In preprints: folding from the top – apical forces drive eye cup formation

How does a flat sheet of cells fold itself into a hemispherical cup capable of focusing light? In other words, how does the vertebrate eye form? Eye development has long fascinated developmental biologists, not only for its complex evolutionary history and rather conserved use of common gene regulatory networks, but also for its morphogenesis, which involves a complex mechanical choreography. The primordium of the mature vertebrate eye, the optic cup (OC), arises from a seemingly simple bi-layered vesicle (the optic vesicle, OV). Studies of the last 15 years or so, mostly in zebrafish embryos, have dissected a series of cellular transformations in the OV neuroepithelium that are thought to drive its bending into an OC.The current view proposes that this transformation depends on the coordinated interplay of molecular and cellular mechanisms and mechanical forces, and therefore is best understood at the interface between biology and physics: a field of study known as mechanobiology (Shellard and Mayor, 2023). At its simplest, eye morphogenesis emerges from the integration of cellular force generation, molecular regulation and extracellular matrix interactions, all governed by the underlying principles of physics. There are at least four well-recognized cellular transformations taking place during OC morphogenesis: neural retina (NR) basal constriction; retinal pigmented epithelium (RPE) cell flattening, rim cell migration; and lens placode invagination (Cardozo et al., 2023). While each of these events has been independently studied, the field lacks an integrative view of how these processes are initiated and how mechanical forces are coupled across tissues to produce the correct OC shape.In new preprinted work, Ramos and colleagues (2025 preprint) take this integrative view, revealing that the apical surfaces of the NR and RPE are responsible for the onset of OV invagination through patterned mechanical forces. This shift in perspective, with apical forces taking the lead, invites us to rethink the steps underlying eye morphogenesis.The OVs emerge from the anterior neural plate as pseudostratified epithelia that rapidly reorganize into bilayers, with the distal layer fated to form the NR and the proximal layer the RPE. Importantly, the apical surfaces of both layers face the interior of the vesicle and are closely apposed, so that bending of the tissue ultimately involves coordinated changes on opposing apical and basal surfaces (Cardozo et al., 2023).Prior studies portrayed basal constriction in cells of the NR (Martinez-Morales et al., 2009; Nicolás-Pérez et al., 2016; Sidhaye and Norden, 2017) and flattening of the RPE layer, implicating its conversion from a pseudostratified to a squamous epithelium (Moreno-Mármol et al., 2021), as two autonomous and independent forces that together generate the tensions necessary for OV bending.Ramos and colleagues re-examine these events with high-resolution 3D morphometrics and time-resolved perturbations. Although they could confirm that basal actin and myosin are enriched in the NR and that RPE cells flatten during the OV-to-OC transition, their temporal and quantitative analyses argued that neither process is the primary initiator of OV folding. At odds with previous findings (Nicolás-Pérez et al., 2016; Sidhaye and Norden, 2017), they showed that basal acto-myosin enrichment in the NR is detected at least 2 hours before measurable invagination and does not change in distribution or intensity when curvature begins. They thus argue that basal constriction is an unlikely trigger of the first bending movements. Likewise, RPE stretching correlated with cup morphogenesis, but its curvature began in synchrony with, although independently of, that of the NR basal surface, indicating that RPE flattening supports and shapes folding but does not start it.A further mechanism, rim involution, in which peripheral cells from the outer layer migrate collectively into the NR through laminin-dependent basal protrusions (lamellipodia), was also thought to promote bending (Heermann et al., 2015; Soans et al., 2022). This epithelial flow contributes to NR growth and to the maintenance and deepening of the cup. However, Ramos et al. claim that rim involution primarily refines and sustains the forming cup but does not initiate the earliest bending events, although the reasons for this claim are unclear.For years, the formation of the lens placode in the abutting surface ectoderm has been proposed to exert a pushing force that drives OV folding, as its formation and inwards invagination occurs roughly in synchrony with OV bending (Magalhães et al., 2021). Yet organoid studies and direct tests argue otherwise: embryoid experiments can produce cup-like folds without a lens, and Ramos et al. showed that acute ablation of the lens placode does not reduce, but paradoxically increases, NR inward curvature. Thus, rather than pushing the OV to fold, the lens seems to be pulled inward as the OC forms. It is likely that the interaction of the lens placode with the OV and the surrounding extracellular matrix optimizes and coordinates folding but is not the initiating mechanical force.After ruling out the previously identified mechanisms as initiators of folding, Ramos and colleagues focused on the apical domains of the OV epithelia. High-resolution imaging of apical junctions revealed that the first detectable bending occurs at the apical surfaces of both the NR and the RPE, approximately 30-45 minutes before basal curvature becomes apparent. This observation places the origin of OC folding at the apical surface, shifting the focus from basal constriction or external pressure to mechanisms intrinsic to the apical layer.To explore how apical activity could generate a curvature, the authors employed TopoSPAM, a programmable spring-lattice model inspired by shape-morphing materials (http://topospam.org/; Singh et al., 2025 preprint). They considered three possible apical mechanisms: (1) a global mismatch in surface area between NR and RPE apical layers, (2) radial gradients of apical expansion, in which some cells expand their apical domains more than others across the tissue, and (3) anisotropic in-plane rearrangements, whereby cells exchange neighbors in a directional manner, introducing oriented strain. Each of these mechanisms could, in principle, bend the epithelium.Biological experiments narrowed down these possibilities. In the RPE, live imaging of mosaic-labeled cells revealed that curvature arises from non-uniform flattening, producing radial gradients of apical expansion across the tissue. In the NR, bending is instead driven by anisotropic rearrangements, consistent with elliptical strain patterns generated by directional neighbor exchanges. Laser ablation confirmed that NR and RPE can bend independently, as previously reported (Moreno-Mármol et al., 2021), even though their apical domains are mechanically connected, thereby excluding the simple mismatch model. These findings establish that each epithelium encodes its own apical program of deformation. The sufficiency of apical forces was then tested computationally and in vivo. In the computational model, introducing apical strain patterns alone reproduced basal bending, indicating that basal invagination is a downstream consequence. In embryos, perturbing the planar cell polarity pathway using a dominant-negative form of the Dishevelled protein, abolished NR apical rearrangements and prevented invagination. Similarly, morpholino-based knockdown of Shroom4, an apical actin-associated protein, produced the same effect, indicating that apical activities are not only correlated with but required for OC morphogenesis.Together, these data define OC invagination as the product of ‘spontaneous apical strain patterns’. In physics, such patterns are well-established in programmable materials such as liquid crystal elastomers or bilayer hydrogels, where local mismatches in expansion encode a target shape that emerges upon release of constraints (Ware et al., 2015; Plucinsky et al., 2018). Comparable principles have been described in epithelia: differential planar growth drives fold positioning in the Drosophila wing disk (Tozluoǧlu et al., 2019), and anisotropic cell shape changes mediate folding at the zebrafish midbrain–hindbrain boundary (Visetsouk et al., 2018).By showing that NR and RPE use independent apical mechanisms (non-uniform flattening and anisotropic rearrangements) that act as ‘spontaneous strain programs’, Ramos et al. extend these physical concepts to the developing vertebrate eye. Their integration of quantitative imaging, mechanical modeling and physical theory points to the logistic for the initiation of OC morphogenesis and new insight into how tissue-level self-organization may drive morphogenetic shape formation. The work suggests that biological form can emerge from cells that actively program their own mechanical behavior. Yet, several crucial questions remain unresolved. How does the system generate and accumulate the energy needed for apical expansion? Is this mechanism evolutionarily conserved in amniotes, in which the NR and RPE are not apically apposed? How is apical curvature transmitted to the basal side of the retinal neuroepithelium? These points are not trivial. Genetic evidence from the ojoplano (flat eye) medaka line, in which mutations in a negative regulator of integrin endocytosis, disrupts actomyosin contractility at basal end-feet leaving the OV flat (Martinez-Morales et al., 2009; Bogdanović et al., 2012), strongly support its essential role in OC formation. Similarly, focal perturbations of myosin II or microtubules in small patches of RPE block cell flattening and reduce OV folding, indicating that RPE flattening is cell-autonomous and a potent mechanical driver of OV folding (Moreno-Mármol et al., 2021). Addressing these questions through further analysis will be essential to fully elucidate how mechanical and molecular processes are integrated during OC morphogenesis.The Centro de Biología Molecular Severo Ochoa is a Severo Ochoa Center of Excellence (CEX2021-001154-S), funded by MCIN/AEI/10.13039/501100011033.