The telencephalon and eye in mammals are originated from adjacent fields at the anterior neuroepithelium ¹. Morphogenesis of these embryonic fields leads to the formation of telencephalon, optic stalk, optic disc, and neuroretina along an axis. Early differentiated retinal ganglion cells (RGCs) in the neuroretina grow axons toward the optic disc and then along the optic stalk to reach the brain. How these telencephalic and ocular tissues are specified coordinately to ensure directional RGC axon growth is unclear in humans.

Cysts are hollow spheres of an columnar epithelium induced from human pluripotent stem cells through embedding hESC sheets in Matrigel; they are used to generate retinal cells ^{28, 33, 47}. Nevertheless, developmental potentials of cysts have not been characterized.

Using the cysts, here we generate a telencephalon-eye organoid that is composed of concentric zones of anterior ectodermal progenitors (CONCEPT) from cysts (Fig. 1A, B). Cysts were generated as previously described ^{28, 33}. The epithelial structure of cysts was demonstrated by apically localized TJP1::GFP reporter at the lumen (Fig. 1C), consistent with previous findings ^{28, 47}. To assess developmental potentials of cysts, individual cysts were manually picked and then seeded onto the Matrigel-coated surface at low densities (Fig. 1A). After attaching to the culture surface, cysts initially grew as dome-shaped individual colonies. Subsequent culture of the colonies in a KSR medium (see Materials and Methods) led to the self-formation of concentric zones of anterior ectodermal progenitors: an elevated central zone surrounded by multiple zones; this morphology was distinguishable under a steoromicroscope (Fig. 1D) or an inverted microscope. CONCEPT structures were highly reproducible, which is demonstrated by consistent gene expression profiles of multiple CONCEPT organoids in whole culture wells (Fig. S1). If multiple cysts were fused together or cysts for seeding were too big or small, CONCEPT structures were affected or incomplete. CONCEPT structures expressed gene markers for telencephalon and eye in concentric patterns. In mice, Foxg1 is specifically expressed in the E10 telencephalon at high levels; it was also expressed in the optic stalk, invaginating optic cup, and lens at lower levels, with the rostral optic stalk connecting the telencephalic vesicle to the invaginating optic cup (Fig. 1E) ⁶; Vsx2 is specifically expressed in the E10.5 neuroretina (Fig. 1F); Pax6 is specifically expressed in the E10.5 neuroretina, RPE, lens vesicles, and surface ectoderm (Fig. 1G) ¹³. In CONCEPT structures, the expression of FOXG1, VSX2, and PAX6 generally exhibited concentric patterns spanning from the center to the periphery (Fig. 1H-O). Similar to FOXG1, telencephalon marker EMX2 ⁴⁸ was also expressed at the central region (Fig. S1F). PAX6 was expressed in multiple concentric zones at distinct levels, mimicking its expression patterns in multiple ocular tissues in E10.5 mouse eyes (Fig. 1G, J, K, N, O). Therefore, the CONCEPT structure is a telencephalon-eye organoid mimicking telencephalic and ocular tissues in the aspects of cell types and patterning at early stages.

Morphogens FGFs and BMPs play crucial roles in patterning the forebrain and eye in vivo. In mice, Fgf8 is specifically expressed at the rostral forebrain at early stages, induces Foxg1 expression ¹⁹, and regulates telencephalic patterning in a dose-dependent manner ²⁰. Bmp4 and Bmp7 are expressed in the dorsomedial telencephalon, optic vesicles, and presumptive lens placodes ^49–52. Bmp4 is required for lens induction ⁵⁰, and Bmp7 is required for proper patterning of the optic fissure ⁵³. In CONCEPT telencephalon-eye organoids, FGF8, BMP4, and BMP7 were expressed in attached cell colonies starting from early stages (Fig. 1P, S, T) and subsequently exhibited circular patterns (Fig. 1R, V, W). At day 10, the expression of BMP4 and BMP7 delineated multiple rings, with a big ring mostly at the center surrounded by numerous small rings (Fig. 1P, S). These observations suggest that the attachment of a single-lumen cyst to the culture surface caused differences in cell behaviors: some cells separated from the original cyst and migrated peripherally; some of the separated cells formed small ring-like structures; the cells that remained at the center formed a big ring-like structure. At day 17, circular patterns of BMP4 and FGF8 emerged (Fig. 1Q, U). At day 25, BMP4, FGF8, and BMP7 clearly exhibited circular patterns (Fig. 1R, V, W). Hence, FGFs and BMPs spontaneously form circular gradients, which would dictate coordinated cell specification in CONCEPT telencephalon-eye organoids.

To assess cell differentiation in CONCEPT telencephalon-eye organoids, we examined marker expression for RGCs, the first type of cells that differentiate in the neuroretina. In mice, transcription factor Pou4f2 is expressed in the early differentiated RGCs and required for the development of a large set of RGCs ⁵⁴. Tubb3 is expressed in the somas and axons of differentiating RGCs.

In CONCEPT organoids, POU4F2 and TUBB3 were detectable as early as day 17 and increased to higher levels at day 22. RGC somas were marked by POU4F2 and TUBB3 co- expression; RGC axons were marked by TUBB3 expression. Interestingly, RGCs grew directional long axons that followed a circular path (Fig. 2A, B). The circular path of RGC axon outgrowth became more evident in organoids at day 26 (Fig. 2C, D). In mice, the initially differentiated RGCs are adjacent to the optic disc; their axons grow towards the optic disc to exit the eye and then navigate within the optic stalk to reach their targets in the brain; the optic disc and stalk specifically express Pax2 at distinct levels (Fig. 2E-G) ^{17, 55, 56}; presumptive optic disc cells expressed VSX2 whereas optic stalk cells do not ¹³ (Fig. 1F; 2E). In CONCEPT organoids, there were two PAX2+ cell populations that formed two adjacent rings (Fig. 2H-L). POU4F2+ RGCs grew TUBB3+ axons toward and then navigated along the adjacent PAX2+ VSX2+ cell population (Fig. 2H-L), mimicking axon growth from the initial RGCs toward the PAX2+ optic disc in vivo. Meanwhile, PAX2+ VSX2- cell population at the inner zone set up an inner boundary of the path for RGC axon growth, mimicking PAX2+ optic stalk that spatially confines RGC axon growth in vivo. Based on these findings, we designate PAX2+ VSX2+ cells and PAX2+ VSX2- cells as the optic disc and optic stalk, respectively. Taken together, our findings demonstrate that RGCs grow directional axons toward and then along a path defined by PAX2+ cell populations in CONCEPT telencephalon-eye organoids.

CONCEPT telencephalon-eye organoids at stages around day 25 contained transparent structures reminiscent of the ocular lens. To determine their cell identity, we performed immunostaining. Starting at day 22, lens markers CRYAA and CRY B were found (Fig. 3A, B) and continuously expressed in the transparent cell structures (Fig. 3C, D). Interestingly, these transparent structures were not stained by DAPI (Fig. 3C-F), indicating denucleation in these lens cells. When CONCEPT organoids were detached using Dispase at around day 28 and continuously grown as suspension cultures, crystal-like clusters with fused transparent spheres were found, and they continuously survived for months (Fig. 3I, J). In mice, terminally- differentiated lens fiber cells are free of organelles and featured by specialized interlocking cell membrane domains shown as ball-and-sockets and protrusions ⁵⁷. Using transmission electron microscopy, we found that the crystal-like lens clusters were free of organelles and exhibited ball-and-socket structures (Fig. 3K, L). Taken together, our findings demonstrate that CONCEPT telencephalon-eye organoids contain lens cells that undergo terminal differentiation; FGFs in CONCEPT organoids likely promote terminal lens differentiation as seen in other settings ⁵⁸.

To fully characterize cell populations in CONCEPT organoids and the mechanisms underlying RGC axon pathfinding, we performed 10x single-cell RNA sequencing of the organoids at day 24, around the stage when RGCs grew long axons toward and along the PAX2+ VSX2+ cell population. Cell Ranger mapping showed that 11158 single cells were sequenced at a depth of 27,842 reads and 2,967 genes per cell, and the dataset were then analyzed using Seurat (v3.2.0) ⁵⁹. Cell filtration (nFeature_RNA > 200 & nFeature_RNA < 6000 & percent.mt < 20) resulted in 10218 cells with high quality. Cell clustering grouped cells into 14 clusters (Fig. 4A; Tab. 1). In addition, cell cycle scores were plotted using an established method (Fig. 4B) ⁶⁰. Several cell clusters, e.g., clusters 4, 8, 9, were separated along cell cycle scores (Fig. 4A, B), indicating the differences in cell cycle phases across these cell clusters.

We next assessed cell identities using known markers. Mesoderm, endoderm, and neural crest are identified by a group of gene markers ^61–64. In CONCEPT organoids, gene markers for mesoderm (TBXT, GATA2, HAND1), endoderm (GATA1, GATA4, SOX17), and neural crest (SNAI1, SOX10, FOXD3) were not expressed (Fig. S2). In contrast, gene markers for the anterior neuroectoderm were widely expressed. Telencephalic progenitor marker FOXG1 was expressed in clusters 10, 3, 12, 1, 4, 8, 9, 13; retinal progenitor markers PAX6 and VSX2 were expressed in clusters 2, 7, 5, 6, 11 (Fig. 4C-E). Cluster 0 was defined by a group of negative markers (Fig. S3); they expressed both FOXG1+ telencephalic cells and PAX6+ and/or VSX2+ retinal cells. In contrast to telencephalon and retinal markers, diencephalon markers (GBX2, WNT3, SOX14) ^{65, 66} and midbrain/hindbrain markers (EN2, PAX7, TFAP2B) ⁶¹ were rarely expressed (Fig. S4). Lens markers CRYAA and FOXE3 were expressed in a very small cell population; those denucleated lens cells were not captured in the scRNA-seq since the absence of nucleus prevented active mRNA transcription. Therefore, lens cells did not form a separate cluster probably due to their very low abundance. Taken together, these findings indicate that CONCEPT telencephalon-eye organoids at day 24 are mostly composed of FOXG1+ telencephalic cells and PAX6+ and/or VSX2+ retinal cells.

We next characterized the FOXG1+ telencephalic cells via assessing differentially expressed genes (DEGs). DEGs in clusters 4, 8, 9, 1 included POU3F3, RGMA, EDNRB, and SOX3, which orthologs in mice are expressed in the pallium. DEGs in clusters 12, 3, 10 included DLX2, RGS16, DLX1, DLL1, DLX6-AS1, NEFL, DCX, and RTN1, which orthologs in mice are expressed in the subpallium (Figs. S5, S6). Therefore, FOXG1+ telencephalic cells in CONCEPT organoids at day 24 comprise cell clusters in both the pallium and subpallium.

PAX6+ and/or VSX2+ retinal cells were also grouped into several clusters. VSX2 was expressed in clusters 2, 7, 5, and 0, with cluster 2 at G1 phase, cluster 7 in G1 and S phases, and cluster 5 in S and G2M phases (Fig. 4A, B, E). Cluster 6 did not VSX2 but expressed PAX6 and was at the G1 phase, suggesting cells in cluster 6 exited the retinal progenitor cell state and underwent cell differentiation (Fig. 4A, B, D, E). Consistently, cluster 6 differentially expressed RPE markers, e.g., PMEL, HSD17B2, DCT, and MITF (Fig. S7), indicating they were mostly differentiating RPE. Cluster 11 did not express VSX2; instead, it expressed PAX6 and was at the G1 phase (Fig. 4A, B, D, E), suggesting that cells in cluster 11 exited the retinal progenitor cell state and underwent cell differentiation.

Taken together, our single-cell RNA sequencing analysis of CONCEPT telencephalon-eye organoids at day 24 confirms their telencephalic and ocular identities, establishing a valuable transcriptomic dataset for mechanistic studies.

To identify PAX2+ cell populations that defined the path for RGC axon outgrowth, we examined PAX2 expression in the dataset. PAX2 was mostly expressed in cluster 2 and subsets of clusters 7, 5, 0, 4, 8, and 9 (Fig. 4A, F). Since clusters 2, 7, 5, and 0 expressed VSX2 (Fig. 4E, F), we deduced that PAX2+ VSX2+ cells in cluster 2 and subsets of clusters 7, 5, and 0 corresponded to those PAX2+ VSX2+ cells that defined the path for RGC axon growth (Figs. 2H-L). Therefore, these PAX2+ VSX2+ cells in the dataset of single-cell RNA sequencing were assigned as the optic disc (OD) (Fig. 4F). PAX2+ VSX2+ cells also expressed optic disc and optic stalk marker SEMA5A (Fig. 4F, G) ⁶⁷. We particularly focused on cluster 2 since PAX2+ VSX2+ cells were mostly found in this cluster. Interestingly, COL13A1 was differentially expressed in cluster 2 (Fig. 4H; Fig. S10); in situ hybridization indicates that COL13A1 is specifically expressed in the optic disc of human fetal retinas ⁶⁸. These findings firmly support that PAX2+ VSX2+ cells in CONCEPT telencephalon-eye organoids corresponded to optic disc cells in human fetal retinas. Cluster 2 also differentially expressed glaucoma gene CYP1B1, signaling molecules LEFTY2, FGF9, and FGF8 (Fig. 4I-L; Fig. S10). The expression of PAX2, SEMA5A, CYP1B1, LEFTY2, FGF9, and FGF8 in CONCEPT organoids was validated using in situ hybridization (Figs. 1T-V; 5B-E; S10). In contrast, PAX2+ VSX2- cells in subsets of clusters 4, 8, 9 (Fig. 4A, E, F) expressed the optic stalk marker VAX1 (Fig. 4F, M); these cells corresponded to those PAX2+ VSX2- cells that set up the inner boundary of the path for RGC axon growth in CONCEPT organoids (Fig. 2H-L).

Therefore, PAX2+ VSX2- cells in subsets of clusters 4, 8, 9 were assigned as the optic stalk (OS) (Fig. 4F). PAX2+ VSX2- optic stalk cells also differentially expressed CNTNAP2, ALDHA3, and LAMP5 (Fig. 4N-P), which orthologs in mice are expressed in the optic stalk/nerve (Fig. S8). Compared to PAX2+ VSX2+ optic disc cells, PAX2+ VSX2- optic stalk cells were closer to telencephalic cells in both the UMAP plot (Fig. 4M-P; Fig. S5A-D) and their positioning in CONCEPT organoids (Fig. 2H-L), mimicking relative positions of telencephalon, optic stalk, and optic disc along the midline-periphery axis in E10.5-E13.5 mouse embryos (Figs. 1E, 2E-G). Consistently, several markers for PAX2+ VSX2- optic stalk cells were also expressed in the telencephalon (Fig. S8). Collectively, we identify two PAX2+ cell populations that mimic the optic disc and optic stalk, respectively, in the scRNA-seq dataset of CONCEPT telencephalon- eye organoids.

To identify RGCs in the dataset, we checked the expression of RGC markers. We found that RGC markers ATOH7, POU4F2, SNCG, and CNTN2 were differentially expressed in cluster 11, indicating its RGC identity (Fig. 4A, 4Q-T). CNTN2 caught our attention because it is a GPI- anchored cell membrane protein, which may be used as a native marker for RGC isolation. In literature, Cntn2 is specifically expressed in developing RGCs in mice and chicks ^{69, 70}.

In CONCEPT at day 25, CNTN2 exhibited an expression pattern very similar to that of TUBB3 (Figs. 2, 5). CNTN2 was found in the cell membrane of RGCs that expressed POU4F2 in the cell nucleus (Fig. 5A, D). PAX2+ cells formed inner and outer concentric zones, mimicking the optic stalk and optic disc, respectively. POU4F2+ RGCs formed a dense circular zone adjacent to PAX2+ optic disc cells; POU4F2+ RGCs were sparse in more peripheral areas (Fig. 5B).

These findings indicate that early differentiated RGCs were adjacent to PAX2+ optic disc cells. Interestingly, RGCs in the dense POU4F2+ zone grew CNTN2+ axons toward and then along the path defined by adjacent PAX2+ optic disc cells (Fig. 5A-E). In areas where there were a gap in PAX2+ optic disc cells, CNTN2+ axons exited the circular path (diamond arrow in Fig. 5A; a gap between two arrowheads in Fig. 5B, F). In regions where POU4F2+ RGCs were a few hundreds of micrometers away from PAX2+ optic disc cells, RGCs grew axons in centrifugal directions (arrow in Fig. 5C). Very similar findings were found using hiPSCs (Fig. S9), indicating the reproducibility of CONCEPT organoids in multiple cell lines. PAX2+ optic disc cells did not express ALDH1A3; instead, the path for RGC axon growth was bordered by two cell populations that highly expressed ALDH1A3 (Fig. 5G-H). In mice, Aldh1a3 expression was low in differentiating RGCs in the central retina but high in peripheral retinal progenitors and optic stalk (Fig. 5K) ⁷¹, consistent with ALDH1A3 expression in CONCEPT organoids (Fig. 5G, H). In mice, Vax1 and Vax2 are expressed in the optic stalk and ventral retina and are jointly required for the optic stalk development ^{55, 72}. In CONCEPT organoids, cells that set up the inner boundary for RGC axon growth also expressed optic stalk marker VAX1/2 (Fig. 5I, J; the antibody recognizes both VAX1 and VAX2). Taken together, our data demonstrate that CNTN2 is a specific cell- surface marker for differentiating human RGCs; RGCs grow axons toward and then along a path that is defined by adjacent PAX2+ VSX2+ ALDH1A3- optic disc cells.

Since cell surface protein CNTN2 was specifically expressed in developing human RGCs, we sought to test whether CNTN2 can be used as a biomarker for isolating human RGCs under a native condition. To that goal, retinal organoids in suspension culture on days 41 - 70 were dissociated using Accutase to generate a single cell suspension, which was then subject to magnetic-activated cell sorting (MACS) with an antibody against CNTN2. From 100 retinal organoids on day 41 to 48, around 385,000 RGCs were isolated. Isolated RGCs were plated onto Matrigel-coated chamber slides for 10-day adherent culture. These cells exhibited neuronal morphology and widely expressed RGC markers TUBB3 and POU4F2 (Fig. 5L). Isolated RGCs tended to form clusters in culture. In contrast to directional axon growth found in CONCEPT organoids, isolated RGCs grew neurites in random directions (Fig. 5L), indicating the differences in axon pathfinding cues between CONCEPT organoids and isolated RGC cultures. RGC neurites were also marked by CNTN2 expression (Fig. 5M). A portion of isolated RGCs expressed POU4F2 but not CNTN2 (Fig. 5M), indicating that CNTN2 was downregulated in dissociated RGC cultures. Isolated RGCs also expressed RGC markers ISL1, RBPMS, and SNCG (Fig. 5N-P). Collectively, we have developed a one-step method for the isolation of differentiating human RGCs in a native condition.

In order to determine the functional properties of isolated RGCs in culture, we examined their electrophysiological properties using whole-cell patch clamp recordings (Fig. 6, see Methods). Using current-clamp configuration, we found that RGCs displayed had a hyperpolarized resting membrane potential (mean ±SD: -20.1 ±6.4mV, Fig. 6A). When cells were held at -70 mV by current injection, in most cases (6/9), depolarizing currents steps could trigger an action potential, which were often followed by a depolarization plateau if the current was injected for more than 10ms (Fig. 6B). In voltage clamp recordings we examined the nature of the voltage- gated conductances (see Methods). From a holding membrane potential of -80 mV, both inward and outward currents (I_m) were observed in response to depolarizing voltage steps (V_m, Fig. 6C). Outward currents were primarily mediated by voltage-gated potassium channels, as application of 20mM TEA significantly reduced their amplitude (mean I_K ±SD at 60mV; 964±537 pA in control; 279±201 pA in TEA, U=24.0; p=0.019, two-sided MannWhitneyU test, Fig. 6D). Conversely, 1µM TTX abolished all inward currents, demonstrating that they were mediated by activation of voltage-gated sodium channels (mean I_Na ±SD at -10mV; -338.2±121 in control; - 18.3±13 pA in TTX; U=0.0; p=0.035, two-sided MannWhitneyU test, Fig. 6E). Taken together, these results indicate that isolated RGCs exhibit functional features traditionally found in excitable cells, such as in neurons.

Directional RGC axon growth toward and then along PAX2+ optic disc cells in CONCEPT organoids could be mediated by signaling molecules secreted from PAX2+ optic disc cells. scRNA-seq analysis identified expression signatures of cluster 2, the major component of PAX2+ optic disc cells (Fig. S10). Interestingly, FGF8 and FGF9 were differentially expressed in cluster 2 (Figs. 4K, 4L; 7A; S10). To assess the roles of FGF8 and FGF9 in CONCEPT organoids, we validated that FGF8 and FGF9 were expressed in PAX2+ optic disc cells of CONCEPT organoids (Fig. 7B-E). TUBB3+ RGC axons grew towards and then along the regions with high-level FGF8 and FGF9 (Fig. 7D, E), suggesting that FGF8 and FGF9 may attract RGC axon growth. FGFR1, FGFR2, FGFR3, MAP2K1, and MAP2K2 were expressed in multiple types of cells in CONCEPT organoids. In RGCs, FGFR1 and MAP2K2 were clearly expressed (Fig. S11), indicating that RGCs expressed the components that transduce FGF signaling. Since FGF8 and FGF9 are probably redundant and it is challenging to inactivate both FGF8 and FGF9 in CONCEPT organoids, we decided to inactivate FGF signaling with FGFR1/2/3 inhibitor PD 161570 during days 17-24, a stage at early RGC differentiation. After the FGFR inhibition, FGF8 expression was grossly unaffected. On the other hand, RGC somas were drastically reduced, and directional axon growth was nearly absent (Fig. 7F-7K). These findings indicate that a) PAX2+ optic disc cells differentially expressed FGF8 and FGF9; b) FGF signaling mediated by FGFRs is required for early RGC differentiation and directional axon growth in humans.

Cysts are hollow spheres composed of homogeneous columnar epithelial cells with the apical surface at the lumen. They are induced from pluripotent stem cells via embedding small sheets of hESCs into Matrigel and subsequent growth either in a solid thin film ⁴⁷ or suspensions ^{28, 33}. Cyst growth in suspensions is cost-effective and scalable. The formation of cysts induced by Matrigel mimics the epithelization of the epiblast by the extracellular matrix (ECM) in the blastocyst ^{73, 74}.

In this study, individual cysts efficiently generate telencephalic and ocular tissues in the absence of any extrinsic factors, indicating their default cell fates of the anterior ectoderm. In literature, anterior neural tissues are generated from re-aggregated single pluripotent stem cells through dual inhibition of Smad signaling ⁷⁵ or inhibition of Wnt/ß-catenin signaling ²⁷. Besides that, undirected cultures without extrinsic factors are used for the generation of retinal cultures ^{29, 30, 76}. In our procedure, we did not add any extrinsic factors; the epithelial structure of cysts and subsequent adherent growth at a low density facilitate neural induction and differentiation. Our findings are consistent with the reports that epiblast cells are epigenetically primed for ectodermal fates ⁷⁷ and neuroectodermal tissues are the default differentiation from pluripotent stem cells ^{78, 79}. Collectively, cysts mimic the anterior ectoderm in the epithelial structure and cell fates.

Tissue patterning is fundamental for the formation of a body plan, which is defined by the anteroposterior, dorsoventral, and left-right axes. Self-formation of CONCEPT telencephalon- eye organoids establishes an efficient way to pattern early telencephalic and ocular tissues along an axis. In CONCEPT organoids on days 22-26, FOXG1+ telencephalon, PAX2+ optic stalk, PAX2+ optic disc, and VSX2+ neuroretinas are positioned along the center-periphery axis, mimicking the relative positions of those tissues in E10.5-13.5 mouse embryos. Our findings indicate that the formation of telencephalic and retinal tissues are highly coordinated, consistent with the prosomere model ^{80, 81}. Collectively, we establish CONCEPT telencephalon-eye organoids for studying the early formation of telencephalic and ocular tissues in humans.

The tissue patterning in CONCEPT telencephalon-eye organoids is originated from the attachment of a radially symmetric epithelium to the culture surface for growth as colonies since individual cysts kept in suspensions do not form any apparent tissue patterning. When a floating cyst initially contacted the culture surface, the contact resulted in ECM-cell adhesions. The initial ECM-cell adhesions caused additional ECM-cell contacts in neighboring cells, resulting in the flattening and spreading of a cyst onto the culture surface. Since the cyst is a radially symmetric epithelium, ECM-cell adhesions between the cyst and the culture surface formed sequentially in a concentric manner. Remodeling of cell-cell and ECM-cell adhesions and self-organization underlay the tissue patterning. Molecular events in concentric patterns during the attachment of a cyst onto the culture surface were eventually translated to concentric gradients of morphogens that specify cell fates. The radially symmetric epithelial structure of cysts is important for the formation of a concentric pattern since such pattern would not be generated when amorphous embryoid bodies are attached to the culture surface for growth as colonies. In literature, timed BMP4 treatment is shown to promote neuroretinal differentiation from pluripotent stem cells ⁸², and FGF8 promotes telencephalic and eye development ^{5, 22}. In our system, BMP4 and FGF8, along with BMP7 and other FGFs, were highly expressed starting at early stages and gradually formed concentric morphogen gradients, which would dictate tissue patterning, resulting in coordinated specification of telencephalic, optic stalk, optic disc, and neuroretinal tissues along the center-periphery axis in CONCEPT telencephalon-eye organoids. Co-culture of these tissues in the concentric zones provides mutual cues for coordinated tissue specification and growth. Nevertheless, anchorage culture of cell colonies prevents three-dimensional morphogenesis as seen in embryos, impeding proper tissue interactions at more advanced stages. When these cell colonies are detached using dispase for suspension culture, retinal progenitor cells self-organize to form protruding, translucent spheres called retinal organoids whereas other types of cells form structures with dark appearance; concentric zones no longer exist ²⁸. It is conceivable that the simple suspension culture does not have proper physical and chemical supports provided by periocular tissues and other surrounding tissues in vivo and therefore optic stalk cells do not form a tube-like structure. It is a challenge ahead to reconstruct this three-dimensional environment to mimic the morphogenesis in vivo.

Concentric patterns of stem cell-derived cultures are also reported in a few other experiments. When dissociated single pluripotent stem cells are grown in micropatterned culture surface at certain cell densities in a medium supplemented with BMP4, concentric zones of progenitors expressing markers for trophectoderm, endoderm, mesoderm, and ectoderm are found ^{43, 44}. Cell density and colony geometry dictate cell fate specification from pluripotent stem cells. A concentric gradient of BMP4 activity regulates the patterning ⁴³. When dissociated single pluripotent stem cells are grown in pre-patterned geometrically confined culture surface in a medium supplemented with dual inhibitors for TGF-β and BMP4, concentric zones of progenitors expressing the markers for neural plate and neural plate border are observed.

Morphogenetic cues—cell shape and cytoskeletal contractile force—dictate the patterning of the neural plate and neural plate border via BMP-SMAD signaling ⁴⁵. When dissociated single pluripotent stem cells are grown as individual colonies in a medium supplemented with knockout serum replacement, multiple zones of ectodermal cells autonomously form ⁴⁶. In all three experiments, dissociated single cells are used to generate cell colonies through either cell re- aggregation or proliferation. In our case, adherent culture of a radially-symmetric epithelium – the cyst – was used to generate CONCEPT organoids. Among the three previous and our experiments, starting cells differ in their developmental potentials, tissue structures, and treatments. Despite the differences, all these organoids comprise concentric zones of distinct cell populations. Formation of CONCEPT telencephalon-eye organoids mimics the coordinated specification of early telencephalic and ocular tissues in humans.

In the mouse retina, multiple RGC axon guidance cues are concentrically organized around the optic disc, regulating RGC axon growth and exit from the eye through the optic stalk ². Early differentiated RGCs are in a short distance from the nascent optic disc, and axons of later differentiated RGCs in more peripheral retinal regions follow the path of the initial axons. It is accepted that the optic-disc regions provide growth-promoting guidance cues whereas peripheral retinal regions provide inhibitory guidance cues ². Pax2 is specifically expressed in the ventral optic stalk, optic vesicles, central neuroretina, optic disc, and optic stalk; Pax2 is essential for optic stalk and nerve development in mice ^15–17.

In CONCEPT telencephalon-eye organoids, coordinated specification of telencephalic and ocular tissues generates PAX2+ optic-disc and optic-stalk cells that define the path for RGC axon growth. To the best of our knowledge, directional RGC axon growth guided by optic-disc and optic-stalk tissues has not been reported.

Single-cell RNA sequencing of CONCEPT organoids identified DEGs of PAX2+ optic-disc and optic-stalk cells. Interestingly, FGF8 and FGF9 were differentially expressed in PAX2+ optic-disc cells; inhibition of FGF signaling with an FGFR inhibitor during early RGC differentiation drastically decreased the number of RGC somas and nearly ablated directional axon growth, indicating that early RGC differentiation and directional axon growth require FGFR-mediated FGF signaling in humans. In chicks, Fgf8 and Fgf3 coordinate the initiation of retinal differentiation ^{22, 23}; FGF8 maintains Pax2 expression in optic nerve explants ²¹. In mice, FGF8 and FGF9 trigger axon outgrowth in motor neuron column explants ⁸³. Therefore, the roles of FGF signaling in the regulation of early RGC differentiation and directional axon growth in CONCEPT organoids are consistent with literature. Dissecting DEGs of PAX2+ optic-disc cells in CONCEPT telencephalon-eye organoids elucidates the mechanisms of RGC axon growth and pathfinding in humans.

RGCs described in this study are still at their early stages in development. One prominent feature of these RGCs are the growth of their axons towards and long the path defined by optic disc cells, which is quite clear in CONCEPT organoids. Optic stalk cells in CONCEPT organoids confine RGC axon growth at inner regions, preventing RGC axons to enter telencephalic areas. We did not observe any clear evidence of synaptic or dendritic connections between RGCs and other cells at tested stages. Engineering assembloids may be needed to establish connections between RGCs and their visual targets.

RGCs are degenerated in glaucoma, a major cause of vision impairment in developed countries. Disease modeling and drug discovery to suppress RGC death will have a huge impact on saving vision. The use of human RGC models is critical for therapeutic studies since humans and rodents differ significantly in RGCs. Additionally, cell replacement therapies for glaucoma are extensively evaluated. Therefore, efficient isolation of human RGCs in a native condition will have a substantial impact on therapeutic studies of RGC-related diseases such as glaucoma. In literature, cell surface marker Thy1 is used for the isolation of adult mouse RGCs, but it is unsuccessful in the isolation of RGCs from 3D retinal organoids ⁸⁴. A study reports the purification of human RGCs via Thy1 from adherent cultures ⁸⁵. Tagging human RGCs with an engineered marker Thy1.2 leads to efficient isolation of human RGCs ⁸⁴, but may not be unsuitable for clinical uses. Using single-cell RNA sequencing, we identified cell surface protein CNTN2 as a specific marker for developing human RGCs. We isolated human electrophysiologically-excitable RGCs using MACS via CNTN2. Therefore, we establish a one- step method for isolating human RGCs under a native condition, facilitating therapeutic studies for RGC-related retinal diseases such as glaucoma.