NC-cell-specific constitutive expression of β-galactosidase in transgenic mice allows monitoring of NC-cell migration and fate during development in vivo 914. This approach was used in the present study to define the ocular structures originating from the NC. Rosa26 Cre reporter (Rosa26R) mice, which express β-galactosidase following Cre-mediated recombination, were mated with transgenic mice expressing Cre recombinase under the control of the Wnt1 promoter. Although Wnt1 is not expressed in any structure of the developing eye (see Additional data file 1), it is expressed in the dorsal neural tube, allowing Wnt1-Cre-mediated recombination in virtually all NC stem cells 1115. In Wnt1-Cre/Rosa26R double transgenic mice, β-galactosidase-expressing NC-derived cells can be visualized by X-gal staining.

NC-derived cells have previously been proposed to contribute to ocular development in mice after embryonic day (E)12 10. Interestingly, we found that NC-derived cells were already detectable at E10 surrounding the optic cup and the lens vesicle (Figure 1f). Until E13.5 (Figure 1f–j), the NC-derived cells were found predominantly in the periocular mesenchyme, whereas the overlying epithelium, the lens, and the retina were consistently X-gal-negative. In addition, we observed that structures of the primary vitreous, located between the lens and the retina, are NC-derived (Figure 1f–j). At later stages (Figure 1d,e), X-gal-positive cells contributed to corneal stroma and endothelium and to structures of the chamber angle at the junction between the cornea and the iris. In mature eyes, the stroma of the iris, the ciliary body, and the trabecular meshwork, as well as cells of the choroid and primary vitreous, are all of NC origin (data not shown). Taken together, these results show that NC-derived cells contribute to eye development as soon as the eye vesicle is formed and, subsequently, to various structures of the maturing eye.

The expression pattern of TGFβ ligands and their receptors during eye development was visualized by immunohistochemistry at various developmental stages (E10.5 to E18), showing that TGFβ2 expression peaked in the forming lens at E13.5 (Figure 2a) and E15, but decreased towards E18 (data not shown), whereas TGFβ1 and TGFβ3 were undetectable (Additional data file 2 and data not shown). At E13.5, TGFβ receptor type 2 (Tgfbr2) was expressed in periocular mesenchyme, lens, retina, and the primary vitreous (Figure 2b). Because in vivo fate mapping revealed a substantial contribution of the NC to the periocular mesenchyme, TGFβ signaling could be important for development of ocular NC derivatives. We therefore analyzed the eyes of mouse embryos after NC-specific inactivation of TGFβ signaling 1116. Tissue-specific signal inactivation was achieved by Wnt1-Cre-mediated deletion of exon 4 of the Tgfbr2 gene (Figure 2c), which leads to loss of Tgfbr2 protein expression in NC stem cells 11. In such Tgfbr2-mutant mice, both Tgfbr2 expression (Figure 2d,f) and TGFβ-induced phosphorylation of the downstream signaling molecule Smad2 (pSmad2; Figure 2e,g) remained undetectable in the periocular mesenchyme.

At E18, main structures of the anterior eye segment, including the forming ciliary body, the iris and the trabecular meshwork, were all well defined in control animals; eye development in the absence of TGFβ signaling in NC-derived cells was therefore analyzed first at E18. Most impressively, eyes from Tgfbr2-mutant embryos were 26 ± 1% smaller than eyes from control littermates (Figures 3a, 4). The cornea in control eyes was properly structured into epithelium and endothelium covering a thick stroma, but in Tgfbr2-mutant mice the cornea lacked an endothelial layer and no normal stroma was formed (Figure 3b). In control mice, corneal structures and the lens were clearly separated to form the anterior eye chamber; in contrast, cornea and lens of Tgfbr2-mutant eyes failed to separate, and no proper anterior eye segment was built (Figure 3c). Moreover, normal formation of the trabecular meshwork and the ciliary body, indicated by a wrinkle in the iris primordium in control eyes, was not observed in Tgfbr2-mutant eyes (Figure 3c). In addition, eye sections from E18 Tgfbr2-mutant embryos revealed a remarkable accumulation of cells between lens and retina, whereas vessels of the hyaloid vascular system were present in corresponding structures of control eyes (Figure 3d). Finally, the retina of control mice was clearly structured into an inner and an outer layer of cells, whereas the retina of Tgfbr2-mutant mice showed diffuse patterning (Figure 3e). Thus, Tgfbr2-mutant embryos show microphthalmic eyes with anomalies of the anterior segment, similar to those seen in human Axenfeld-Rieger's anomaly, and the embryos also had defects of the posterior eye segment.

In normal mice, the primary vitreous, including the hyaloid vascular system, persists until postnatal day (P)30. Its regression starts postnatally around P14 to form the avascular and transparent secondary vitreous 17. In patients with congenital persistent hyperplastic primary vitreous, developmentally abnormal primary vitreous becomes a fibro-vascular membrane, formed behind the lens (retrolentally) 18. Much as in human persistent hyperplastic primary vitreous 19, irregular retrolental structures present in Tgfbr2-mutant mice consisted of several different cell types (Figure 5a–e). These included fibroblast-like cells, prospective melanocytes expressing dopachrome tautomerase mRNA (Dct; also called Trp-2; Figure 5c), smooth muscle α-actin-positive pericytes (Figure 5b), and vessels of the hyaloid vascular system (Figure 5e). Moreover, staining with an antibody to Ki-67, a protein expressed only in dividing cells, revealed proliferative cells in the retrolental tissue (Figure 5d).

Effects on the retina have been reported in patients with persistent hyperplastic primary vitreous 20. Moreover, as instructive signals from the lens promote normal patterning of the retina 21, the irregular retrolental structures in Tgfbr2-mutant mice might alter normal interaction between the lens and the retina. To test whether retinal development in Tgfbr2-mutant mice was affected, retinas from embryos of different ages were immunohistochemically stained for factors known to be expressed at distinct stages of development 22. At E15, the inner parts of the retina from control mice strongly expressed the transcription factors Brn3A in retinal ganglion cells and Pax6 in amacrine cells of the ganglion cell layers; in contrast, Tgfbr2-mutant embryos had lower numbers of both Brn3A- and Pax6-positive retinal cells (Figure 5f,g). Moreover, at E15 the number of cells positive in the TUNEL-staining procedure, which detects apop-totic cells, was higher in the retinas of Tgfbr2-mutant embryos than in those of control embryos (13.3 ± 2.5/5 μm section (mutant) versus 5.6 ± 0.5 (control); p < 0.01; not shown). At E18, expression of Brn3A, Pax6 and neurofilaments defines distinct layers of the developing retina in control eyes (Figure 5h). In Tgfbr2-mutant mice, however, patterning into cell layers was disturbed, and the thickness of the retina was increased in the mutants (Figures 4, 5h). Eyes of Tgfbr2-mutant mice are therefore affected by anomalies similar to persistent hyperplastic primary vitreous and by disturbed retinal patterning.

Anterior eye segment anomalies in Tgfbr2-mutant mice were reminiscent of human Axenfeld-Rieger's anomaly (Figure 3). In vivo fate mapping revealed that migration of TGFβ-dependent NC cells to the corneal stroma, the endothelium, and the trabecular meshwork was unaffected in Tgfbr2-mutant mice (Figure 6a). This indicates that the ocular malformations arise from impaired differentiation rather than from NC-cell migration defects. Interestingly, the anomalies observed in the Tgfbr2-mutant embryos recapitulate aspects of ocular defects found in Foxc1-null or Pitx2-null mice 23. Loss of TGFβ responsiveness in the cells of the periocular mesenchyme might therefore affect expression of the transcription factors Foxc1 and Pitx2. To test this hypothesis, we analyzed eyes from Tgfbr2-mutant and control embryos at different developmental stages for the presence of Foxc1 and Pitx2. We confirmed previous reports 223 that the two factors are expressed in the periocular mesenchyme during early development (Figure 6b and data not shown); at E15, however, Foxc1 localizes to the corneal endothelium and structures of the forming trabecular meshwork (Figure 6d), and Pitx2 to the corneal stroma (Figure 7a). In contrast, in eyes of Tgfbr2-mutant embryos Foxc1 was hardly detectable in the periocular mesenchyme at E13.5 and in the forming chamber angle and corneal endothelium at E15. Moreover, Tgfbr2-mutant cells that failed to express Foxc1 appeared to subsequently undergo apoptosis around E15, as revealed by TUNEL staining (Figure 6e).

Pitx2 was strongly expressed in the corneal stroma at E15 in control eyes, but was undetectable in the eyes of Tgfbr2-mutant embryos (Figure 7a). Interestingly, some Tgfbr2-mutant cells of the corneal stroma expressed Dct rather than Pitx2, pointing to incorrect fate acquisition towards melanocytes or misguidance during migration (Figure 7b). At E18, the corneal stroma of control embryos consisted of thin keratocytes organized in a lamellar structure and embedded in extracellular matrix, which provides corneal stability and transparency (Figure 7c). High levels of collagen were detectable in the corneal stroma of control mice, whereas collagen staining was negative in the malformed cornea of E18 Tgfbr2-mutant mice, and stromal cells had an abnormal polygonal shape (Figure 7c,d). In summary, NC-derived ocular cells that lack responsiveness to TGFβ fail to express Foxc1 and Pitx2 and fail to undergo correct differentiation into corneal endothelial cells and collagen-synthesizing keratocytes of the corneal stroma.

The absence of Foxc1 and Pitx2 expression in the developing eyes of Tgfbr2-mutant mice raises the question of whether TGFβ signaling can regulate the expression of Foxc1 and/or Pitx2. To address this issue, cultured rat embryonic fibroblasts were treated with TGFβ and analyzed by western blot for the presence of Foxc1 and Pitx2 (Figure 8a). In the absence of TGFβ, the cells showed weak expression of Foxc1, and Pitx2 expression was undetectable. TGFβ treatment, however, strongly increased Foxc1 expression and induced Pitx2 expression, concomitant with increased levels of pSmad2 (Figure 8a). In addition to fibroblasts, postmigratory NC-derived cells of mouse periocular mesenchyme were also responsive to TGFβ, as shown in short-term tissue cultures of eyes from E11 embryos (Figure 8b): again, treatment with TGFβ resulted in elevated Foxc1 expression. Moreover, Pitx2 expression, which was undetectable in untreated samples, was induced upon addition of TGFβ. In summary, TGFβ treatment upregulates both Foxc1 and Pitx2 expression in a fibroblast cell line and in embryonic eye tissue cultures. TGFβ signaling is therefore not only required for the expression of transcription factors associated with developmental eye disorders, but it is also sufficient to regulate their expression.

The primary vitreous is situated directly behind the lens and contains the hyaloid vascular system beneath NC-derived cells. Normally, the primary vitreous regresses during postnatal eye maturation through tissue remodeling by apoptosis and phagocytosis, thereby generating the avascular, transparent secondary vitreous 17. In patients suffering from persistent hyperplastic primary vitreous, a dense cell membrane persists between the lens and the retina. This congenital disorder is often accompanied by cataracts, secondary glaucoma, and a variable degree of microphthalmia 1828. Similarly, the primary vitreous in the eyes of Tgfbr2-mutant mice appears as a dense cellular membrane, and mutant eyes are smaller than those in control mice. Much as in human persistent hyperplastic primary vitreous 19, the persistent retrolental cell mass in Tgfbr2-mutant mice contains fibroblast-like cells, pigmented cells, and vessels of the hyaloid vascular system, and proliferating cells are also seen.

Other mouse mutants have been reported to have a phenotype similar to persistent hyperplastic primary vitreous, including those mutant for the Arf1, Bmp4, or p53 genes 293031. In these models, normal postnatal regression of the primary vitreous fails, resulting in a variable degree of anomalies reminiscent of persistent hyperplastic primary vitreous. Similarly, a dense cell mass in the posterior eye has also been observed previously in Tgfb2 null mice, but this was not analyzed further 12. Treatment of pregnant mice with retinoic acid, which is known to interfere with TGFβ signaling 32, induces anomalies similar to persistent hyperplastic primary vitreous in the offspring 33. Thus, we conclude that TGFβ signaling in NC-derived cells constituting the primary vitreous is important for tissue morphogenesis.

In the posterior eye segment, retinal development is also disturbed upon ablation of Tgfbr2 in NC cells, separately from the generation of persistent hyperplastic primary vitreous. In particular, we observed increased retinal apoptosis at E15 and abrogated retinal patterning, as shown by histology and layer-specific tissue marker expression (Figure 5f–h). Because there is no NC contribution to the retina, this phenotype is probably due to a secondary, non-cell-autonomous effect. The dense persistent primary vitreous in Tgfbr2-mutant mice might conceivably constrain instructive signals from the lens to the retina, but such putative signals remain to be identified.

In addition to the defects reminiscent of persistent hyperplastic primary vitreous, all Tgfbr2-mutant mice have several developmental defects in the anterior eye. The anterior chamber of the eye is absent in the mutant because the cornea and the lens fail to separate. Furthermore, normal formation of the ciliary body and of the chamber angle with the trabecular meshwork requires TGFβ signaling, as these structures are defective in the mutant mice. The abnormalities presented by Tgfbr2-mutant mice are characteristic of the disorders found in patients with Axenfeld-Rieger's anomaly 10. In this disorder, anterior segment dysgenesis impairs the regulation of the intraocular pressure, which frequently leads to developmental glaucoma.

Other mouse mutants have also been implicated as models for developmental anterior eye disorders. Mice homozygous for an inactivating mutation of Pax6, a candidate for human Peter's anomaly, lack eyes 7. Heterozygous Pax6^+/- mice have defects in the anterior eye segment, although less severe than those found in Tgfbr2-mutant mice 3435. The expression of Pax6 in eyes of Tgfbr2-mutant mice is not affected, however (data not shown), suggesting that their defects do not depend on Pax6 modulation. In human Axenfeld-Rieger's anomaly, mutations have been found in the genes encoding the transcription factors FOXC1 and PITX2 1. Deletion of either Foxc1 or Pitx2 in mice 23 leads to defects in the anterior eye segment, very similar to those in Tgfbr2-mutant mice described in this study. In the eye, Foxc1 is expressed in the forming corneal stroma and endothelium and, at later stages, in the structures of the prospective trabecular meshwork 2. Intriguingly, these structures express Foxc1 in a TGFβ signal-dependent manner, and Tgfbr2-mutant prospective corneal endothelial and trabecular meshwork cells undergo apoptosis that is not observed in control eyes. Furthermore, TGFβ upregulates Foxc1 expression in fibroblasts and cultured eye tissue, in agreement with a previous report that described Foxc1 as a target gene of TGFβ in human cancer-cell lines 36. Thus, the data suggest that lens-derived TGFβ signaling controls the survival and development of the NC-derived periocular mesenchyme that gives rise to corneal endothelium and trabecular meshwork by regulating Foxc1 expression in these cells (Figure 9).

Pitx2 is expressed predominantly in NC-derived corneal stromal cells that become collagen-synthesizing keratocytes. In Tgfbr2-mutant mice, however, corneal stromal cells do not express Pitx2 and consequently fail to develop into collagen-synthesizing keratocytes. Recently, mutations in the human TGFBR2 gene have been reported to cause Marfan's syndrome, a disorder also associated with defective extracellular-matrix synthesis 37. Thus, we conclude that corneal NC-derived cells must have TGFβ-dependent expression of Pitx2 and differentiation to become stromal keratocytes that produce the collagen matrix (Figure 9). In support of this hypothesis, Pitx2 expression is strongly induced in fibroblasts and eye tissue upon TGFβ signal activation.

In Axenfeld-Rieger's anomaly patients who have a disease-linked mutation in the PITX2 gene, ocular anomalies appear to be accompanied by additional defects, including tooth abnormalities, redundant periumbilical skin, and heart defects (all together referred to as Rieger's syndrome) 1. Apart from its expression in NC-derived cells of the forming eye, Pitx2 is expressed in several other tissues during development, including the teeth, umbilicus, and the heart 23. In contrast to the mesenchymal expression pattern in the eye, in other organs the expression of Pitx2 is restricted to structures that are not NC-derived, but these structures, and especially the tooth anlagen, are surrounded by or are in close contact with NC-derived cells 14. Nevertheless, Tgfbr2-mutant embryos show no defects in the tooth anlagen or umbilicus at E18 (data not shown). Therefore, Pitx2-dependent anomalies in Tgfbr2-mutant mice appear to be restricted to the eyes, although because of embryonic lethality we could not determine whether there are additional Pitx2-dependent defects at a developmental stage later than E19.

We recently reported that inactivation of TGFβ signaling in NC stem cells also leads to cardiac and craniofacial defects and parathyroid and thymic gland anomalies reminiscent of human DiGeorge syndrome 11. Moreover, depending on the cellular context, TGFβ promotes non-neural cell fates in cultured NC cells 3839. Hence, together with the findings from the present study, there is good evidence that TGFβ is a key modulator of non-neural differentiation of post-migratory NC cells during development of multiple tissues, including the eye.

The generation of mice used in this study has been described before 9111641. Briefly, loxP-sites for Cre-mediated recombination were introduced into the mouse Tgfbr2 locus flanking exon 4, which encodes the transmembrane domain and is an important part of the functional intracellular domain of the Tgfbr2 protein. Mice expressing the Cre recombinase under the control of the Wnt1 promoter and heterozygous for this Tgfbr2 'floxed' allele were mated with mice homozygous for the floxed allele. Inactivation of TGFβ signaling in NC-derived cells was achieved in embryos inheriting Wnt1-Cre and two Tgfbr2 floxed alleles 11. 100% of all mutant embryos had the phenotype described in this study, as assessed by the analysis of at least three embryos per stage and staining condition. In contrast, littermates lacking the Wnt1-Cre transgene or carrying a wild-type Tgfbr2 allele expressed Tgfbr2 normally and did not exhibit any overt phenotype, thus serving as control animals. Genotyping was performed as described 11. All animal experiments were performed on the C57BL/6 background, which has never been associated with genetic mutations causing retinal degeneration.

The Rosa26 reporter (Rosa26R) mouse strain expresses β-galactosidase following Cre-mediated recombination 41. To define the distinct contribution of the NC during ocular development, Rosa26R mice were crossed with Wnt1-Cre transgenic mice 9. At least three whole embryos per stage were stained with the β-galactosidase substrate 5-bromo-4-chloro-3-indolyl-β-D-galactopyranoside (X-gal; Sigma, Buchs, Switzerland) and subsequently fixed in 4% paraformaldehyde overnight at 4°C. Subsequently, embryos were embedded in paraffin, sectioned at 7 μm, and dewaxed for mounting with DFX (Fluka, Buchs, Switzerland). Some sections were counterstained with eosin (Fluka).

Embedding, sectioning and staining procedures were performed as described 11. Briefly, the eyes of at least three embryos per stage were stained with primary antibodies to TGFβ1, TGFβ2, and TGFβ3 (Santa Cruz Biotechnology Inc., Santa Cruz, USA), Tgfbr2 (Santa Cruz), pSmad2 (Cell Signaling Technology Inc., Beverly, USA), Ki-67 (Lab Vision (UK) Ltd, Newmarket, UK), Brn3A 42, Pax6 (Chemicon International Inc., Temecula, USA), neurofilament (Chemi-con), Foxc1 (Santa Cruz), Pitx2 23, and GFAP (Sigma). For visualization the ABC elite Kit (Vector Laboratories Inc., Burlingame, USA) with Metal enhanced DAB (Pierce Biotechnology Inc., Rockford, USA) or AP substrate kit I (Vector) as substrates was used. In situ hybridization with digoxigenin-labeled riboprobes to Dct was performed as described 915. TUNEL assays were performed following the manufacturer's guidelines (Roche Diagnostics, Basel, Switzerland). Standard protocols were used for tissue processing of semi-thin sections and subsequent toluidine blue staining 43. The Van Gieson's staining procedure was used to visualize collagen formation in the cornea.

At least three Tgfbr2-mutant and control embryos per stage were embedded and sectioned. Mid-organ sagittal sections of both eyes were measured using an Eclipse E600 microscope (Nikon, Tokyo, Japan) equipped with a CCD camera (Kappa, Gleichen, Germany) and the PicEd Cora software version 8.08 (JOMESA, Munich, Germany).

Rat embryonic fibroblasts (rat2 cell line; American Type Culture Collection, Mannassas, USA) were cultured in DMEM:F12 medium (Gibco/Invitrogen, Carlsbad, USA) containing 10% fetal bovine serum (Sigma). Following a 60 min incubation in DMEM:F12 medium containing 0.1% bovine serum albumin at 37°C, cells were treated with TGFβ (5 ng/ml) for 90 min at 37°C as described 11. For short-term tissue-culture experiments, the eyes with periocular tissue were removed from nine embryos at E11 by microdissection. Left and right eyes were pooled separately and kept in DMEM:F12 medium containing 0.1% bovine serum albumin and antibiotics with and without TGFβ (5 ng/ml), respectively, for 6 h at 37°C. Western blot analysis of rat embryonic fibroblast extracts and mouse eye tissue extracts were carried out as described 44. Primary antibodies used were against Actin (Chemicon), pSmad2 (Cell Signaling), Foxc1 (Santa Cruz) and Pitx2 23. Each experiment was performed at least three times.

Results are shown as mean ± standard error of the mean (S.E.M.). Graphs and statistical analyses used Prism 4.01 (GraphPad Software Inc., San Diego, USA).