cells. Just as in mouse organoids (A), ganglion cells (BRN3) were mostly GFP negative.

Characterization of Transgenic hPAX6GFP Expression in the Developing and Mature Mouse Retina

(A) Images of GFP and PAX6 immunostained hPAX6GFP reporter mice retina sections at embryonic days (e) 12 and 15, and postnatal days (p) 0 and 6 show GFP in progenitors and subsets of PAX6+ postmitotic horizontal and amacrine neurons.

(B) Similar to the retinal organoids, GFP+ cells in the mouse retina also co-expressed the markers CALB1 (amacrine, horizontal cells) and ELAVL3/4 (amacrine, horizontal, ganglion cells), but not the ganglion cell marker BRN3. Subsets of GFP+ cells also co-expressed PROX1 and bHLHB5 (amacrines).

(C and D) Summary scheme (C) and comparison table (D) show that transgenic hPAX6GFP expression showed similar retinal cell types in the retinal organoid system and mouse retina in vivo. GFP expressed (+)/not expressed (−); RPC, retinal progenitor cell; RGC, retinal ganglion cell; HC, horizontal cell; AC, amacrine cell; BP, bipolar cell; MG, Müller glia; NE, neuroepithelium.

  1. Völkner, M.. Retinal Organoids from Pluripotent Stem Cells Efficiently Recapitulate Retinogenesis. Stem Cell Reports6, 525–538 (2016).