The Corneal Epithelium and Lens Develop Independently From a Common Pool of Precursors


  • Elodie Collomb,

    1. FRE CNRS 3405, AGIM, Université Joseph Fourier Grenoble, Site Santé, France
    Search for more papers by this author
    • Drs. Collomb and Yang contributed equally to this work.

  • Ying Yang,

    1. FRE CNRS 3405, AGIM, Université Joseph Fourier Grenoble, Site Santé, France
    Current affiliation:
    1. Institut de la Vision, Institut National de la Santé et de la Recherche Médicale, Paris, France
    Search for more papers by this author
    • Drs. Collomb and Yang contributed equally to this work.

  • Sarah Foriel,

    1. FRE CNRS 3405, AGIM, Université Joseph Fourier Grenoble, Site Santé, France
    Search for more papers by this author
  • Sébastien Cadau,

    1. FRE CNRS 3405, AGIM, Université Joseph Fourier Grenoble, Site Santé, France
    Current affiliation:
    1. Centre Hospitalier Affilié Universitaire de Québec, Pavillon Enfant-Jésus, Québec, Canada
    Search for more papers by this author
  • David J. Pearton,

    1. FRE CNRS 3405, AGIM, Université Joseph Fourier Grenoble, Site Santé, France
    Current affiliation:
    1. AgResearch Ltd, Ruakura Research Centre, Hamilton, New Zealand
    Search for more papers by this author
  • Danielle Dhouailly

    Corresponding author
    • FRE CNRS 3405, AGIM, Université Joseph Fourier Grenoble, Site Santé, France
    Search for more papers by this author

Correspondence to: Danielle Dhouailly, UJF, Site santé, Centre Jean Roget, La Tronche, 38706, France. E-mail:


Background: The corneal epithelium (CE) overlays a stroma, which is derived from neural crest cells, and appears to be committed during chick development, but appears still labile in adult rabbit. Its specification was hitherto regarded as resolved and dependent upon the lens, although without experimental support. Here, we challenged CE fate by changing its environment at different stages. Results: Recombination with a dermis showed that CE commitment is linked to stroma formation, which results in Pax6 stabilization in both species. Surgical ablation shows that CE specification has already taken place when the lens placode invaginates, while removal of the early lens placode led to lens renewal. To block lens formation, bone morphogenetic protein (BMP) signaling, one of its last inducing factors, was inhibited by over-expression of Gremlin in the ocular ectoderm. This resulted in lens-less embryos which formed a corneal epithelium if they survived 2 weeks. Conclusion: The corneal epithelium and lens share a common pool of precursors. The adoption of the CE fate might be dependent on the loss of a lens placode favoring environment. The corneal fate is definitively stabilized by the migration of Gremlin-expressing neural crest cells in the lens peripheral ectoderm. Developmental Dynamics 242:401–413, 2013. © 2013 Wiley Periodicals, Inc.


The ocular surface ectoderm is a multipotent region of head ectoderm that gives rise to the lens, the corneal and conjunctival epithelium as well as the lachrymal glands, and the eyelid epidermis. Here, we examine the mechanism of ectodermal specification and commitment toward a corneal epithelium (CE) fate. The territory of the presumptive CE can be identified, although not peripherally delimited, when the lens vesicle detaches from the ectoderm. In the chick embryo, this event is followed by two waves of migration of neural crest cells into the space between the ectoderm and the lens. These cells will subsequently become the endothelium and the stromal fibroblasts (Hay, 1980). The stages of chick anterior eye segment development are presented in the diagram which summarizes our results (Fig. 6). The fully formed cornea is composed of three layers: a stratified epithelium expressing the K3/K12 keratin pair (O'Guin et al., 1987) overlaying a stroma composed of highly-aligned collagen sparsely populated by fibroblasts and with a single layer endothelium underneath. The cornea is the transparent part of the integument, characterized by the expression of crystallin-type proteins (for a review: Jester, 2008), including classical lens crystallins (Ren et al., 2010). Moreover, its development and postnatal maintenance, similarly to the lens, depends on the activity of Pax6 (for a review: Ashery-Padan and Gruss, 2001). Chick corneal epithelium commitment was found to be autonomous from its stroma by the first week of development (Zak and Linsenmayer, 1985), whereas experiments from our group have shown that in mammals the fate of corneal epithelium can be changed to epidermis even in adult tissue, when associated with an embryonic dermis (Ferraris et al., 2000), and that this is dependent on the down-regulation of Pax6 in the induced hair placodes (Pearton et al., 2005). We believed that this difference in the results between the two different species required further examination, particularly with respect to the role of Pax6 expression in chick corneal epithelium and how it responds to a new environment.

Corneal induction is usually considered to be the last in a series of inductive events in eye development, with the lens vesicle interacting with its overlying surface ectoderm (for recent reviews: Moose et al., 2009; Graw, 2010). However, there is no experimental support for this hypothesis, which is based only on the close association of these two tissues (Amprino, 1949, reviewed in Hay, 1980). Likewise, for almost a century, lens induction was thought to occur at the time the optic vesicle and lens placode came in contact (Spemann, 1901). This view has subsequently changed in the light of new evidence. First, cells fated to become lens are part of a preplacodal domain in the neurula ectoderm (for review: Streit, 2004; Bhattacharyya and Bronner-Fraser, 2004). Second, the optic vesicle (OV) is not the primary inducer, but rather a contributor to the final phase of lens formation (work of the Grainger group reviewed in Sullivan et al., 2004), Furthermore, Bmp4 and Bmp7 knockout studies in mice, and loss- of function experiments in chick have shown that bone morphogenetic protein (BMP) signaling is essential for lens formation (Dudley et al., 1995; Furuta and Hogan, 1998; Wawersik et al., 1999; Pandit et al., 2011). Despite such studies in mice and chickens, the possible effects of the absence of the lens on corneal development has not been adequately studied. A recent study showed that the lens vesicle may not be required for CE differentiation given that the expression of keratin 12 (K12) is unaltered in transgenic mice where the lens fibers are destroyed. This was obtained by expression of an attenuated version of diphtheria toxin (Tox176) driven by a hybrid promoter that is active in the fiber cells of the lens (Zhang et al., 2007). This result suggested that the lens placode, rather than the lens vesicle, might be critical for CE specification. When the expression of Tox176, is driven by a modified Pax6 promoter, the lens placode is permanently removed, and skin forms instead of cornea (Zhang et al., 2008b). However, the later result has two possible interpretations. As Pax6 is required both for lens placode and CE formation, and is still required in these two structures in adulthood, the Pax6 promoter will drive expression in both, thus: either both lens and CE precursors are killed, leaving open the possibility that the lens placode might be required for cornea formation, or lens and CE have the same precursors. It does appear from this that corneal epithelium and lens placode fates are inextricably linked.

The question is thus still open: how and when is corneal fate specified and subsequently committed? In other words, what are the steps involved in the induction of corneal epithelium and how are they integrated with the other aspects of eye development, in particular those of the lens? We show here that corneal epithelium commitment is linked to stroma formation and stabilization of Pax6 expression, but that its specification occurs well before the appearance of the lens placode. We propose that both CE and lens arise from a common pool of precursors and that the final partition between CE and lens occurs when the peripheral ocular ectoderm no longer receives the signals responsible for the final phase of lens induction.


Chick Corneal Epithelium Is Committed Several Days Before Its Differentiation

The differentiated corneal epithelium expresses the K3/K12 keratin pair, rather than the K5/K10 characteristic of the epidermis (O'Guin et al., 1987). Moreover, the differentiated CE does not possess the anucleate stratum corneum, nor form cutaneous appendages, such as feathers, which are features of epidermis. In both the embryo and the adult, a key difference between those two epithelia is that only the CE expresses Pax6, the master gene of eye formation (Gehring and Ikeo, 1999). Among other functions, Pax6 acts as a co-activating factor for the activation of the K12 gene (Liu et al., 1999). During chick embryonic development, K12 expression indicative of CE differentiation starts at E12 and K12 is expressed in all the corneal epithelial cells by embryonic day (E) 14 (Chaloin-Dufau et al., 1990). This is, however, 7 days after the formation of the three corneal cell layers (at E5) and a full 9 days after the initial formation of the CE (after detachment of the lens vesicle at E3) (Hay, 1980). To determine when the corneal epithelium is committed, we challenged its ability to maintain or change fate in the context of a novel embryonic environment by performing recombinants with an E7 dorsal dermis. The inductive capacity of embryonic dermis at E7 in chick, as well as at a corresponding stage in mouse, has been extensively studied since its discovery (Dhouailly, 1977), and shown to be linked to the Wnt/beta catenin pathway (among others: Gat et al., 1998; Zhang et al., 2008a; Chang et al., 2004). A mouse dermis at this stage is able to transform an ectoderm into an epidermis (Fliniaux et al., 2004) and to initiate cutaneous appendage morphogenesis, even in an adult mammalian corneal epithelium (Ferraris et al., 2000; Pearton et al., 2005).

Due to the long delay between tissue recombination and the initiation of K12 expression a long period of in vivo culture (2 weeks) is required to detect it. When using a CE of E5 and E7, this was done by grafting the epithelial/mesenchymal recombinants (2 mm2 in size) under the kidney capsule of nude mice. When E7 chick corneal epithelium was associated with E7 dorsal dermis (n=8), the fate of the CE was not changed: Pax6 and K12 were expressed throughout the whole epithelium and no epidermal differentiation was observed (Fig. 1A–A′; Supp. Fig. S1A, which is available online). When E5 chick CE was associated with E7 chick dorsal dermis (n=6), the resulting epithelia never formed feathers and showed tiny restricted areas of non–Pax6-, non–K12-expressing cells, but most of the epithelium maintained Pax6 and K12 expression (Fig. 1B–B′). At E3, the CE is very small, but the recombination could be done directly in ovo, as the migration of stromal cells has not yet occurred (Fig. 6; and Hay, 1980). Thus, to examine the potential commitment of CE at E3, we replaced the lens vesicle with an equivalent volume of inductive (E7) dermis and recovered embryos within 4 hr, after 12 hr and after 11 days. The expression of Pax6 in the overlying ectoderm started to decrease within a few hours of the lens replacement (compare Fig. 1C,D), and was completely lost by 12 hr (Fig. 1E), while the expression of Pax6 in the retina was unchanged (Fig. 1E′). When embryos were grown until E14, the entire CE was transformed into an epidermis complete with feathers (Fig. 1F–I; n=16) with no sign of K12 expression (data not shown). It should be noted, however, that the eyelids formed even in the absence of cornea formation (Fig. 1F). An alternative explanation to induction of epidermis by the embryonic dermis in this case might be that the implanted dermis is inhibiting the migration of the neural crest-derived stroma, which may be at least partially responsible for directing CE formation. To test whether the stroma is able to induce CE, naive (uncommitted) E7 chick dorsal epidermis (one epidermal cell layer) was recombined with the central (n=9) or peripheral (limbal, n=11) corneal stroma from an E7 chick embryo, and grafted under the kidney capsule of nude mice. The grafts were recovered after 15 days and showed that a single cell layer of embryonic epidermis had developed to form several strata including a thick keratinized stratum corneum (Fig. S1B), and showed no expression of Pax6 nor K12 (Fig. S1C).

Figure 1.

During the fifth day of incubation, the chick corneal epithelium loses its capacity to down-regulate Pax6 and becomes committed. A–B′: Two weeks after the recombination of embryonic day (E) 7 and E5 corneal epithelium (ce) with E7 chick dorsal dermis (de), grafted under a nude mice kidney capsule. A–A′: At E7, the corneal characteristics are maintained throughout the epithelium. B–B′: At E5, the corneal epithelium can be still partially transformed into an epidermis (epd), as shown by a tiny patch of epithelium which has lost both the expression of Pax6 and K12. It never formed feather buds (the round structures are a result of twisting of the recombinant). C–I: At E3, the corneal epithelium fate is totally labile. In the control left eye (C), at E3+4h, Pax6 is expressed in all the eye structures, lens vesicle (lv), retina (r), and prospective corneal epithelium (ce). In the operated right eye (D), 4 hr after replacement of the lens vesicle in ovo by a piece of E7 dermis (de), Pax6 expression has already decreased in the ocular epithelium (arrow), while it is not changed in the retina. After 12 hr (E), the eye ectoderm (ect) overlying the dermis is negative for Pax6, while a lower magnification (E′) involving the retina shows that this non CE structure continues to express Pax6. F–I: After eleven days, a feathered skin has formed instead of a cornea, while the morphogenesis of eyelids (eyl) has occurred (F). Whereas the retina (r) expressed Pax6 (G), the ocular ectoderm was transformed into an epidermis (epd) with feathers (fs). G–I: Those feathers did not express Pax6 (G, H) and reached the stage of barb ridge (br) morphogenesis (H,I). Immunofluorescence with Pax6, K12 and αK monoclonal antibodies.

The chick corneal epithelium at E5 is therefore almost completely committed, with only a few cells still able to down-regulate Pax6 expression under the influence of a dermis, whereas at E3 its fate can be changed to that of a feathered epidermis. Moreover, the nonresponsiveness of the reverse recombinant (corneal stroma with uncommitted epidermis) means that noncranial ectoderm, which does not express Pax6, is not competent to respond to stromal signals, and/or that corneal stroma, even from the limbal region, is not able to induce Pax6 expression in an associated epithelium.

Lens Formation Is Not Required for Corneal Epithelium Specification

The lens has been postulated to be critical for CE formation but the experimental evidence for this hypothesis is lacking. To test whether the CE can form in the absence of lens, we decided to surgically remove the lens anlagen at different steps of its development. After ablation of the right lens vesicle at Hamburger and Hamilton (HH) stage 15 (25–27 somites, n=8) (compare Fig. 2A,B), small right eyes developed (Fig. 2H); the lens did not form and the retina was extensively folded (Fig. 2C). By E5, corneal stromal precursors had migrated to fill the space underneath the ectoderm; suggesting that cell migration occurred earlier and appeared to involve more cells in the “lens-less” eye than in the contralateral, unoperated eye (Fig. 2DE). Histological evidence suggests that in the operated eye migration of neural crest cells to form the stroma occurs in a single step rather than in two separate waves as happens during normal development. Furthermore, in contrast to the control eye (Fig. 2F), the operated eye (n=10) did not form a distinctive corneal endothelium located below the stroma by E7 (Fig. 2G). The anterior eye chamber did not form and aberrant connections were formed with the future iris area. Despite this aberrant morphological development and the small size of the eye (Fig. 2H), the CE was well differentiated in all cases, expressing K12 by E14 (Fig. 2J) similarly to the control eye (Fig. 2I). The lens vesicle is therefore not required for CE specification. However, the postulated “lens cornea induction” suggested by the close contact between lens vesicle and the surface ectoderm, might occur at an earlier point by means of polar induction within the invaginating lens vesicle, or even at the stage of the lens placode, with its peripheral ectoderm.

Figure 2.

Surgical ablation of the chick lens vesicle at Hamburger and Hamilton (HH) stage 15 does not prevent the differentiation of corneal epithelium, but inhibits the formation of endothelium. A–C: At HH stage 15, the lens vesicle (arrow) is detached from the head ectoderm (A). When the lens vesicle is ablated (arrow) through a split in the ectoderm (B), no lens regeneration has occurred four days later (C), but the retina (r) was folded, and a cornea (c) had formed. D–J: Comparison of corneal morphogenesis in the control left eye (D,F,I) and in the lens-less ablated right eye (E,G,J). By E5, in the control (D) the acellular stroma (ast) is formed and limited by a single layer of fibroblasts, corresponding to the first mesenchymal cell migration, while in the lens-ablated eye (E), a large wave of fibroblast migration has occurred. By E7, the second fibroblast cell migration occurred in the control (F), and the corneal stroma (st) is well delimited by the formation of an endothelium (end). In contrast, the operated eye (G) showed no evidence of endothelium, but abnormal connections between the stroma and the retina have formed, with a consequent lack of the anterior chamber (ac). H: At E14, the right eye (black arrow) is smaller than the left control eye (white arrow) In the case of lens ablation (J), a well differentiated corneal epithelium (ce) expressing keratin K12 was formed, as also occurred in the control (I). Hematoxylin staining (C–G), Immunofluorescence with K12 monoclonal antibody (I,J). le, lens; ir, iris; on, optic nerve.

When surgery was conducted at early stage HH14 (21–22 somite; n=30), the lens pit as well as its surrounding ectoderm were ablated (Fig. 3A–B′). At this stage, three different phenotypes were obtained in the embryos recovered from E4 to E14. In 30% of cases, there was formation of a microphthalmic eye where the lens did not regenerate (Fig. 3C–G). At E14, three cases resulted in apparently eyeless embryos. Histological examination of these revealed a total lack of a lens, and a tiny structure comprising a scleroma surrounding both neural and pigmented retina (Fig. 3E). Use of a polyspecific αKeratin antibody, which stains most epithelia, identified a cavity connected with the epidermis localized on top of the tiny eyeball (Fig. 3F). Labeling with the anti-K12 specific antibody showed that the inner part of this cavity was fringed by a K12 positive differentiated corneal epithelium (Fig. 3G). Surprisingly, the lens renewed in 70% of cases. In half of these cases, the resulting right eye was quite similar to the left control eye (Fig. S2A). In the other half, the right eye was abnormal and smaller than the control left eye (Fig. S2B), the lens remained open (Fig. S2D,E) and did not detach from the ectoderm. However, the later still differentiated into a corneal epithelium, which expressed K12 (Fig. S2C–G).

Figure 3.

Surgical ablation of the chick lens pit at Hamburger and Hamilton (HH) stage 14 can lead to a microphthalmic lens-less eye, without preventing the differentiation of corneal epithelium. A–C′: At HH stage 14, the lens placode is invaginating (inv lp) (A,A′); the ablation (B,B′), not only of the lens pit, but also of its peripheral ectoderm (ect) in front of the optic cup (oc) resulted in lens-less eye at E4 (C,C′). D: At E6, compare the size of the control (white arrows) and of the operated eye (black arrow). E–G: At E14, the lens-less eye consists of a tiny black ball formed by nervous (nr) and pigmented (pr) retina, surrounded by a scleroma (scl) (E). However, note a cavity fringed by an epithelium (ept) contiguous with the epidermis (epd), both being labeled by the αK antibody (F). The labeling of Keratin 12 allowed for the identification of differentiated corneal epithelium (ce), as the deeper part of this cavity (G). eyl: eyelid; fs: feathers; spl: split. A′,B′,C′,E: Hematoxylin staining. Immunofluorescence with αK (F) and K12 (G) monoclonal antibodies.

Based on previous results in chick embryos (Hyer et al., 2003; Dias da Silva et al., 2007), we would expect no lens to form after the ablation of the placodal ectoderm at early HH stage 13 (18 somites). To fully remove the lens placode, we removed a large, regular section of the ectoderm overlying and peripheral to the apposed right optic vesicle (Fig. 4A,B). This was easy with the use of Nile blue. When this was performed in ovo, the survival rate was very low so an embryo culture technique was used that resulted (after washing out all the Nile blue) in all the ablated embryos surviving for at least 38 hr and hence gave interpretable data. One hour after lens placode ablation, the surface ectoderm had already begun to heal and had covered the base of the optic cup (Fig. 4C; n=3). Three to 6 hr later (n=6), on the operated side, the ectoderm was either still migrating or had already healed (Fig. 4D), and the optic vesicle had transformed into an optic cup, whereas the control eye showed a lens vesicle and optic cup formation (Fig. S3A–C). Seventeen to 38 hr after surgery, six of nine survivors formed a typical lens placode (Fig. 4E) or vesicle (Fig. 4F) in the operated eye. It should be noted that the renewed lens vesicle was smaller and appeared later than the lens of the contralateral eye. In only three cases, the general development of the embryo was delayed, the un-operated eye had formed a lens placode, and no lens placode was distinguishable on the operated side, and they consequently cannot be informative.

Figure 4.

At early Hamburger and Hamilton (HH) stage 13 (18 somites), the chick head ectoderm is able to reform a lens. A,B: At the 18 somite stage, it was easy to ablate the ectoderm (ect), not only corresponding to the lens-placode, but also until a ring largely outside (arrow and white dashes) its contact with the optic vesicle (ov). C: One hour after the ablation, the ectoderm was already starting to recover (arrow and white dashes) the optic vesicle. D: Then, at 6 hr after ablation, the healing ectoderm has completely recovered the optic cup (oc), which has differentiated on the operated right side in the absence of lens vesicle. E,F: Then, the right operated side regrew first a lens anlage (arrow), and finally a detached lens vesicle (lv) (arrow), respectively, at 17 and 30 hr after ablation. Hematoxylin staining (B,D).

In brief, although some cases showed CE differentiation in the absence of lens after lens pit ablation at HH stage 14, the surgical removal of lens placode at HH stage 13 led to de novo lens formation in 100% of cases. It was, therefore, impossible to disentangle the role of the early stages of lens development on CE specification using this approach, so we adopted a strategy of developmentally inhibiting the formation of the lens placode.

Inhibition of BMP Signaling Prevents the Formation of Lens But Not of Corneal Epithelium

We decided to interfere biochemically with one of the final steps in the formation of the lens placode so as to prevent its formation. In chick, similar to the mouse, BMP signaling has been shown to be a lens-promoting event (for a review: Sullivan et al., 2004; Gunhaga, 2011). The activity of BMPs is well known to be modulated by extracellular binding proteins, including Noggin and Gremlin (reviewed by Mehler et al., 1997). Loss-of-function experiments have previously been carried out in chick embryos at HH stage 10–13 (Adler and Belecky-Adams, 2002; Huillard et al., 2005) and HH stage 11–17 (Pandit et al., 2011) by overexpression of these genes in the intra-optic vesicle. Those treatments resulted in various degrees of smaller lenses with complete lens loss in the most severely affected embryos. The variation in the results at these different stages of treatment emphasizes that the timing is crucial in these experiments.

To permanently remove the lens placode we, therefore, decided to inhibit BMP signaling using in ovo electroporation of the BMP antagonist Gremlin into the ectoderm facing the right optic vesicle at HH stage 10 (Fig. S4A), i.e., well before placode formation, which occurs at HH stage 13. Fourteen hours after electroporation, Gremlin expression was more pronounced on the electroporated side than in the control (Fig. S4B). It should be noted that in normal embryos, Gremlin expression occurs at HH stage 16 surrounding the lens vesicle (Fig. S4C and Tzahor et al., 2003). Twelve hours later, the majority of the embryos had died and showed a severe reduced head size phenotype, which can be explained by the fact that the BMPs have pleiotropic roles in head morphogenesis. As our goal was to recover E13 to E15 embryos to be able to identify a differentiated corneal epithelium by the expression of K12, we examined the manipulated embryos each day and recovered only those which appeared to be dying. At time zero plus 4 to 8 days, the right eye of 30% (n=18) of recovered embryos was clearly microphthalmic (Fig. 5A). Among the five embryos which survived until E14/E15, three appeared to be eyeless on the right, electroporated side (Fig. 5B), with a normal eye on the left, nonelectroporated side. At a higher magnification, however, tiny eyelid openings were visible on the right side of the head (Fig. 5C,D). Dissection and examination revealed the presence of tiny black balls which, on histological study, proved to be composed of both neural and pigmented retina, surrounded by a scleroma (Fig. 5E). Most importantly, no lenses were present. The eyelid openings were connected by a sack-like structure lined by an epithelium that stained with a polyspecific keratin antibody (Fig. 5F). Labeling with the anti-K12 specific antibody showed that the inner part of the conjunctival cavity was fringed by a K12 positive differentiated corneal epithelium (Fig. 5G).

Figure 5.

Differentiation of a corneal epithelium, and retinal tissue in the lens-less microphthalmic eye after RCAS-Gremlin overexpression at Hamburger and Hamilton (HH) stage 10. A: At embryonic day (E) 8, a frontal view allows a comparison of the right microphthalmic eye (arrow) and the left control eye. B–G: At E15. B: The eye appeared totally absent (arrow) on the right side. Note the crossed beak. C,D: Two higher magnifications allowed a detection of the eyelid split (spl), and the eyelids (eyl) which were deprived of feathers (f). E: Note the absence of lens. The different components of the lens-less eye involved a tiny black ball formed by nervous (nr) and pigmented (pr) retina, surrounded by a scleroma (scl). F: A sinuous cavity started from the eye split (spl), fringed by an epithelium (ept) contiguous with the epidermis (epd), both being labeled by an anti-epithelial keratin antibody (αK). G: Labeling of Keratin 12 allowed identification of the differentiated corneal epithelium (ce) as a part of the conjunctival sack. np, nasal pit; b, beak. Hematoxylin staining (E) Immunofluorescence with αK (F) and K12 (G) monoclonal antibodies

Figure 6.

From eye ectodermal precursors at Hamburger and Hamilton (HH) stage 10 to committed corneal epithelium (CE) at embryonic day (E) 5. The Pax6 ectodermal cells which segregate from the Dlx5 nasal precursors at stage HH 8 (see Battachyyara et al., 2004) can be identified as ocular ectodermal precursors as they can give rise either to lens or corneal epithelium. During HH stage 14, the final lens bone morphogenetic protein (BMP) promoting event finishes. The commitment of CE starts by E5 and correlates with neural crest migration which forms the stroma and prevents Pax6 down-regulation. ce, corneal epithelium; end, endothelium; ivp, invaginating pit; le, lens; nr, neural retina; oc, optic cup; ov, optic vesicle; pr, pigmented retina; 1st, first stroma; st, stroma.


The corneal epithelium and lens both arise from the ectoderm and share several other important characteristics; they are both transparent and require sustained Pax6 expression even in the adult. Our results, when combined with other work demonstrating the specification of the lens at an early neurula stage (for a review, see Streit, 2007), give a comprehensive account of corneal epithelium development from early specification through to commitment (Fig. 6) and demonstrate its developmental relationship with other eye structures. In addition to our primary conclusions, we also show that not only the cornea, but also the invagination of the optic vesicle and the development of the eyelids in chick, are independent of lens morphogenesis, as is the case in mice (Zhang et al., 2008b; Swindell et al., 2008). That is, the transformation, of the optic vesicle into an optic cup occurred even in the absence of lens placode formation, and eyelids formed in lens-less chick embryos.

Commitment of Corneal Epithelium Correlates With Both the Loss of Its Capacity to Down-regulate Pax6 as Well as to the Formation of the Stroma in Chick, Similarly to Rabbit

Heterotypic mesenchymal/epithelial recombinations performed with chick embryos have shown previously that development of the avian CE is autonomous from its underlying stroma several days before its final differentiation (Zak and Linsenmayer, 1985), but that its fate is still labile on day 3 of incubation (Coulombre and Coulombre, 1971). Here, we confirm and extend those results, showing that the commitment of the chick CE correlates with sustained Pax6 expression and stroma formation from E5. This is only after corneal commitment which appears superficially different to what we have previously shown in mammals. Indeed at an early stage in rabbits, i.e., E12.5 (Chaloin-Dufau et al., 1990) which, for the development of eye, corresponds to E3 in chicks, the mammalian CE is able to be directly de-specified and develop into an epidermis when associated with a competent dermis (Ferraris et al., 1994). This ability is linked to a rapid (within a few hours) loss of Pax6 expression as shown here for chick and also confirmed in rabbit (E. Collomb, unpublished data). By contrast, at later developmental stages in the rabbit, i.e., after corneal stroma formation at E17 (Chaloin-Dufau et al., 1990), the sequence for formation of an epidermis differs and occurs by means of a process of transdifferentiation. The committed rabbit CE dissociated from its stroma at E20 (E. Collomb, unpublished data), at E23 (Ferraris et al., 1994), or adult (Ferraris et al., 2000; Pearton et al., 2005) is able to give rise to an epidermis, but only indirectly and after a long period (3 weeks) of association with a cutaneous appendage forming dermis, with all ages following the same steps and timetable. In all cases, the down-regulation of Pax6 is a prerequisite for the development of the new epidermis. This down-regulation occurs only in patches of the basal layer of the epithelium which proliferate to give rise to hair pegs, whereas Pax6 expression is retained in the inter-follicular basal layer as well as in all upper epithelial cell layers (Pearton et al., 2005). The new epidermis derives from the new hair stem cells (Pearton et al., 2005), which arise as a consequence of hair peg formation (Nowak et al., 2008; S. Cadau, unpublished data). The re-entry of corneal epithelial cells into the cell cycle requires the loss of Pax6 activity (Li and Lu, 2005), and this might thus be a prerequisite for their loss of commitment. We suggest that this can be induced by embryonic dermal instructions in mammals where hair placode formation occurs by means of cell division. This is in contrast to the case in birds where feather placode formation occurs by means of cell rearrangement and elongation rather than cell division (Dhouailly, 1977). This lack of re-entry into the cell cycle might explain the inability of the older chick CE to undergo the same transdifferentiation process as the older rabbit CE. Briefly, when the corneal stroma forms, in mammals as in birds, Pax6 expression appears to be stabilized in corneal epithelium, at which point it can be considered to be committed.

Corneal Stroma Is Involved in Stabilizing Pax6 Expression

The CE plays a dominant role in promoting the nondermal fate of the stroma, as previously shown in mice embryos. It controls the expression of several genes in the migrating ocular mesenchyme (Matt et al., 2005), which represses Wnt signaling in this mesenchyme (Kumar and Duester, 2010), the initial promoting factor of skin development (Chang et al., 2004). In the reverse case, we have here demonstrated that the embryonic corneal or limbal stroma in chick, as well as in rabbit (our unpublished results), is unable to induce Pax6 expression in an early single layer embryonic dorsal epidermis. In contrast, several studies have examined the potential of stem cells—both embryonic as well as epidermal tissue-specific, or induced pluripotent stem cells—to be directed to a corneal fate by being exposed to a corneal stromal environment (among others: Homma et al., 2004; Ahmad et al., 2007; Blazejewska et al., 2009; Shalom-Feuerstein et al., 2012). Of interest, the last authors recently showed that a medium conditioned by corneal fibroblasts induces a decrease in miR-450b-5p, which is involved in the maintenance of the pluripotent state by repressing Pax6. Consequently, corneal stroma effect might be to prevent Pax6 down-regulation after it has been turned on during the early neurula stage in the head ectoderm by a complex interaction of genetic pathways (for a review: Lang, 2004).

Lens Vesicle Has a Limited Role in Corneal Endothelium Morphogenesis, Mesenchymal Cell Migration, and Eye Growth

We show that in chick, as in mice (Zhang et al., 2007), the lens vesicle is not required for the last step of CE differentiation, as measured by the expression of the corneal specific keratin 12, but is critical for corneal endothelium morphogenesis. This is consistent with the observation that in human congenital aphakia (absence of lens) the CE formed, but the corneal stroma is not bounded by an endothelium (Valleix et al., 2006). We confirm that lens vesicle has an inhibitory effect on early migration of neural crest cells before E5 (Lwigale and Bronner-Fraser, 2009). We also confirm that the lens vesicle has a primary role in the general growth of the eye. With the lens being a primary source of the pro-growth FGF (for reviews Govindarajan et al. 2000; Robinson, 2006): the removal of the lens vesicle, lens pit, or the prevention of lens formation leads respectively to the formation of a small, microphthalmic or an “apparently absent” eye.

Lens Pit or Lens Placode Are Not Required for Corneal Epithelium Specification

The surgical excision of the lens pit as well as its peripheral ectoderm covering the optic cup at HH stage 14, resulted in lens renewal in 70% of cases. The particular stage targeted during HH stage 14 might explain the variation of the results. A possible explanation is that at one particular moment during HH14, all of the potential lens cells are contained in the volume of cells excised, or, more likely, that the final lens promoting event is over. Indeed, the 30% of cases where a lens-less microphthalmic eye, which nevertheless still developed a cornea, was formed might indicate that some “eye ectodermal precursors” remain.

As our surgical ablation of placodal ectoderm at HH stage 13 invariably led to lens renewal, we decided to definitively prevent the “lens placode favoring events” discussed in (Sullivan et al., 2004). In chick, lens placodes do not form if the OV is ablated (Kamachi et al., 1998), but in contrast to mice which express Bmp4 in the OV (for a review: Lang, 2004; Sullivan et al., 2004; Reza and Yasuda, 2004), Bmp4 is only expressed by the preplacodal ectoderm (Trousse et al., 2001). Of interest, BMP receptor distribution in not only the preplacodal ectoderm but also in the OV appears similar in chickens and mice (among others: Furuta and Hogan, 1998; Trousse et al., 2001; Belecky-Adams et al., 2002). Furthermore, the culture of chick eye ectodermal progenitor cells showed that the generation of lens cells requires continued exposure to BMP (Sjödal et al., 2007; Pandit et al., 2011). Gremlin is well known to bind and inhibit the action of BMPs (among others: Hsu et al., 1998) and, in addition, it specifically binds to the BMP4 precursor protein inside cells, which prevents the production of mature BMP4 protein, and thus functions as a highly efficient intracellular BMP4 antagonist (Sun et al., 2006). By overexpressing Gremlin in the eye surface ectoderm, we expected to prevent the BMP4 production by the ocular ectoderm. A detailed analysis of BMP function by means of the study of ocular ectodermal gene expression was, however, hindered by the high lethality of the manipulation. Most of the dead embryos, however, lacked a lens even if other eye structures were present, and those that reached E14/E15 showed a lens-less eye with a differentiated corneal epithelium expressing K12. Thus even if lens placode formation was prevented, we did not stop the formation of corneal epithelium.

Corneal Epithelium Precursors Are Specified Early, Concurrently With Those of the Lens and Thus Constitute a Pool of Eye Ectodermal Precursors

Our results are in accordance with previous findings showing that the potential to form a lens is not restricted to the ectoderm facing the optic vesicle, as a broader region of the chick head Pax6-expressing ectoderm is capable of forming a lens (Barabanov and Fedtsova, 1982; Bailey et al., 2006). The migration of neural crest cells, which express Gremlin and Noggin (Bardot et al., 2001; Tzahor et al., 2003), is known to inhibit lens placode formation outside the ocular region (Bailey et al., 2006). Their arrival, in periphery of the lens, might thus contribute to the stabilization of the corneal epithelium fate of this ectodermal area. Likewise, following removal of the lens in Xenopus tadpoles, the corneal epithelium reforms a new lens (first described by Freeman [1963], and reviewed by Henry and Tsonis [2010]), and this requires BMP signaling (Day and Beck, 2011).

It should be noted that in Xenopus the cornea is not fully differentiated until after the process of metamorphosis is completed, suggesting that lens renewal from corneal epithelium appears to be prevented in adult and older tadpoles by stroma formation (Reeve and Wild, 1981). In chick, earlier experiments showed that removal of the placodal ectoderm at early HH stage 13 (Hyer et al., 2003) leads to loss of lens formation. This is in contrast to what occurred in our experiments where new lens formation was observed. We suggest that the Hyer group removed not only the ectodermal cells strictly fated to be lens, but also almost all of those fated to give the corneal epithelium. The excision performed was larger than ours, seeming to comprise almost half of the head ectoderm, and this, added to the toxic effect of the Nile blue, might have prevented lens placode renewal. Moreover, in the majority of their experiments (Hyer et al., 2003), the wound remained open over the course of the experiment.

In our experiments, by contrast, wound healing occurred in 100% of cases, starting immediately after excision, and was completed within 6 hr. Given this timeframe, it is clear that the ectodermal cells sealing the wound come from migration and not from proliferation, as by a HH stage close to 13, chick embryo cells proliferate with an average doubling time of 12 hr (McMaster and Modak, 1977). This migration allowed the ectodermal cells to arrive adjacent to the optic vesicle in time to potentially receive any remaining lens promoting signals, which enhance, among others, L-Maf expression which is required for placodal cell fate (review in: Reza and Yasuda, 2004). If we had managed to remove all of the lens-fated cells in our ablation of the placodal ectoderm then this suggests that those remaining cells originally destined to be corneal epithelial cells might have the capacity to give rise both to lens and cornea. The new lens thus does not “regenerate” in the classical sense, rather there is a re-distribution of the remaining cells of a precursor cell pool which normally contribute both to the lens and corneal epithelium.

Conclusion (Fig. 6)

We suggest that most of the cells expressing Pax6, that have been shown to segregate from the olfactory precursors at HH stage 8 (Bhattachyyara et al., 2004), form a common pool that we propose to be “eye ectodermal precursors.” These eye ectodermal precursors can give rise to either lens or corneal epithelium, depending on the presence or absence of a final BMP signal, which promotes lens placode formation. The loss of this BMP signaling in the wake of lens vesicle formation combined with the migration of neural crest cells expressing BMP inhibitors (Bardot et al., 2001; Tzahor et al., 2003), and the production of retinoic acid (Matt et al., 2005), which is known to interfere with BMP signaling (Shenga et al., 2010), might contribute to the stabilization of the corneal epithelium fate. Our results are in support of the previous “refracton hypothesis” (Piatigorsky, 2001), which states that similar genetic programs may be responsible for the development of both the lens and corneal epithelium. In fact, CE shares the same program to the lens ectoderm until almost the end: CE specification is essentially the penultimate step in lens specification. Finally, we show here that, after having been earlier specified as eye ectodermal precursors, and persisting in this state after lens placode formation, corneal epithelium is finally committed at the stage of stroma formation, as determined by the persistence of its intrinsic Pax6 signaling.



Fertilized eggs (JA957 strain, St Marcellin, France) for microsurgery and fertilized SPF (pathogen-free) White Leghorn eggs for in ovo electroporation (Cerveloup, Vourey, France) were incubated at 38°C until the embryos reached the required Hamburger Hamilton stages (Hamburger and Hamilton, 1951). HH staging on the second day of incubation was determined by the number of somites: i.e., from 10 somites (HH stage 10) to 26 somites (HH stage 15). Eggs were opened, the embryos moistened with phosphate buffered saline (PBS), and the eggs sealed with tape.

Preparation and Grafting of Epithelial/Mesenchymal Heterotopic Recombinants

Recombinants were performed between both corneal epithelium and stroma, and either skin, epidermis and dermis, components. The embryonic CE from E5 to E7 chick embryos was dissected out of the relevant embryos and dissociated from its underlying stroma by treatment with saturated EDTA in a Ca2+ and Mg2+ free medium. The dorsal skin at E7 was surgically removed, and the dermis and epidermis were separated by means of protease treatment (1.25% Trypsin+2% Pancreatin). The epithelium and mesenchyme were reassociated on agar medium for 1 hr and were then grafted on the chick chorioallantoic membrane at E11, for short-term grafts (3 to 6 days), or under the kidney capsule of athymic nude mice for long-term grafting. The recombinants were recovered sequentially to follow Pax6 expression as well as after 2 weeks to allow full epithelial differentiation.

Insertion of Dorsal Dermis Under the Corneal Ectoderm After Removal of the Lens Vesicle

At the stage where the lens vesicle has formed (E3; HH 20), the first migration of the neural crest mesenchyme has still not occurred, but the presumptive corneal epithelium is too small to be easily removed and used in recombination studies. In this case, the lens vesicle was removed through a dorsal incision of the ectoderm at the level of the future iris and a 1 mm2 piece of E7 dorsal dermis, obtained as previously described, was inserted under the presumptive corneal epithelium. The embryos were recovered at time points ranging from a few hours to 12 days after grafting.

Removal of the Lens and Pre-lens Tissue

Lens vesicle and lens pit

Embryos were very lightly stained for 2 min with a piece of agar impregnated with 0.01% Neutral red. Lens tissue was surgically removed from the right eye with sharpened tungsten needles at HH stage 14 (when the lens placode invaginates) and at HH stage 15 (when the lens vesicle is detaching from the ectoderm). In each case, four embryos were recovered at time 0 to confirm correct surgical ablation by histology, while in the remaining cases the eggs were resealed and the operated embryos allowed to develop, then recovered and fixed at time points ranging from a few hours to 13 days after surgery.

Lens placodal ectoderm

To visualize the ectoderm and to facilitate the excision of the placode, we used of very low concentrations of Nile blue sulfate (0.05%), and just touched the ectoderm above the right optic vesicle of early HH13 embryos (18 somites) with a thin capillary. However, in our hands, this in ovo technique almost invariably leads to the death of the embryo within 4 hr of application with only a few surviving up to 30 hr after surgery. We thus modified a previous in vitro technique (Chapman et al., 2001), that allowed us to remove all of the remaining Nile blue dye after surgery. In this case, a ring of filter paper was placed on the yolk such that the embryo with its membranes up to 4 mm outside the sinus terminalis was in the central aperture. The entire blastoderm was maintained under tension by the ring of filter paper, removed and placed in a black dissecting dish to allow both the counting of the somites and microsurgery after using the Nile blue on the eye ectoderm. The blastoderm plus the paper ring were rinsed, and then explanted dorsal side up in a center well organ culture dish containing a semi-solid medium. All of the operated embryos explanted at HH early stage 13 (18 somites) survived and developed until HH stage 16–19. For the in ovo series, embryos were recovered at time zero and, those few that survived longer than a few hours were recovered before their death, from 10 to 30 hr after surgery. For the in vitro series, embryos were recovered at time 0 (t0) and then at t0+1 to 6 hr to follow the ectoderm wound healing, then from t0+17 to 38 hr to follow eye morphogenesis.

Production of Drm-Expressing Retrovirus and In Ovo Electroporation

Replication-competent RCAS BP(A) retrovirus encoding murine drm (RCAS/mdrm) was obtained by inserting a 1,202-bp fragment containing the coding region and 5′ and 3′ untranslated sequences of mouse drm cDNA (a gift from Dr. Maria Marx) into the RCAS(BP)A retroviral vector. Retroviral stocks were prepared as described previously (Morgan and Fekete, 1996). For electroporation, plasmid solution at 1.5 μg/μl was mixed with a 1/3 volume of 0.5% Fast Green dye (Sigma) just before use. A total of 1 μl of the DNA solution was injected across the vitelline membrane on the right side of the head of HH stage 10 embryos around the ectoderm and underneath the optic vesicle. A 25V 50-msec square pulse was applied eight times. Eggs were sealed with tape and returned to the incubator. Embryos were harvested at day 2–15 (E2–E15).

Immunofluorescence and Histology

For routine histology, specimens were processed and stained with hematoxylin. Immunohistochemistry was performed on cryosections (7 μm) according to standard protocols (Harlow and Lane, 1988). Primary antibodies were visualized using a secondary antibody against the appropriate species labeled with Alexa Fluor 488 or Alexa Fluor 548 (Molecular Probes). We used the AK12 monoclonal antibody, which is monospecific for the acidic corneal-type keratin K12, and the αK (11E1O) monoclonal antibody that recognizes several keratins, mainly K3, K12, K14, and K18 (both produced in D.D's laboratory, Chaloin-Dufau et al., 1990). Nuclei were counterstained with DAPI, and the slides mounted in Molwiol 4.88 anti-fading medium (Calbiochem.). Some cryo-sections were stained with hematoxylin.


The authors thank Prof. M. Mouillon, former head of ophthalmology at Grenoble Hospital for his continuous support, two unknown reviewers for suggestions, Prof. A. Streit for critical discussion and suggestions, Dr. Bertrand Pain for experimental suggestions, Dr. J. P. Viallet for teaching in ovo electroporation to EC, Mrs Brigitte Peyrusse for iconography, Dr. P. Kaur, Miss K-A. Bright, and Mrs Shonah Wraith for manuscript reading. Y.Y. received a Vision Research fellowship, E.C. received a AFM fellowship, and S.C. received a FRM fellowship.