Introduction

The tissues of the embryo develop in an evolutionarily selected confined space with robust physical and biochemical constraints. Precise spatio-temporal gradients of growth factors aligned with complex mechanical forces guide cell and tissue interactions, permitting the coordinated formation of functional tissues and organs. Eye development is an example of complex organ formation relying on the coordinated interactions of tissues of different embryonic origin. This ultimately results in the positioning of a light-focusing lens in the center of the light-sensing retina (Cardozo et al., 2023; Casey et al., 2021; Chow and Lang, 2001; Cvekl and Ashery-Padan, 2014).

While the lens originates from the head surface ectoderm (SE), the retina originates from optic vesicle (OV), a lateral evagination of the forming diencephalon. In vertebrates, eye development is initiated at the end of gastrulation when the anterior ectoderm is patterned into the neural plate, non-neuronal ectoderm and the neural plate border at their interface. Lens-committed progenitors are first found in the anterior neural plate border, also called the anterior pre-placodal region (Bailey and Streit, 2005; Bhattacharyya et al., 2004; McCabe and Bronner-Fraser, 2009; Schlosser, 2006; Toro and Varga, 2007). Retina-committed progenitors, on the other hand, are established within the anterior neural plate, by the specification of the eye field, followed by the evagination of the optic vesicle (Adelmann, 1937; Eagleson et al., 1995; Kenyon et al., 2001; Li et al., 1997). Once the retina-forming OV contacts the lens-competent ectoderm, interactions of OV, SE and the surrounding tissues result in the formation of the lens placode within the head surface ectoderm. The concomitant invagination of lens placode and morphogenesis of the OV then lead to the formation of the lens vesicle in the center of the forming optic cup (Casey et al., 2021; Chow and Lang, 2001; Cvekl and Ashery-Padan, 2014; Gunhaga, 2011). Initially the lens vesicle is composed of lens progenitor cells. It is further patterned into lens epithelium at the anterior part of the lens, while posteriorly located progenitors differentiate into lens fiber cells (Kallifatidis et al., 2011; Zhou et al., 2006). The differentiation of lens fiber cells is characterized by the production and accumulation of crystallin proteins, the main structural proteins of the lens (Cvekl et al., 2015; Donaldson et al., 2017). This transition is marked by the extensive elongation of cells (Cheng et al., 2017; Fudge et al., 2011; Rao and Maddala, 2006) and the subsequent degradation of all organelles and de-nucleation (Bassnett, 2009; Brennan et al., 2021; Chaffee et al., 2014; Wride, 2011), minimizing light scattering and providing optimal refractive properties, all together resulting in the optical properties of the lens (Bassnett et al., 2011).

Several studies have shown that the conserved process of lens formation depends on the intricate interplay of intrinsic factors and extracellular signalling pathways (Cvekl and Ashery-Padan, 2014; Cvekl and Zhang, 2017; Lang, 2004; McAvoy et al., 1999; Ogino and Yasuda, 2000). The transcription factors Pax6, c-Maf, Sox1, Prox1, Hsf4, Six3, Foxe3, Pitx3, Sox2 characterise early lens development and are directly involved in the processes of initial lens morphogenesis and lens fiber cell differentiation (Cvekl and Duncan, 2007; Cvekl and Zhang, 2017; Lang, 2004; Ogino et al., 2012; Ogino and Yasuda, 2000). Establishment of lens-competent progenitors as well as differentiation of lens fiber cells relies on the interplay of the FGF, BMP, and WNT signalling pathways (Faber et al., 2001, 2001; Furuta and Hogan, 1998; Grocott et al., 2011; Gunhaga, 2011; Jarrin et al., 2012; Kreslova et al., 2007; Le and Musil, 2001; Litsiou et al., 2005; Lovicu and McAvoy, 2005; McAvoy et al., 1999; Rajagopal et al., 2009).

Although many aspects of lens formation are well characterized, addressing the molecular mechanisms in the complex environment of the developing embryo is often challenging. Advances in stem cell and developmental biology allowed the introduction, fast development and prominent use of organoids as new models to study complex processes of ex vivo organ formation. Those allow to gain insight into cellular and molecular processes in normal and diseased conditions in a controlled and experimentally accessible in vitro environment (Chen et al., 2024; Hofer and Lutolf, 2021; Kretzschmar and Clevers, 2016; Sahu and Sharan, 2020; Zhao et al., 2022). Mouse, human and fish pluripotent embryonic stem cells have been shown to self-organise into retinal tissue when aggregated and cultured under minimal growth factor 3D suspension culture conditions. Retinal organoids recapitulate key aspects of embryonic development with respect to cell fate commitment, tissue patterning, cell differentiation and physiological functions (Eiraku et al., 2011; Kuwahara et al., 2015; Nakano et al., 2012; Völkner et al., 2016; Zilova et al., 2021). Since the culture conditions for the generation of retinal organoids favor neuroectodermal differentiation over non-neuronal ectodermal fates, they did so far not allow for concomitant lens formation (Eiraku et al., 2011; Kuwahara et al., 2015; Nakano et al., 2012; Zilova et al., 2021).

Generation of lens-like structures (lentoid bodies) by stepwise differentiation of human ES or iPS cells has been reported (Cvekl and Camerino, 2022; Dincer et al., 2013; Fu et al., 2017; Leung et al., 2013; Yang et al., 2010). However, most lens differentiation protocols directly favor the promotion of placodal fate on the expense of the neuronal fate (Dincer et al., 2013; Fu et al., 2017; Yang et al., 2010). This rendered the generation of more complex organoids composed of cell types of diverse embryonic origin challenging. Interestingly, two independent studies had reported the appearance of lens marker-expressing cells in close vicinity to retina tissue in 3D organoid culture (Gabriel et al., 2021; Mellough et al., 2015). However, considering the complex nature of tissue interactions required for the coordinated formation of both lens and retina in vivo, this raises the intriguing question of how a lens is formed in neuronal organoids and to what extend that process recapitulates in vivo organ formation.

Here we used organoids derived from Oryzias latipes blastula cells as a model system to investigate the formation of ocular organoids composed of both, retina and lens. By following the individual populations of lens– and retina-committed progenitors, we found that similar to the in vivo situation, lens and retina originate from spatially confined and separated populations of progenitor cells. The subsequent differentiation into their respective lineages was found to be dependent on BMP and FGF signaling pathways. Their differentiation and morphogenesis are characterized by the expression of key regulatory factors implicated in eye (lens and retina) organogenesis, however following an entirely different morphogenetic route towards forming an eye cup enclosing a centrally positioned lens.

Results

Fish retinal organoids form embedded lenses under specific culture conditions

We have previously reported the generation of retinal organoids from fish pluripotent embryonic cells grown in 3D suspension culture in minimal growth factor-containing media. Such organoids acquire anterior neural fate and consist of retina and presumptive forebrain tissue (Zilova et al., 2021). Here, we modified our previously described protocol (Figure 1a) in order to generate organoids, that differentiated into retinal and lens tissues.

Figure 1.

Aggregation of fish primary pluripotent cells in DMEM/F12-based differentiation media supplemented with 5% knockout serum replacement and 20 mM HEPES pH 7.4 (see Materials and Methods for detailed composition) resulted in the efficient formation of retinal organoids containing a single transparent, spherical lens-like structure surrounded by neuronal tissue (Figure 1b). The lens-like structure was clearly visible at day 4 but first signs of its formation could be observed from day 2 onwards (Figure 1b, asterisk). We showed that organoids were composed of both retina and lens tissue, arranged in a way that was reminiscent of the tissue arrangement in the embryonic eye by using reporter lines and molecular markers. Differentiated lens cells expressing crystallin were labelled in the Gaudí^RSG transgenic reporter line (Centanin et al., 2014) that carries a crystallin::ECFP (Cry::ECFP) reporter. Lens fiber cells were identified with the “lens fiber cell” marker antibody ZL-1 (referred to as LFC), and early and late lens cells were identified with an anti-Prox1 antibody (Figure 2a and 2b). Prox1 is expressed by cells of the lens placode and differentiating lens fiber cells of different species including mouse, zebrafish and medaka (Mikula Mrstakova and Kozmik, 2024; Pistocchi et al., 2008; Wigle et al., 1999). Anti-LFC labels lens fiber cells in zebrafish and is suggested to target one of the crystallin proteins (Greiling et al., 2010; Shi et al., 2006). In the medaka embryo, the expression of both, crystallin promoter-driven ECFP and LFC marker was observed exclusively in the lens 3 days post-fertilization (dpf) (Figure 2a). Prox1 immunolabeling was detected in the lens and a subpopulation of differentiating retinal cells in the central region of the retina (Figure 2a). The expression of all three lens markers was detected in day 4 ocular organoids (Figure 2b and 2c), demonstrating that the spherical structure formed by the organoids is indeed the lens. ECFP labeling was sparse (25 % Cry::ECFP, Figure 2b) reflecting the mode of generation of the organoids by a combination of 25 % cells carrying the Cry::ECFP reporter and 75 % wild-type cells (see Materials and Methods for details).

Figure 2.

The transgenic reporter lines Atoh7::EGFP (Del Bene et al., 2007) and Rx3::H2B-GFP (Rembold et al., 2006) were used to identify retinal tissue in lens carrying organoids. In embryos Atoh7::EGFP activity was confined to the retina, labelling differentiating retinal neurons at 3 dpf (Figure 2a). In organoids, labeled retinal cells surrounded the forming lens, reminiscent of the situation in the eye. To address the distribution of retinal and lens cells we used blastula cells from Atoh7::EGFP and Rx3::H2B-GFP retina-specific reporter lines (Figure 2d and 2e) and co-labeled Rx3::H2B-GFP-derived day 4 organoids with anti-LFC antibody (Figure 2e).

Analysis of the organoid lens revealed a great similarity with structural features of the developing embryonic lens, such as the elongation of Prox1⁺/LFC⁺ cells, typical for differentiating lens fiber cells and atypically looking nuclei indicating disorganization of nuclear material, a process tightly connected to lens fiber cell differentiation (Figure 2b and 2h, arrowhead).

In 462 organoids derived from 38 independent experiments we observed lens formation, in 83.5 % (n=386) of organoids. A careful evaluation of different culture conditions and their impact on the lens formation revealed that lens formation depends on supplementation of the DMEM/F12-based differentiation media with 20 mM HEPES buffer (Figure S1a and 1b). While organoids grown without HEPES supplementation never formed lenses (n=63 organoids in 6 independent experiments, see also Zilova et al., 2021), 92 % of organoids grown in media supplemented with HEPES (n=83 organoids in 6 independent experiments) formed lenses (Figure S1b).

We tested the impact of extracellular matrix (ECM) supplementation on lens formation, as existing protocols for generation of lentoid bodies from human pluripotent embryonic stem cells mostly rely on supplementation with ECM proteins (Dincer et al., 2013; Fu et al., 2017; Leung et al., 2013; Yang et al., 2010), usually in form of Matrigel. Interestingly, lens cell fate specification as well as lens fiber cell differentiation, judged by the expression of Prox1 and LFC, occurred independently of Matrigel (2 % final concentration) supplementation in medaka-derived organoids (Figure S1c).

The lens originates from a spatially distinct population of placodal progenitors located in the central region of the organoid

We further investigated the origin of lens cells and how its formation compares to lens development in vivo in the developing organism. In the embryo, the lens originates from the lens competent head surface ectoderm (SE, magenta), which directly abuts the evaginating optic vesicle (OV, prospective retina, green) (Figure 3a). OV and SE invaginate in a coordinated way to form a lens sphere surrounded by the retina. In medaka embryos, those processes occur between 1 dpf and 2 dpf (Figure 3a, 3b). Concomitant with the completion of lens formation, the differentiation of lens fiber cells is initiated and marked by the expression of crystallin proteins (LFC; cyan) (Figure 3b). To track the distribution of lens and retinal progenitors, we generated organoids from the Rx3::H2B-GFP reporter line, which highlights retina-committed cells by nuclear GFP expression (Figure S2a and 2b). To visualize lens-committed progenitor cells within the SE (Mikula Mrstakova and Kozmik, 2024), we co-labeled those organoids with an anti-Prox1 antibody (Figure S2b) and followed the expression of Rx3-specific GFP and Prox1 from day 1 to day 2 (Figure 3c). Retinal progenitor cells were distinctly localized in the cortical region around the surface of day 1 organoids (Figure S2d, Figure 3c, Video S1), while lens progenitors (expressing Prox1) were not detectable at this stage (Figure 3c). Expression of retinal markers was confined exclusively to the cortex and was never detected in the central part of the organoid (Figure 2b, Figure S2d, Video S1). From day 1.5 onwards (stage S22 – 1 day and 14 h) (Iwamatsu, 2004), embryos as well as organoids (30 hours post aggregation) started expressing Prox1 (Figure 3b and 3c). In the embryo, Prox1 expression was localized to the surface of the head, labeling the presumptive lens region (Figure 3b). In the organoid, the expression was confined to the central core of the organoid with Prox1⁺ cells organized into a sphere (Figure 3c). Prox1 expression in the central core of the organoids was followed by the expression of lens fiber cell differentiation marker LFC at day 2, concomitant with the onset of Prox1 and crystallin expression in the embryo (Figure 3b and 3c).

Figure 3.

To specifically address the dynamics, onset, and cellular re-arrangements of lens-committed progenitors, we generated a Foxe3::GFP reporter line. Foxe3 (winged helix/forkhead domain transcription factor) is necessary for proper lens development and is expressed by the cells of the presumptive lens ectoderm and lens epithelium of different species, including medaka (Blixt et al., 2000; Brownell et al., 2000; Mikula Mrstakova and Kozmik, 2024; Shi et al., 2006). This pattern is fully recapitulated by Foxe3::GFP with an onset of expression starting at 1 dpf (stage S20/S21; Iwamatsu, 2004). It was further expressed by the majority of lens cells at 2 dpf, where it preceded the expression of lens fiber cell marker LFC (Figure 3d). During development, the lens is composed of anteriorly located lens epithelium (proliferative lens progenitors) and posteriorly and centrally located differentiating lens fiber cells (Greiling et al., 2010). We could differentiate those parts of the developing lens with Foxe3::GFP expression in both, lens fiber cells and lens epithelium in 3 dpf embryo, and the expression of Prox1 and LFC restricted to differentiating fiber cells, but excluded from the lens epithelium (Figure S3a, arrows). Analysis at later stages showed that at 4 dpf, Foxe3::GFP expression was maintained in the lens epithelium (proliferatively active progenitors) and was nearly undetectable in differentiated lens fiber cells (Figure S3b, arrows).

In organoids, lens progenitor cells expressing Foxe3::GFP were first found in the central core at day 1 (26 hours post aggregation) (Video S2, Figure 3e, Figure 4, Video S3). These cells appeared in a pattern complementary to the retinal marker Rx3::H2B-GFP (Video S1), indicating that lens progenitor cells originate from a spatially distinct population in the central core of the organoid. We never detected any Rx3-positive cells eventually acquiring a lens fate. To address the distribution of lens and retina domains, we generated organoids from mixed blastula cells derived from the retina-specific reporter line Rx2::H2B-RFP (Inoue and Wittbrodt, 2011) and the lens specific reporter line Foxe3::GFP (Figure S3c). By day 2, lens progenitor cells expressing Foxe3::GFP contributed to the formation of a spherical lens in the central core of the organoid surrounded by unlabeled cells and eventually by the cortical retinal tissue (Figure 3g, Video S4). This was confirmed by time-lapse imaging of ocular organoid development from day 1 to day 2 where Foxe3::GFP-expressing cells appear initially in the core region of the organoid where they organize into a spherical lens within 16 hours (Video S2, Figure 4). This process was accompanied by gradual displacement of the forming lens from center to the periphery where the lens is eventually reaching the surface of the organoid (Video S2, Figure 3f and 3g, Figure 4).

Figure 4:

Further analysis showed that the organoid lens consists of proliferating as well as differentiating cells (Figure S3d and 3e), however not in distinct domains as in the embryo. We found that Foxe3-driven GFP was co-expressed with the LFC differentiation marker in day 2 organoids (Figure S3e). EdU incorporation also showed that the lens is formed by proliferative (EdU⁺) Foxe3::GFP-expressing cells (Figure S3d). On the surface of day 2 organoid lenses, we initially detected stretches of Foxe3⁺/LFC^- cells, reminiscent of the lens progenitor cells of the lens epithelium (Figure S3e and 3f, arrows). However, at day 4, organoid lenses did not show any apparent lens epithelium (Prox1^-/LFC^-) (Figure 2g and 2h). In fact, we did not detect any Prox1^-/LFC^- cells on the surface of the day 4 lenses, indicating that in organoid lenses a lens epithelium is absent at later stages.

Lens size scales with organoid size

Since retinal and lens progenitors occupied distinct regions within the organoid, we next investigated how their distribution is affected by total organoid size. We generated organoids from different numbers of cells at day 0 (4,000, 2,000, 1,000 and 500 cells). This approach resulted in organoids of different sizes at day 1: organoids derived from 4,000 cells had an average diameter of 521 ± 13 µm, those from 2,000 cells measured 379 ± 15 µm, from 1,000 cells 310 ± 22 µm, and from 500 cells 201 ± 55 µm (Figure 5a, 5b and 5d). Interestingly, the thickness of the tissue acquiring retinal progenitor cell identity in the cortex of the ocular organoid (Rx3::H2B-GFP) remained constant, independent of the initial size of the aggregate reaching to the depth of: 59.7 μm ± 1.6 μm for 4,000 cells; 55.6 ± 3.1 μm for 2,000 cells; 56.4 μm ± 3.2 μm for 1,000 cells and 55.4 μm ± 6.8 μm for 500 cells (Figure 5b and 5e). While the thickness of retinal neuroepithelium stayed constant, the lens size scaled proportionally with the size of the organoid. By day 4, larger organoids developed larger lenses, whereas smaller organoids formed smaller lenses (224 μm ± 12 μm for 4,000 cells; 199 μm ± 26 μm for 2,000 cells; 115 μm ± 33 μm for 1,000 cells and 68 μm ± 20 μm for 500 cells) (Figure 5c and 5f). Although 500 cell derived organoids were able to form lenses, the efficiency of lens formation was considerably lower (33.3 %; Figure 5g) than in organoids of the bigger sizes (between 86.7 – 95 % for organoids generated by 1,000 or more cells) (Figure 5g). Ocular organoids generated by the aggregation of less than 500 cells did not form any lenses (Figure 5g). Taken together, these data indicate that lens size scales with the size of the organoid and is defined by the constant thickness of the forming retina.

Figure 5.

Lens formation depends on temporal activity of FGF and BMP signaling

The competence to form a lens is established during the formation of the anterior neural plate within the region of the anterior neural plate border and is regulated by the coordinated activity of BMP and FGF signaling (Gunhaga, 2011; Litsiou et al., 2005; Sjödal et al., 2007). Once lens competence is established within the SE, FGF and BMP signaling induce its differentiation into the lens placode and subsequently into the lens, accompanied by the expression of structural lens proteins (Faber et al., 2001; Furuta and Hogan, 1998; Le and Musil, 2001; Rajagopal et al., 2009; Sjödal et al., 2007; Wawersik et al., 1999; Zhao et al., 2008). To address the contribution of BMP and FGF signaling in organoid lens formation, we exposed developing organoids to FGF and BMP antagonists, SU5402 and Noggin or Dorsomorphin, respectively. To separate the establishment of lens competence from the process of lens differentiation, we divided the treatments into two discrete time windows: day 0 to day 1 corresponding to the time before the lens placode formation (establishment of lens competence), and day 1 to day 2 covering lens placode formation and differentiation of lens fiber cells (Figure 6a). We then assessed the organoids at day 2 for the expression of the early lens progenitor marker Foxe3::GFP and the differentiation marker LFC.

Figure 6.

Blocking BMP signaling with Noggin or Dorsomorphin completely prevented the expression of lens-specific markers if applied from day 0 to day 1 (Figure 6b and 6c) but did not impact on lens formation if administered from day 1 to day 2 (Figure 6b and 6d). In contrast, inhibition of FGF signaling with SU5402-mediated did not affect initial lens formation if applied from day 0 to day 1 and allowed expression of both, lens placode and lens differentiation markers. However, it efficiently blocked the differentiation of lens fiber cells when applied from day 1 to day 2, and completely prevented the expression of the LFC marker (Figure 6b, 6c and 6d). This indicates that similar to embryonic development, the establishment of lens competence (between day 0 and day 1) depends on the activity of BMP signaling, while FGF signaling is required for the lens differentiation process (day 1 to day 2).

Lens organoid gene expression recapitulates the onset of lens-specific genes in vivo

To address the temporal dynamics of lens-specific gene expression in in vivo and in organoid-based lens development, we performed total RNA sequencing analysis. Embryos at 1, 2, 3, and 4 dpf as well as organoids at corresponding developmental stages, were analyzed for the expression of key markers of lens development. For total RNA sequencing we extracted RNA from whole organoids (composed of both lens and retina), the anterior part of day 1 and 2 embryos (dissected right behind optic vesicles or optic cups), and the eyes of 3 dpf and 4 dpf embryos (Figure 7a). Within the total RNA reads we focused specifically on key genes implicated in the different stages of lens development. Among the genes analysed, the earliest expressed was the transcription factor Foxe3, with expression peaking at day 1 of development, followed by the upregulation of the transcription factor Pitx3 (Figure 7c, Data S1). Day 2 marks the onset of expression of main structural lens proteins in embryos and organoids, namely: crystallins: cryaa (αA-crystallin), cryba1a (βA1-crystallin), crybb1l2 (βB1 like 2), crybb1l3 (βB1 like 3), cryba2b (βA2), cryba4 (βA4), crybb1 (βB1), crygn2 (γN2-crystallins), crygmx (γMX), crygm3 (γM3); major lens membrane structural proteins (MIPs/acquaporins) mipa and mipb coding for water channels ensuring water transport between lens cells (Agre and Kozono, 2003; Donaldson et al., 2024; Varadaraj et al., 1999); Gap junction proteins Gja8 (connexin-50) and Gje1 (connexin-23) and lens intrinsic membrane proteins lim2.2 and lim2.4 facilitating intercellular transport of metabolites and ions in lens fiber cells (Mathias et al., 2010). Day 3 was characterized by the onset of expression hsf4 (heat shock transcription factor 4), coding for a transcription factor required for de-nucleation and organelle degradation during lens terminal differentiation (Cui et al., 2013; Gao et al., 2017). Accordingly, dnase1l1l (deoxyribonuclease I-like 1-like) and plaat1 (phospholipase A/acyltransferase) implicated in DNA and lens organelle degradation, respectively (Morishita et al., 2021; Zhang et al., 2020) were also expressed from day 3 of development. At the same stage, the expression of proteins making up the structural scaffold critical for fiber cell architecture and mechanical properties of the lens was initiated. Those include beaded filament proteins Bfsp1 (filensin) and Bfsp2 (phakinin) and Lgsn (lengsin) (Harding et al., 2008; Perng et al., 2007; Song et al., 2009).

Figure 7.

Our comparative analysis indicates highly reminiscent gene expression patterns of key marker genes when comparing organoid and embryonic lens development (Figure 7c, Data S1). This demonstrates that our conditions promote the development of ocular organoids in which the expression of key markers genes is temporally recapitulating the order of expression in the development of the embryonic lens.

Discussion

In vitro organoid development builds on the remarkable ability of embryonic stem cells or progenitor cells to autonomously self-organize in higher order structures with sophisticated microarchitecture, often closely resembling cell type compositions, 3D structure, and the function of the organ (Camp et al., 2015; Dye et al., 2015; Eiraku et al., 2011; Lancaster et al., 2013; McCracken et al., 2014; Nakano et al., 2012).

Here, we used medaka derived pluripotent cells as a model for the generation of complex ocular organoids composed of both, retina and lens. We showed that fish retinal organoids efficiently form embedded lenses, in a process that employs the molecular pathways including expression of key transcription factors and utilization of extracellular signaling pathways, as the embryo. In contrast to in vivo development, organoid cells did not follow the canonical process of lens formation, but used an alternative mode of cellular assembly to generate and grow the lens. While in the embryo, the lens sphere is shaped by the coordinated invagination of lens committed head surface ectoderm and retina-forming optic vesicle neuroepithelium, lens progenitors in the organoid were established in a different fashion directly inside of the retinal organoid to form the lens.

By following early progenitors of both retina (Rx3) and lens (Foxe3) we found that retina progenitor identity was acquired at the surface, while a spherical domain with lens progenitor identity was established in the center of the organoid (Figure 3c and Figure 3e). While in the embryo the lens originates from the head surface ectoderm, non-canonical routes to lens regeneration have been described in some species. In newts, the lens can be regenerated from the pigmented epithelial cells of the iris after injury or lens removal (Agata et al., 1993; Grogg et al., 2005; Maki et al., 2007). Lens formation was also observed in the cultures of chick neuroretinal cells (Iida et al., 2017; Okada et al., 1975) showing that retinal-fated cells can give rise to lens cells under certain circumstances. However, in medaka ocular organoids, retina and lens are formed by spatially separated populations of retina– and lens-committed progenitor cells.

The size of the lens scaled with the size of the organoid and was limited by the volume of the organoid that acquired retinal fate (Figure 5). The constant depth of retinal fated cells in the cortex under the surface of the organoid, regardless of the size of the aggregate, suggests that retinal fate is influenced by environmental gradients, which cause the diffusion-limited availability of retina-inducing factors. It has previously been shown that cells on the surface of the organoid experience higher level of oxygen tension, higher level of nutrients and signaling molecules than cells located more centrally (Grün et al., 2023; McMurtrey, 2016; Qian et al., 2019; Tse et al., 2021) when grown in 3D suspension culture. Additionally, cells at the surface experience different mechanical forces, mostly due to the interactions with ECM (Vining and Mooney, 2017). Both, retinal as well as lens fates, were established either before ECM addition or in the absence of Matrigel (Zilova et al., 2021, this study), making the environmental gradient more likely to contribute to the observed tissue arrangement in fish organoids.

Interestingly, Bmp inhibition in the early stages of organoid development, before the appearance of Foxe3-expressing lens progenitor cells (day 0 to day 1) completely abolished lens formation, while BMP inhibition during the lens formation process (from day 1 to day 2) did not show any impact. Thus, we hypothesize that a source of BMP in the forming neuroretina establishes initial lens competence in the center of the organoid. During normal development, BMP4 is secreted by the neuroepithelium of the developing OV and was found to be essential for lens induction (Furuta and Hogan, 1998; Rajagopal et al., 2009). Accordingly, ablation of the prospective neuroectoderm in chick (Kamachi et al., 1998) or defective OV formation in mouse (Klimova and Kozmik, 2014; Mathers et al., 1997; Porter et al., 1997) prevents lens formation, indicating that the developing neuroepithelium stimulates lens formation within the pre-placodal ectoderm. The emergence of lens fate in the center of ocular organoids consequently depends on a critical threshold of the morphogen to be reached in the center. The scaling of lens size with organoid size supports this hypothesis as larger organoids are better suited to reach this threshold – potentially due to increased total ligand production (in the forming neuroretina) resulting in the formation of a more pronounced gradient.

During normal development, developing retinal cells have been suggested to serve as a source of FGFs stimulating expression of lens structural genes and inducing lens differentiation (Chow et al., 1995; Faber et al., 2001; Gunhaga, 2011; Le and Musil, 2001; McAvoy et al., 1999; Robinson, 2006; Zhao et al., 2008). Accordingly, blocking of FGF signaling activity during organoid lens formation prevented lens differentiation. The dependency of lens formation on the activity of BMP and FGF signaling pathways has been well documented also in other in vitro systems where transient activation of BMP and FGF signaling is necessary for the in vitro differentiation of lens-like structures (lentoid bodies) (Dincer et al., 2013; Fu et al., 2017; Leung et al., 2013; Yang et al., 2010).

Since organoids show remarkable similarities to their in vivo counterparts when it comes to cellular composition and key structural features, we naturally tend to infer that corresponding structures are formed by corresponding cellular processes in both systems. Here we show an example of an alternative mode of cellular assembly that leads to a similar structural and molecular outcome.

Organoids can recapitulate many aspects of embryonic development. However, the absence of embryonic constraints can lead to significant differences in tissue organization, cellular interactions, and overall structure. While the goal of many organoid studies is to replicate in vivo conditions as closely as possible (Chen et al., 2024; Hofer and Lutolf, 2021; Kretzschmar and Clevers, 2016; Sahu and Sharan, 2020; Zhao et al., 2022), the unconstrained environment of organoids offers a unique opportunity to tap the organoid’s potential of self-organization. This can uncover alternative developmental pathways that are not apparent in vivo and could expand our understanding of the plasticity, robustness and adaptability of developmental programs when unleashed from evolutionary constraints. Insights gained from alternative assembly processes could inform the design of synthetic biological systems and potentially lead to innovative strategies for creating complex tissues and organs in vitro.

Materials and Methods

Fish handling and maintenance

Medaka (O. latipes) stocks were maintained according to the local animal welfare standards (Tierschutzgesetz §11, Abs. 1, Nr. 1, husbandry permit AZ35-9185.64/BH, line generation permit number 35–9185.81/G-145/15 Wittbrodt). The following medaka lines were used in this study: Cab strain as a wild type (Loosli et al., 2000), Rx3::H2B-GFP (Rembold et al., 2006), Rx2::H2B-GFP (Inoue and Wittbrodt, 2011), Atoh7::EGFP (Del Bene et al., 2007), Gaudí^RSG(Centanin et al., 2014), Foxe3::GFP (this study).

Generation of organoids

Organoids were generated as previously described (Zilova et al., 2021) with slight modification of the protocol. Blastula stage embryos, stage 11 (Iwamatsu, 2004) were de-chorioneated, cells mass was isolated and washed twice with PBS (Thermo Fisher Scientific, Cat#:10010023). Cell mass was dissociated in PBS by pipetting up and down and centrifuged through 40 μm strainer (pluriSelect, Cat#:43-10040-51) for 3 min at 180 x g. Pellet was re-suspended in differentiation media: DMEM/F12 (Gibco, Cat#:11320033), 5 % KSR (Gibco, Cat#:10828028), 1x NEAA (Sigma-Aldrich, Cat#:M7145), 1x sodium pyruvate (Sigma-Aldrich, Cat#:S8636), 1x penicillin/streptomycin, 20 mM Hepes pH 7.4 (Roth, Cat#:9105.4), 0.1 μM ϕ3-mercaptoethanol (Gibco, Cat#:21985-023). If not stated otherwise, 3.000 cells per aggregate in 100 μl were seeded into individual wells of low-binding 96-well plate (Thermo, Cat#:174925 Nunclon Sphera). Cells were incubated in 26°C in the incubator without CO₂ control. At day 1, aggregates were washed in differentiation media, Matrigel (Corning, Cat#:356238) was added to final concentration of 2 % and aggregates were incubated in humidified tissue culture incubator under 5 % CO₂ condition.

To label specific population of cells, organoids were generated from fish reporter lines expressing fluorescent protein under the control of tissue specific promotor/enhancer (Cry::ECFP; Rx3::H2B-GFP; Rx2::H2B-RFP; Atoh7::EGFP; FoxE3::GFP). In that case, cells were extracted from the blastula stage embryos originating from the outcross of fish reporter line and wild-type fish, and in some cases further mixed with cells originating from wild-type embryos of the same stage to achieve sparse labeling. The proportion of cells coming for transgenic reporter line genetically able to express the transgene is indicated in the figures in percentage (%).

Treatment of cells/aggregates/organoids

Cells or aggregates were treated with Noggin (Merck Millipore, Cat#:GF173, 100 ng/ml), Dorsomorphin (Merck, Cat#:171260, 100 ng/ml), or SU5402 (Sigma, Cat#:SML0443, 10 μM). For treatment from day 0 to day 1, blastula stage cells were resuspended directly in differentiation media containing an antagonist in desired concentration. At day 1 (16 hours post aggregation), aggregates were washed twice with fresh differentiation media, supplemented with 2% Matrigel (Corning, Cat#:356238) in differentiation media and incubated till day 2. For treatment from day 1 to day 2, day 1 aggregates (16 hours post aggregation) were washed with differentiation media and transferred into antagonist-containing media and immediately supplemented with 2% Matrigel diluted in antagonist-containing media and incubated till day 2. All organoids were fixed at day 2 and analyzed for the presence of lens and retinal tissues by immunohistochemistry.

Fluorescent labeling

Fish embryos and organoids were fixed in 4% PFA overnight at 4°C, washed with PTW (PBS supplemented with 0.1 % Tween 20), permeabilized by incubation in acetone for 15 min in – 20°C and washed tree times with PTW. Samples were blocked in PTW supplemented with 4 % sheep sera (Sigma, Cat#:14C509), 1 % BSA (Sigma, Cat#:4503) and 1 % DMSO (Sigma, Cat#:D4540) and incubated with primary antibody diluted in 0.1 x blocking solution overnight at 4°C. Primary antibodies used in this study: rabbit anti-Rx2 (Reinhardt et al., 2015) (1:500), chicken anti-GFP (Invitrogen, Cat#:A10262, 1:500), mouse anti-lens fiber cell (Abcam, Cat#:ab185979, ZL-1, 1:500), rabbit anti-Prox1 (Millipore, Cat#:AB5475, 1:2000), rabbit anti-RFP (Clontech, Cat#:632496, 1:500). Samples were washed with PTW and incubated with the mixture of DAPI nuclear stain (Sigma, Cat#:D9564, 1:500) and secondary antibodies diluted in 0.1 x blocking solution overnight at 4°C. Secondary antibodies used in this study: donkey anti-mouse 647 (Jackson, Cat#:105869, 1:750), donkey anti-chicken 488 (Jackson, Cat#:703-545-155, 1: 750), donkey anti-rabbit 594 (Invitrogen, Cat#:A21207, 1:750). Samples were washed with PTW. For detection of proliferating cells, Click-iT EdU Alexa Fluor 647 Imaging Kit (Invitrogen, Cat#:C10340) was used according to manufacturer instructions. All immunolabeled samples were transferred into clearing solution composed of: 20% urea (Vetec, Cat#:V900119), 30% D-sorbitol (Vetec, Cat#:V900390), 5% glycerol (Vetec, Cat#:V900122) in DMSO (Vetec, Cat#:V900090) (Zhu et al., 2019) and imaged with laser scanning confocal microscope Sp8 (Leica).

Imaging

Fixed immunolabeled samples were imaged in MatTek glass bottom dish (Mattek, Cat#:D35G-1.5-10-C) with Sp8 confocal microscope (Leica). Organoids overview bright-field pictures were taken with wide-field microscope (Mi8; Leica). For live imaging of lens formation (Video S2, Figure 4), organoids generated from Foxe3::GFP reporter line were imaged from day 1 (24 h post aggregation) to day 2 (40 h post aggregation). Imaging was caried out at 26°C directly in low binding 96-well plate (Thermo, Cat#:174925 Nunclon Sphera) with ACQUIFER Imaging Machine (ACQUIFER Imaging GmbH, Heidelberg, Germany) (Pandey et al., 2019). For each well, a set of 10 z-slices (70 µm step size) was acquired in the bright field (50% LED intensity, 40 ms exposure time) and 470 nm fluorescence (50% LED excitation source, FITC channel, 200 ms exposure time) channels. Imaging was performed over 16 hr of organoid development with 30 min intervals.

Transcriptome analysis

RNA was isolated from 1 dpf, 2 dpf, 3 dpf and 4 dpf medaka embryos as indicated in Figure 7a. Material from 6 embryos per condition was pooled. For embryos 1 and 2 dpf, head region containing optic vesicle and head structure anterior to the optic vesicles and eye cup and tissue anterior to the eye respectively were dissected in ice cold PBS. For stages 3 dpf and 4 dpf, eyes were dissected in ice cold PBS. Considering 3 h delay in organoid development with the respect to embryonic development (Zilova et al., 2021), organoids were collected for RNA isolation 3 h later than the embryonic tissue. For organoids, 20 organoids were pooled per condition. Samples were prepared in three replicates. Samples were rinsed with ice cold PBS and homogenized in 200 μl of Trizole and RNA was isolated using Direct-zol RNA Microprep (Zymoresearch, Cat#:R2060) according to the manufacture protocol. Library preparation was performed with the NEBNext Ultra II RNA Library Prep Kit for Illumina with NEBNext Poly A Selection Kit and NEBNext Multiplex Oligos with UMI. Input RNA was quantified with the Qubit BR RNA assay, and qualified on the Agilent Tapestation (RNA Assay), libraries were quantified with the Qubit DNA HS Assay, and qualified on the Agilent Tapestion (D1000 Assay). Libraries were sequenced on the Illumina NextSeq 550 Sequencer.

For downstream bioinformatic analysis, UMIs were extracted using UMI-Tools v1.1.2 (Smith et al., 2017). The first 12 bases of each read were trimmed, and reads shorter than 50 bases with an average quality lower than 20 were discarded using Trimmomatic v0.39 (Bolger et al., 2014). Processed reads were mapped to the Japanese medaka HdrR reference genome ASM223467v1 using Bowtie 2 v2.4.4 (Langmead and Salzberg, 2012), and deduplication of the mapped reads was performed with UMI-Tools v1.1.2 (Smith et al., 2017). Gene counts were extracted with featureCounts v2.0.3 (Liao et al., 2014) using the Ensembl version 113 GTF as a reference. The subsequent analysis was performed using R v.4.4.1 (https://www.r-project.org/). Genes with fewer than 10 counts across samples were excluded, and DESeq2 (Love et al., 2014) was used for normalization and variance stabilizing transformation of the gene counts. The transformed counts were then used for principal component analysis and heatmap visualization of lens-specific gene expression.

Generation of FoxE3::GFP reporter line

1.5 kb region upstream of Foxe3 (ENSORLG00000017669) ATG was PCR amplified from medaka fish genomic DNA with Q5 High Fidelity DNA Polymerase (NEB, Cat#:M0491) using following primers: KpnI-FoxE3 (forward primer): 5’-TAAGGTACCTGGAATCAAGCAGAAATAATCCACCAAAAGCA and FoxE3_R1 (reverse primer): 5’-GGGGTACCGTTGCACATGCCAGAGAAGTAG, sub-cloned into pJet1.2/blunt cloning vector (Thermo Fisher Scientific), released via restriction digest with XhoI (NEB, Cat#:R0146) and NcoI (NEB, Cat#:R0193) restriction enzymes. The plasmid ISceI-pBSII KS+ (NovoPro, Cat#:V009906) (containing GFP sequence inserted via EcoRI and XbaI restriction sites) was digested with XhoI and NcoI, and ligated with the 1.5 kb Foxe3 upstream regulatory element to generate I-Sce/FoxE3(1.5kb)::GFP. Plasmid DNA was isolated with QuiAprep Spin Miniprep Kit (Quiagen, Cat#:27106), sequenced (Eurofins Genomics) and analyzed using Geneious R8.1 (Biomatters). I-Sce/FoxE3(1.5kb)::GFP (10 ng/μl) was co-injected with Meganuclease (I-SceI) (NEB, Cat#:R0694L) into the cytoplasm of one cell stage medaka zygotes as previously described (Thermes et al., 2002) to generate Foxe3::GFP reporter line. Embryos were raised and maintained at 26°C in 1× Embryo Rearing Medium (ERM, 17 mM NaCl, 40 mM KCl, 0.27 mM CaCl2, 0.66 mM MgSO4, 17 mM HEPES) and screened for ocular GFP expression 1 dpf on a Nikon SMZ18.

Quantitative analysis of organoid and lens size

Organoid and lens sizes (Figure 5d and 5f) were measured as diameter (mean of 4 measurements per organoid) of the organoid on central optical section through organoid (organoid size at day 1) or on maximal projection of multiple z planes of the organoid (lens size at day 4). Depth of Rx3::H2B-GFP expression (Figure 5e) was measured on central optical section through the organoid (mean of 8 measurements per organoid) as a thickness of tissue carrying GFP⁺ cells. Statistical analysis and plots were prepared with GraphPad. Figures were assembled with Affinity Designer 2.

Acknowledgements

The authors thank to E. Hasel de Carvalho and J. Benjaminsen for valuable discussions and comments on the manuscript, T. Kellner and B. Wittbrodt for technical help and maintaining fish stocks. Library preparation and sequencing was performed by CCTP Deep Sequencing Facility of Heidelberg University.

This work was supported by grants of the Excellence Cluster “3D Matter Made to Order” (EXC-2082/1-390761711) funded through the German Excellence Strategy via Deutsche Forschungsgemeinschaft (DFG, German Research Foundation), PostDoc take-off Grant funded through and the Excellence Cluster ‘3D Matter Made to Order’ to L.Z. and J.W.

Additional information

Data Availability Statement

Raw data from bulk RNA sequencing (Figure 7) will be deposited upon acceptance in publicly available repository of Heidelberg University, HeiData.

Author Contributions Statement

Conceptualization: L.Z., J.W.

Investigation: E.S., M.A.D.T., L.Z., A.P., I.S.

Visualization: E.S., M.A.D.T., I.S., L.Z.

Funding acquisition: J.W., L.Z.

Project administration: J.W. Supervision: L.Z.

Writing – original draft: L.Z., J.W.

Funding

Deutsche Forschungsgemeinschaft (EXC-2082/1-390761711)

Additional files

Supplemental PDF. Figures S1-S3.

Video S1. Cells of the outer layer of the organoid acquire retina fate, related to Figure 3.

Video S2. Time laps imaging of multiple organoids from day 1 to day 2 showing the formation of the lens, related to Figure 4.

Video S3. Lens fate is acquired by cells localized in central part of the organoid, related to Figure 3.

Video S4. Arrangement of retinal and lens tissue in day 2 organoid, related to Figure 4.

Data S1. Excel table, related to transcriptome analysis of Figure 7. Normalized and variance stabilized counts labeled with Ensemble IDs.